Вы находитесь на странице: 1из 108

Progress in Surface Science 79 (2005) 47154

www.elsevier.com/locate/progsurf

Review

The surface and materials science of tin oxide


Matthias Batzill *, Ulrike Diebold
Department of Physics, Tulane University, 6400 Freret St., Stern Hall 2001, New Orleans,
LA 70118, United States

Abstract
The study of tin oxide is motivated by its applications as a solid state gas sensor material, oxidation catalyst, and transparent conductor. This review describes the physical and chemical properties
that make tin oxide a suitable material for these purposes. The emphasis is on surface science studies
of single crystal surfaces, but selected studies on powder and polycrystalline lms are also incorporated in order to provide connecting points between surface science studies with the broader eld of
materials science of tin oxide. The key for understanding many aspects of SnO2 surface properties is
the dual valency of Sn. The dual valency facilitates a reversible transformation of the surface composition from stoichiometric surfaces with Sn4+ surface cations into a reduced surface with Sn2+ surface cations depending on the oxygen chemical potential of the system. Reduction of the surface
modies the surface electronic structure by formation of Sn 5s derived surface states that lie deep
within the band gap and also cause a lowering of the work function. The gas sensing mechanism
appears, however, only to be indirectly inuenced by the surface composition of SnO2. Critical
for triggering a gas response are not the lattice oxygen concentration but chemisorbed (or ionosorbed) oxygen and other molecules with a net electric charge. Band bending induced by charged
molecules cause the increase or decrease in surface conductivity responsible for the gas response signal. In most applications tin oxide is modied by additives to either increase the charge carrier concentration by donor atoms, or to increase the gas sensitivity or the catalytic activity by metal
additives. Some of the basic concepts by which additives modify the gas sensing and catalytic properties of SnO2 are discussed and the few surface science studies of doped SnO2 are reviewed. Epitaxial SnO2 lms may facilitate the surface science studies of doped lms in the future. To this end lm
growth on titania, alumina, and Pt(1 1 1) is reviewed. Thin lms on alumina also make promising test
systems for probing gas sensing behavior. Molecular adsorption and reaction studies on SnO2 surfaces have been hampered by the challenges of preparing well-characterized surfaces. Nevertheless
some experimental and theoretical studies have been performed and are reviewed. Of particular
*

Corresponding author. Tel.: +1 504 862 3187.


E-mail address: mbatzill@tulane.edu (M. Batzill).

0079-6816/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.progsurf.2005.09.002

48

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

interest in these studies was the inuence of the surface composition on its chemical properties.
Finally, the variety of recently synthesized tin oxide nanoscopic materials is summarized.
 2005 Elsevier Ltd. All rights reserved.
Keywords: Tin oxide; Transparent conductor; Catalysis; Gas sensor; Single crystalline surfaces; Surface
structure; Epitaxy; Surface chemistry; Nanostructured oxides; Chemisorption

Contents
1.
2.

3.

4.

5.

6.

7.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Applications of tin oxide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1. Transparent conductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2. Heterogeneous catalysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3. Solid state gas sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1. Pure metal oxide gas sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2. The role of additives in gas sensing materials. . . . . . . . . . . . . . . . . . . .
Comparison between stannic and stannous oxide . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1. Crystal structure of stannic and stannous oxide . . . . . . . . . . . . . . . . . . . . . . . .
3.2. Phonon modes of stannic and stannous oxide . . . . . . . . . . . . . . . . . . . . . . . . .
3.3. Auger- and core level spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Surface composition and structure of SnO2 low index surfaces . . . . . . . . . . . . . . . . . .
4.1. Experimental studies of the SnO2(1 1 0) surface. . . . . . . . . . . . . . . . . . . . . . . . .
4.2. Oxygen chemical potential dependent properties of SnO2(1 1 0), (1 0 0), and (1 0 1)
low index surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1. Surface structure and composition I: Experimental. . . . . . . . . . . . . . . .
4.2.2. Surface structure and composition II: Theoretical . . . . . . . . . . . . . . . .
4.2.3. Surface structure and composition III: comparison of experimentally
and theoretically determined phase transitions . . . . . . . . . . . . . . . . . . .
4.3. Conclusions for surface structure and composition . . . . . . . . . . . . . . . . . . . . . .
Electronic structure of tin oxide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1. Comparison between the bulk electronic structure of SnO and SnO2 . . . . . . . . .
5.2. Surface electronic structure of SnO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1. SnO2(1 1 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.2. SnO2(1 0 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.3. SnO2(1 0 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.4. Homogeneity of surface electronic structure. . . . . . . . . . . . . . . . . . . . .
5.3. Conclusions for the surface electronic structure of SnO2 . . . . . . . . . . . . . . . . . .
Epitaxial SnO2 films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1. SnO2 on TiO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2. SnO2 on Al2O3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3. SnO2 on Pt(1 1 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4. Conclusions and outlook for SnO2 epitaxy . . . . . . . . . . . . . . . . . . . . . . . . . . .
Additives in SnO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1. Clusters and ad-layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1.1. Tin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1.2. Palladium. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1.3. Copper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1.4. Vanadium and vanadia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1.5. Conclusions and outlook for surface dopants . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.

49
51
51
57
59
60
65
66
68
69
71
72
74

. 79
. 81
. 86
. 89
. 90
. 91
. 91
. 93
. 93
1 00
104
105
106
108
108
109
111
112
113
113
113
114
116
116
117

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

7.2.

Lattice dopants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.1. Antimony-doped SnO2 . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.2. Cobalt-doped SnO2 . . . . . . . . . . . . . . . . . . . . . . . . . . .
8. Surface chemistry of SnO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.1. General remarks on molecular reactions at SnO2 surfaces . . . . . . .
8.2. Adsorption and reactions of specific molecules . . . . . . . . . . . . . .
8.2.1. Oxygen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.2. Carbon monoxide CO . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.3. Carbon dioxide CO2 . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.4. Nitrogen dioxide NO2 . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.5. Ammonia NH3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.6. Methanol (CH3OH) . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.7. BF3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.8. Formic acid (HCOOH) . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.9. Benzene (C6H6) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.10. Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9. Nanomaterials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10. Synopsis and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

49

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

118
118
120
120
121
122
122
124
125
126
126
126
127
127
128
128
132
136
139
139

1. Introduction
The wide variety of electronic and chemical properties of metal oxides makes them
exciting materials for basic research and for technological applications alike. Oxides span
a wide range of electrical properties from wide band-gap insulators to metallic and superconducting. Tin dioxide belongs to a class of materials that combines high electrical conductivity with optical transparency and thus constitutes an important component for
optoelectronic applications.
In chemical applications oxides are used as support materials for dispersed metal catalysts but also often exhibit catalytical activity by themselves. Furthermore, it becomes
increasingly apparent that the active phase of some metal oxidation catalysts is in fact their
oxides rather than the pure metal. For CO oxidation for instance, it has been shown that
the oxides of Ag [1], Pd [2], and Ru [3] are the active phase under operating conditions.
The catalytic cycle on oxides may include the reaction of lattice oxygen from the catalyst
surface with the reactant and replacement of the lattice oxygen by gas phase oxygen at the
end of the catalytic cycle. This is the so-called Marsvan Krevelen mechanism [4]. For this
mechanism to be viable the surface oxygen concentration of the catalyst has to be readily
adjustable. This is also the case for SnO2 and may explain its suitability as an oxidation
catalyst.
Another eld in which oxides play a dominant role is in solid state gas sensors. A wide
variety of oxides exhibit sensitivity towards oxidizing and reducing gases by a variation of
their electrical properties, but SnO2 was one of the rst considered, and still is the most
frequently used, material for these applications. There is an obvious close relationship
between the gas sensitivity of oxides and their surface chemical activity and thus gas
sensing applications and catalytic properties should be considered jointly.

50

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

The great diversity of oxide materials could not be better demonstrated than in the variety of self-assembled nanoscale materials that have been recently discovered. For tin oxide,
for instance SnO nanodiskettes [5], SnO2 nanobelts, and other nanoscopic materials [6]
form. SnO2 nanobelts have the bulk-like rutile structure but are only of the order of
10 100 nm in the cross-section and up to several millimeters in length. Their surfaces
are low index bulk terminations and thus should exhibit similar properties to single crystal
surfaces. For growing many of these tin oxide nanostructures a simple vapor phase transport growth is employed. Tin oxide is not unique in its self-formation of nanostructured
materials other oxides also grow nanowires (ZnO [79], In2O3 [10], SiO2 [1113], Ga2O3
[1416], and GeO2[17]), nanobelts (ZnO, SnO2, Ga2O3, In2O3, CdO, and PbO2 [1821]),
and nanorods (MgO [22]). Many of these materials have gas sensing properties. This
and their large surface to volume ratio make them promising materials for well dened,
highly sensitive gas sensors.
One aspect of surface science is to elucidate the interaction of molecules with solid surfaces by studying well-dened systems. As it becomes more apparent that oxides often play
an intricate role in chemical processes, surface science studies of well-ordered oxide lms
or bulk-oxide surfaces are becoming more popular [2334]. Recent surface science studies
of metal oxides and thin oxide layers on metal substrates, however, indicate that these surfaces can be rather complex [35,36]. For the same oxide many dierent surface structures
and compositions exist and the surface phase one obtains depends largely on the preparation conditions in surface science experiments. Consequently, one can project that also
during catalysis and gas sensing applications dierent compositions and structures of
the oxide are present depending on the operation conditions. Not all oxide phases may
be equally active and thus by tuning the operation conditions to favor one phase over
the other better activities, selectivities, or gas sensing sensitivities may be obtained.
For the large majority of surface science studies the necessity of a well-dened environment requires ultrahigh vacuum conditions. This causes a quite reducing environment and
oxides may exhibit dierent surface compositions and structures than those encountered at
elevated pressures. For oxides, this is the quintessence of the often discussed pressure gap
between fundamental surface science studies and realistic applications. In this article we
discuss the inuence of the pressure on the composition and structure of oxides on the
example of SnO2. Tin oxide is special in the respect that tin possesses a dual valency, with
tin preferably attaining an oxidation state of 2+ or 4+. This dual valency facilitates a variation of the surface oxygen composition. A variation of the sample composition has been
observed for other oxides of multivalent metals. For example studies on vanadium oxide
[36] and iron oxide [35] have shown that surfaces with dierent compositions can be prepared. Thus the variability of surface compositions is a general scheme for many metal
oxides. This property of oxides is, however, not just a nuisance but may be instrumental
in explaining and tailoring many of the unique chemical properties of these materials.
In this article we attempt to connect applied research that utilizes SnO2 as a transparent
conductor, catalyst, and gas sensor with fundamental surface science studies. In most
applications SnO2 in its pure form is rarely used but is usually modied by impurity
dopants and other additives. Understanding the inuence of additives on the properties
of oxides is a challenge for surface scientists that need to be addressed if one wants to gain
a better understanding of materials used in applications.
The investigation of the surface science of SnO2 is still relatively unexplored compared
to for instance TiO2 [34]. This is mainly due to the lack of commercially available single

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

51

crystals or established methods for growing well-dened thin lms. Suitable SnO2 single
crystals were grown in the past. New elds like dilute ferromagnetic oxides (briey discussed in Section 7.2.2) as well as new approaches in the study of traditional applications
of SnO2 may sporn again the growth of suitable single crystals for surface science investigations [37]. It is the intend of this article to increase the interest in the study of SnO2
within the surface science community and to show avenues for possible future
investigations.
This article is organized as follows: First we discuss some fundamental aspects of the
applications in which SnO2 nds use. Then the structure of SnO2 and SnO are described
followed by a description of the surface structure and composition of low index SnO2 surfaces as a function of the oxidation potential of the gas phase. The dierences of the surface electronic structure for reduced SnO2 surfaces compared to stoichiometric surfaces
are discussed in Section 5. Systems for the growth of thin epitaxial SnO2 lms are summarized in Section 6 and the inuence of additives on SnO2 properties are discussed in Section
7. Studies of the surface chemistry are presented in Section 8. A summary of the rich variety of tin oxide nanomaterials investigated in recent years is given in Section 9. We nish
by giving an outlook on possible future research directions.
2. Applications of tin oxide
The surface and materials properties of SnO2 (and impurity doped SnO2) should be discussed in context of its three major applications. These applications are (i) as a transparent
conducting oxide (TCO), (ii) as an oxidation catalyst, and (iii) as a solid state gas sensing
material. For the latter two applications the surface of the material is where the action is
and thus surface science investigations are of direct relevance. For the rst application it is
the bulk properties that are responsible for making SnO2 a TCO. However, many applications of TCOs require interfacing them with a dissimilar material. Thus the surface
and interface properties of SnO2 are also important in the use of SnO2 in TCO
applications.
In the next three subsections we try to give a brief overview of the current understanding of the key properties of SnO2 and related materials that make them suitable for these
applications. All three applications involve complex mechanisms that are topics of ongoing research. Thus this introduction cannot give a complete review; it is rather an attempt
to place surface science studies on SnO2 into a framework where it can benet from
research done in other elds and vice versa.
2.1. Transparent conductors
SnO2 belongs to the important family of oxide materials that combine low electrical
resistance with high optical transparency in the visible range of the electromagnetic spectrum. These properties are sought in a number of applications; notably as electrode materials in solar cells, light emitting diodes, at panel displays, and other optoelectronic
devices where an electric contact needs to be made without obstructing photons from
either entering or escaping the optical active area and in transparent electronics such as
transparent eld eect transistors [3841]. Another property of SnO2 and other TCOs is
that although they are transparent in the visible they are highly reective for infrared light.
This property is responsible for todays dominant use of SnO2 as an energy conserving

52

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

material. SnO2-coated architectural windows, for instance, allow transmitting light but
keeping the heat out or in the building depending on the climate region. More sophisticated architectural windows, so-called smart windows, rely on TCOs to electrically contact
electrochromic lms that are changing their coloring and transparency by applying a
voltage across the lms [4244].
There is a large number of TCOs, the most commonly known ones are the binary systems, i.e. SnO2, ZnO, In2O3 (better known as ITO if doped with Sn), Ga2O3, and CdO
[45,46]. A large variety of ternary and more complex TCO materials are being developed
[47,48] and continuous eorts are being made to nd p-type conducting TCOs [49] in addition to the above-mentioned n-type materials. For dierent applications dierent materials
may possess advantageous properties [50]. Hartnagel et al. give a review of properties and
preparation procedures for TCOs [51]. A summary of physical properties of SnO2 and two
other common TCO materials, i.e. ZnO and In2O3 is given in Table 1.
Often the highest possible conductivity is sought. Many of the binary TCOs already
possess a high conductivity due to intrinsic defects, i.e. oxygen deciencies. This is also
the case for SnO2, which, as a wide band-gap semiconductor, is in its stoichiometric form
a good insulator. Non-stoichiometry, in particular oxygen deciency, makes it a conducTable 1
Summary of physical properties of the transparent conducting oxides, In2O3, ZnO and SnO2 (adapted and
extended from Ref. [56])
Property

In2O3

ZnO

SnO2

Mineral name
Abundance of the metal
in the earths crust (ppm)
Crystal structure
Space group
Lattice constants [nm]

0.1

Zincite
132

Cassiterite
40

Cubic, bixbyite
I213
a = 1.012

Hexagonal, wurtzite
P63mc
a = 0.325
b = 0.5207
5.67
4
kc: 2.92
?c: 4.75
2240
420
10

Tetragonal, rutile
P42mnm
a = 0.474
b = 0.319
6.99
6.5
kc: 3.7
?c: 4.0
>1900a
232
5 109

9.7
3.75
9

3.6
3.4
kc: 8.75
?c: 7.8

6.0
3.6
kc: 9.6
?c: 13.5

0.3 [53]

0.34 [52]

kc: 0.58, 0.59


?c: 0.6, 0.59
[52,55]
B, Al, Ga, In,
Si, Ge, Sn, Y, Sc,
Ti, Zr, Hf, F, Cl

kc: 0.23
?c: 0.3 [54]
kc: 0.20
?c: 0.26 [52]

Density q [g cm3]
Mohs hardness [50]
Thermal expansion
coecient (300 K) [106 K1]
Melting point [C]
Melting point of metal [C]
Vapor pressure of metal
at 500 C [Torr]
Heat of formation [eV]
Band gap [eV]
Static dielectric constant er
(the complex dielectric functions
are calculated in Ref. [51]
Eective electron mass of conduction
electrons m*/m0 (experimental)
Eective electron mass of conduction
electrons m*/m0 (computational)
Common extrinsic n-type dopants

7.12
5
6.7
2190
157
106

Sn, Ti, Zr, F,


Cl, Sb, Ge, Zn, Pb, Si

Decomposition into SnO and O2 at 1500 C.

Sb, F, Cl

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

53

tor, however. Kilic and Zunger [57] showed that the formation energy of oxygen vacancies
and tin interstitials in SnO2 is very low and thus these defects form readily, explaining the
often observed high conductivity of pure, but non-stoichiometric, SnO2. In all applications
of these materials the charge carrier concentration and thus the conductivity is further
increased by extrinsic dopants. In the case of SnO2 these are commonly Sb as a cation dopant and F as an anion dopant [50]. Although these dopants increase the conductivity of
SnO2 signicantly it does not reach the low conductivity values achievable with ITO or
with doped ZnO. Fig. 1 shows the progress that has been made in reducing the resistivities
of these three materials over the last 30 years.
The transmission window of TCOs is between wavelengths of 0.4 lm and 1.5 lm. At
short wavelengths (high energies) electron-interband transitions from the valence to the
conduction band limit the transmission. While for long wavelengths (low energies) light
is reected because of the plasma edge. The transmission window for SnO2 is shown in
Fig. 2. For increasing conductivity of the material the long wavelengths transmission
decreases due to a lowering of the plasma frequency [58] of the material which is a direct
consequence of the increase in charge carriers. The plasma frequency is given by
xp ne2 =e0 e1 m

1=2

where n is the carrier concentration, e is the electron charge, e0 is the permittivity of free
space, e1 is the high frequency permittivity, and m* is the conductivity eective mass. Consequently, there exists a trade-o between conductivity of the TCO material and transparency in the infra-red region.
The coexistence of high electrical conductivity and optical transparency of TCOs are
due to the combination of several properties of the band structure of these materials.
These can be summarized in a simplied form as follows: (i) TCOs exhibit a wide optical
band gap prohibiting interband transitions in the visible range. (ii) Intrinsic dopants

Fig. 1. Decrease of the resistivity achieved for TCO materials over the last 30 years (from Ref. [47], reproduced
with permission from the MRS Bulletin).

54

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 2. Optical transmission of SnO2. (From Ref. [58], reproduced with permission from the MRS Bulletin.)

(oxygen deciency) or impurity dopants donate electrons into the conduction band. (iii)
The conduction band is a single band of s-type character that is strongly dispersed with
a minimum at the C-point of the Brillouin zone. This causes a light eective mass of conduction electrons (in contrast to the heavy holes in these materials) and results in a uniform distribution of the electron charge density and hence relative low scattering. These
eects contribute to the high mobility of the conduction electrons. (iv) A large internal
gap in the conduction band prohibits inter-conduction-band adsorption of photons in
the visible range.
The features of the electronic structure mentioned above for TCO materials can be
found in the band structure calculations shown in Fig. 3(a) and (c) for pure SnO2 and
In2O3 respectively. A critical point for TCOs is the creation of charge carriers. There
are three possible scenarios for generating charge carriers: (i) Intrinsic or extrinsic dopants
donate electrons to the conduction band of the host lattice and no additional energy such
as light or heat is required to create carrier electrons. (ii) Localized impurity bands exist
below the conduction band of the host lattice. In this scenario an activation energy is
needed to promote the electrons into the conduction band. (iii) The donor band becomes
the conduction band. However, hybridization of the donor states and valence band levels
of the host lattice may make a strict distinction in donor bands and host lattice valence
band not always a reasonable proposition.
For non-stoichiometric SnO2 Kilic and Zunger [57] showed that defect levels for oxygen
vacancies lie just 114 meV below the conduction band minimum (CBM) and thus can be
easily thermally ionized. For Sn interstitials a level 203 meV above the CBM was found
implying a spontaneous donation of electrons into the conduction band. This analysis
for non-stoichiometric, pure SnO2 suggests that the rst two scenarios outlined above
may be active. Experimentally the position of the dopant level of SnO2 was deterimed
from the activation energy for charge carrier generation. Marley and Dockerty found a
value for the ionization energy on SnO2 single crystals of 150 meV [59]. Other measurements yielded activation energies of 1040 meV [60], 24 4 meV [61], 60 meV [62],
700 meV (measured on cassiterite minerals) [63], and 4160 meV (Zn doped SnO2 samples)
[64]. Mizokawa and Nakamura [65] concluded from electron spin resonance (ESR) mea-

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

55

Fig. 3. Band structure calculations for (a) SnO2, (b) Sb doped SnO2 (reprinted with permission from Ref. [66], 
(1995) by the American Physical Society) and for (c) In2O3 and (d) Sn doped In2O3 (ITO) (from Ref. [67], 
(2001) by the American Physical Society).

surements the existence of two donor levels with activation energies of 100 meV and
300 meV respectively.
The modication of the band structure due to impurity dopants was addressed for Sb
doped SnO2 [66] and recently for Sn doped In2O3 [67,68]. In both cases the impurity atom
has one valence electron more than the cations of the host lattices. Thus a donation of the
extra electron into the conduction band upon substitutional replacement of a cation by the
impurity dopant may be expected. Band structure calculations for Sb doped SnO2
(Fig. 3(b)) suggested the formation of a Sb 5s like band in the SnO2 band gap with a free
electron-like character at the C-point. Therefore it was concluded that this band could be a
half-lled metallic band and that additional thermal excitation into the Sn-like bands
could increase the conductivity. However, these calculations were performed on a unit cell

56

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

with a Sn3SbO8 composition, i.e. a much higher Sb concentration than is reasonable in real
materials with Sb concentrations of a few percent at most.
Sn doped In2O3 appears to behave similar to Sb doped SnO2 (Fig. 3(d)). In these calculations 1 out of 16 In atoms were replaced by Sn, thus giving a Sn concentration of 6.5%,
which is about the solubility limit of Sn in In2O3. A partially lled conduction band separated by an additional gap from the remainder of the conduction band forms due to
strong hybridization of Sn and In 5s states. The splitting of the conduction band contributes to the lowering of optical adsorption due to inter-conduction-band transitions. On the
other hand, the dispersion of the conduction band is slightly less for Sn doped In2O3 than
for pure In2O3 which lowers the charge carrier mobility with increasing Sn concentration.
The strong dispersion of the s-type conduction band common for most TCOs gives rise
to a shift in the optical adsorption edge to higher energies with increasing charge carrier
concentration. This has been observed for In2O3 [45,53,6974], ZnO [75,76], CdO
[77,78], Cd2SnO4 [7981], and SnO2 [8287] and is known as the BursteinMoss eect
[88,89]. Filling of the strongly dispersing conduction band results in an increase of the
energy necessary to promote an electron from the valence band into an empty state in
the conduction band. This is schematically illustrated in Fig. 4, which shows this eect
for the case of InSb [88]. It is apparent from this representation that the optical band
gap is increased by an amount that depends on the curvature of the conduction and
valence band, which is proportional to their eective masses. Assuming parabolic bands
and a spherical Fermi surface one can derive that the BursteinMoss increase in energy
required for a transition DEBM
is given by
g

Fig. 4. Schematic of the broadening of the optical band gap due to the BursteinMoss eect. EG indicates the
fundamental band gap and EO, the optical band gap. The broadening of the optical band gap due to lling of the
conduction band is the dierence between EO and EG, and is indicated as DEBM
g . Such a broadening is observed
for materials with strongly dispersing conduction bands. The eective masses for the valence band and
conduction band are indicated as mp and mn, respectively (from Ref. [88],  (1954) by the American Physical
Society).

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

DEBM
h2 k 2F =2mcv
g

57

where kF is the Fermi wave vector and mcv is the reduced eective mass given by
k F 3p2 n1=3

1=mcv 1=mn 1=mp

and
respectively. n is the charge carrier concentration and mn and mp are the conduction and
valence band eective mass respectively.
The widening of the optical band gap due to the BursteinMoss eect is counteracted
by a narrowing of the fundamental band gap due to many body eects. However, the
BursteinMoss eect usually dominates [70,77]. In Section 7.2.1, the inuence of the
BursteinMoss eect in photoemission studies of Sb doped SnO2 is further discussed.
In many applications of TCOs in optoelectronic devices the interface properties, i.e. the
contact between the transparent electrode and the optical active material is critical. Surface science studies can address important issues such as surface structure and composition
of pure SnO2 and segregation of dopants to the surface. Furthermore, the impact structural and compositional changes have on the electronic surface/interface properties can
be addressed by photoemission studies. This will not just allow a better understanding
of the interface properties but also to devise ways for tuning these properties and thus
achieving better device operation. Such surface science studies are summarized in Sections
4 and 5 and Sections 7.2 and 8.2.9.
2.2. Heterogeneous catalysis
Many oxides mainly act as a support material for dispersed metal catalysts; tin oxide,
however, is an oxidation catalyst in its own right. Tin-oxide based catalysts exhibit good
activity towards CO/O2 and CO/NO reactions [9097]. As in most oxide catalysts the oxidation reactions are supposed to follow the Marsvan Krevelen mechanism. In this mechanism the molecules are oxidized by consuming lattice oxygen of the oxide catalyst which
in turn is re-oxidized by gas-phase oxygen. This is possible because transition and posttransition oxides have multivalent oxidation states that allow the material to easily give
up lattice oxygen to react with adsorbed molecules and can be subsequently re-oxidize
by gas-phase oxygen. In Section 4.2 the variable oxygen composition of SnO2 is discussed
for its three major low index surfaces. It is shown that for dierent oxygen chemical potentials surfaces with Sn4+ or Sn2+ are stable. This indicates that an easy reduction and reoxidation of SnO2 surfaces can be expected in catalytic oxidation reactions.
The activity and selectivity of tin-oxide catalysts can be substantially improved by
incorporation of heteroelements [98]. For instance the addition of copper, palladium
[99101], chromium [102,103] and antimony increases the total oxidation of carbon monoxide and hydrocarbons. The use of copper and chromium promoted tin(IV) oxide for the
ecient use in three-way automotive emission control has been discussed [104]. These
studies showed that the performance of such catalysts is similar to a Pt/Al2O3 catalyst
for CO and hydrocarbon oxidation. Antimony [105110], bismuth, molybdenum [111
116], and vanadium [117119] is added to tin oxide for the partial oxidation and ammoxidation of hydrocarbons, and phosphorus and bismuth are additives used for oxidative

58

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

coupling and oxidative dehydrogenation reactions. A summary of reactions catalyzed by


pure and modied SnO2 is given by Harrison [120].
The role of many of these additives is not fully understood. Most of the additives are
oxidized during operation conditions of the catalyst. A strong synergetic eect between
the additive and SnO2 is usually proposed that is manifesting itself in several ways. Special
active sites may be stabilized at the interface between the additive and SnO2. For example
Mo5+ has been identied in MoO3 catalysts supported on SnO2. It was suggested that
these Mo5+ sites play an important role in methanol and ethanol oxidation to formaldehyde and acetaldehyde or acetic acid, respectively [111,112].
In most cases the additives form clusters supported on SnO2, for antimony, however, it
is suggested that antimony oxide forms a solid solution with SnO2 with a strong surface
segregation of Sb. Thus Sb3+ surface species may be the active sites (see also Section
7.2.1 for a summary of studies of Sb doped SnO2).
In addition to creating active sites, electronic control of energy levels in supported clusters has to be considered for the activation of certain reactions by additives. In the electronic theory of catalysis the relative positions of the energy levels in the solid and the
redox potential of the reacting molecule is crucial [121127]. Essentially, the electronic theory of catalysis states that depending on the relative position of the Fermi level and the
redox potential of the molecules electrons are being transferred from the solid to the molecule or vice versa. This charge transfer may result in electrostatic bending of the energy
levels which will be discussed in more detail for gas sensor applications of SnO2 in the next
section. Oxidation of organic molecules over oxide catalysts usually proceeds by consumption of lattice oxygen in the oxidation reaction followed by reoxidation of the catalyst.
Thus the oxidation cycle involves two adsorbed redox couples:
RH O2 ) R-O H 2e


1=2 O2 2e ) O

2

5
6

In the rst step of this oxidation cycle electrons from the redox couple are injected into the
oxide. This can only occur if the redox couple of the adsorbed molecule is situated above
the Fermi level and above the bottom of the conduction band. For the second step, the
reoxidation of the catalyst, electrons need to be extracted from the oxide to activate adsorbed oxygen [128]. The necessary alignment of the energy levels for a catalytic cycle
to proceed is shown schematically in Fig. 5.
The adjustment of the energy levels in the catalyst can be achieved by several means of
which some may be relevant for SnO2 based catalysts. (i) Band alignment at the interface
of a monolayer oxide on an oxide support causes a band shift in the monolayer compared
to a respective bulk material. This may be relevant for instance for molybdenum, vanadium, and chromium oxide lms on SnO2 supports. (ii) Doping semiconductors by altervalent ions changes the catalytic activity due to a shift of the Fermi level [129]. Antimony
is one of the most common n-type dopant for SnO2, thus addition of Sb will shift the
Fermi level up into the conduction band as has been discussed above. This change in
the electronic structure of Sb doped SnO2 should be taken into account when assessing
the catalytic properties of this material. (iii) An external potential will shift the Fermi level
due to a eld eect. (In this context it may be worth mentioning that this external eld
may also be applied by charged, chemisorbed species. Band shifts of 0.7 eV have been
observed due to chemisorbed oxygen on SnO2 surfaces. This is the origin of the gas sen-

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

59

Fig. 5. Scheme of electron transfer across the adsorbate/oxide interface in the course of the heterogeneous
catalytic oxidation of hydrocarbon molecules by gas phase oxygen (from Ref. [127],  (2003), with permission
from Elsevier).

sitivity of SnO2 and is discussed in the next subsection.) (iv) Surface defects (intrinsic and
extrinsic) may cause acceptor or donor states within the band gap mediating the electron
exchange between adsorbed molecule and catalyst (see Section 5.2 for energy levels of
reduced and defective SnO2 surfaces). (v) Variations in the surface composition of the catalyst may change its work function, which inuences the relative alignment of the redox
levels of adsorbed molecules with respect to the band structure of the solid. This eect
may be pronounced for SnO2. This has been demonstrated by measuring the binding
energy of the molecular orbitals for benzene adsorbed on reduced and oxidized
SnO2(1 0 1) surfaces and is discussed in Section 8.2.9.
The combination of the various electronic eects on semiconducting oxides and the creation of special sites make it dicult to nd a fundamental description and understanding
of the role of additives; even if the catalysts were fully characterized. Dynamic changes of
the catalyst during operation adds to the challenge of understanding complex heterogenous catalysts. The probably best known example, where decades of research have not
yet been able to completely explain the synergy between an oxide and a metal is the hugely
important methanol synthesis catalyst Cu/ZnO. Since ZnO shares a similar electronic
structure with SnO2 that gives rise to similar applications (gas sensing, and transparent
conductor) some of the lessons learned from studying ZnO may be cautiously applied
to SnO2.
2.3. Solid state gas sensors
Materials that change their properties depending on the ambient gas can be utilized as
gas sensing materials. Usually changes in the electrical conductance in response to

60

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

environmental gases are monitored. Many metal oxides are suitable for detecting combustible, reducing, or oxidizing gases. For instance all the following oxides show a gas response
in their conductivity: Cr2O3, Mn2O3, Co3O4, NiO, CuO, CdO, MgO, SrO, BaO, In2O3,
WO3, TiO2, V2O3, Fe2O3, GeO2, Nb2O5, MoO3, Ta2O5, La2O3, CeO2, Nd2O3. However,
the most commonly used gas sensing materials are ZnO and SnO2 [130]. The gas sensitivity
of oxides is often divided into bulk- and surface-sensitive materials. TiO2 for example
increases its conductivity due to the formation of bulk oxygen vacancies under reducing
conditions and thus is categorized as a bulk sensitive gas sensing material. SnO2 on the
other hand, although bulk defects aect its conductivity, belongs to the category of surface
sensitive materials. It has not been addressed quantitatively why many TCO materials like
In2O3, ZnO, and SnO2 are also excellent gas sensing materials. However, the dispersing
conduction band with its minimum at the C-point and the high mobility of the charge carriers ensures that a change in charge carrier concentration results in a strong change in
electrical conductance of the material. Consequently, adsorbate induced band bending
has the potential to result in strong conductivity changes in these materials and thus trigger a gas response signal. In contrast TiO2 has an indirect band gap with the conduction
band minimum not at the C-point. Consequently band bending does not have a huge
impact on the conductivity of TiO2.
Pure oxides exhibit gas sensing properties by themselves. In practice additives are often
included to increase the sensitivity and selectivity to certain gases. Before we discuss the
inuence of additives on the gas response we describe the fundamental gas sensing mechanism of pure SnO2 in the next subsection.
2.3.1. Pure metal oxide gas sensors
Most gas sensors are made of porous thick SnO2 lms with a high surface to volume
ratio. During operation the gas sensing material is heated to around 300 C. The exact fundamental mechanisms that cause a gas response are still controversial, but essentially trapping of electrons at adsorbed molecules and band bending induced by these charged
molecules are responsible for a change in conductivity. Charge transfer in chemisorption
has been discussed briey in the previous subsection and is discussed in detail by Gopel
[131] and Madou and Morrison [132] with respect to gas sensors. When a molecule
adsorbs at the surface electrons can be transferred to this molecule if the lowest lying unoccupied molecular orbitals of the adsorbate complex lie below the Fermi level (acceptor levels) of the solid and vice versa electrons are donated to the solid if the highest occupied
orbitals lie above the Fermi-level of the solid (donor levels). Thus molecular adsorption
may result in a net charge at the surface causing an electric eld. This electrostatic eld
causes a bending of the energy bands in the solid. A negative surface charge bends the
bands upward, i.e. pushes the Fermi level into the band gap of the solid, eectively reducing the charge carrier concentration and resulting in an electron depletion zone. Depleting
electrons causes a positive space charge region that compensates for the negative surface
charge. This is illustrated in Fig. 6 on the example of negatively charged chemisorbed
adsorbate. The charge density distribution in the depletion zone can be determined by
solving the one-dimensional Poisson equation:
qz epz  nz D z  A z er e0 d2 V =dz2

where e is the elementary charge, p and n are the charge carrier densities, i.e. the number of
holes and electrons, and D+ and A are the densities of singly charged donors and accep-

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

61

Fig. 6. Schematic representation of band bending in the near surface region of an n-type semiconductor induced
by a (partially) lled donor state of an adsorbed molecule.

tors and er is the dielectric constant of the material. Charge neutrality of the system invokes that integration of the charge q(z) over the entire volume has to be equal to the surface charge trapped in the adsorbed molecule. Generally, the Poisson equation cannot be
solved analytically, but for the simplest case with negligible acceptor and hole concentration (n-type semiconductor) and complete ionization of the donors, i.e. D+(z) =
D+ = const., the Poisson equation can be written as
enz D er e0 d2 V =dz2

8
+

Assuming an abrupt transition of the charges with n(z) = D in the bulk and n(z) = 0 in
the space charge region makes the rst integration of (8) straightforward and yields a
linear relationship of the electric eld E:
Ez dV =dz eD z  D=er e0

where D is the Debye length or depth of the space charge region. Charge neutrality of the
system requires that the space charge is equal to the surface charge, i.e.
D D N s

10

where Ns is the number of elementary surface charges (singly charged adsorbed molecules)
per unit area.
To nd the variation of the potential in the space charge region the second integration
of the Poisson equation is calculated
2

V eD z  D =2er e0

11

If we dene the potential to be zero within the bulk V(z > D) = 0 and by using (10) to
eliminate D the surface (z = 0) potential can be written as
V s eD D2 =2er e0 eN 2s =2er e0 D

12

The above calculated variations of the space charge, electric eld, and potential energy are
illustrated in Fig. 7. Eq. (12) also indicates that the amount of band bending depends only
on the surface charges and the concentration of ionized donor atoms or defects.

62

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 7. Variation of the (a) space charge, (b) electric eld, and (c) electrical potential with distance from the
surface (z) into the bulk.

As a consequence of the charge carrier depletion zone in the presence of surface charges
the sheet conductivity r of the surface is altered. This change in conductivity is commonly
used as the signal in gas sensing devices. Returning to a general description the change in
conductivity due to surface band bending e DVs can be expressed as
 Z

Z
Dr e ln fnz  nb g dz lp fpz  pb g dz
13
where ln and lp are the electron and hole mobility, and nb and pb are the bulk electron and
hole densities respectively. For strongly n-type materials such as SnO2 the hole concentration may be neglected.
The band bending eect thus alters the sheet conductance of the surface layer of the
material. For samples with dimensions smaller than the space charge region a particular
strong change in the conductance may be expected [133]. This is true for many gas sensing
materials that consists of small grain sizes or is at least true along the necks that connect
individual grains and determine the conductivity of the lm [134,135]. Also, novel nanomaterials (discussed in Section 9) and epitaxial thin lms (Section 6) fulll this criterion
of small dimensions and are thus suitable materials (and are better dened than porous,
polycrystalline lms) to study gas sensor responses.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

63

In polycrystalline gas sensing materials another eect related to band bending at the
surface is often considered to contribute and amplify the gas sensing response. This eect
is due the formation of Schottky barriers at grain boundaries [136140]. Fig. 8 shows schematically the situation that can be found along a grain boundary. Upward band bending
induced by surface charges result in a barrier that conduction electrons have to overcome
to carry the current across the grain boundary.
Many chemisorbed or ionosorbed molecules that are associated with charge transfer
from the solid to the adsorbed molecule will induce a band bending and consequently a
variation in the conductivity. In the special case of monitoring reducing and oxidizing
gases, for which SnO2 nds its majority of applications, the adsorption of oxygen is of
particular interest (see also Section 8.2.1).
In general reducing gases increase the conductivity of the SnO2 gas sensing material
while the opposite is observed for oxidizing gases. Adsorbed negatively charged oxygen species are believed to be responsible for this phenomenon [62]. The negative charge trapped in
these oxygen species causes an upward band bending and thus a reduced conductivity compared to the at band situation. Reaction of these oxygen species with reducing gases or a
competitive adsorption and replacement of the adsorbed oxygen by other molecules
decreases and can reverse the band bending, resulting in an increased conductivity.
There are two main species of charged oxygen detected on oxide surfaces in an oxidizing environment. In a simplied fashion these can be described by the following
reaction steps:
e O2 () O
2

14

O2 () 2O
e O () O

15
16

O2 is also a possible surface species, but because of its lled atomic orbitals it is not paramagnetic and thus escapes detection in electron paramagnetic resonance (EPR)
measurements.

Fig. 8. Schematic diagram of charge carrier concentration in SnO2 grains. Negatively charged chemisorbed
oxygen species cause an upward band bending and consequently a depletion layer in the near-surface region. This
causes a Schottky-like barrier across grain boundaries.

64

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

The concentration of such adsorbates may be approximated from simple adsorption


models. In the Lennard-Jones model the rate of chemisorption is determined by an activation barrier between a physisorbed state and the chemisorbed state and an activation
barrier of desorption as is illustrated in Fig. 9(a). Using this model the rate of chemisorption dH/dt is expressed as the dierence in adsorption and desorption rates:
dH=dt k ads expfDEA =kT g  k des H expfDEA DH chem =kT g

17

where DEA is the activation barrier for chemisorption and DHchem is the heat of chemisorption. Under steady state conditions dH/dt = 0, i.e. the rate of adsorption is equal
to the rate of desorption, the coverage H is dependent on the heat of chemisorption
DHchem and is given by
H k ads =k des expfDH chem =kT g

18

Thus, generally the coverage will decrease with temperature. At low temperatures the molecules are, however, trapped in a physisorbed state and cannot overcome the activation
barrier DEA. This results in a maximum coverage at a temperature Tmax as is illustrated
in Fig. 9(b).
In the above model it was assumed that DHchem is independent of the coverage. This is
not generally the case though. Apart from site availability and repulsion between adsorbates, the heat of adsorption is changed as a consequence of surface charge induced-band
bending. In the simplest case for ionosorption the heat of adsorption is related to the separation of the acceptor level in the molecule and the electrochemical potential of the electrons (Fermi level) in the substrate. An upward bending of the energy levels in the solid,
due to a negative surface charge, reduces the energy separation between acceptor level and
Fermi energy and thus reduces the heat of adsorption. Thus with increasing coverage the
heat of adsorption is expected to decrease.

Fig. 9. (a) Lennard-Jones model for physisorption and chemisorption of a molecule. An energy barrier DEA has
to be overcome for a physisorbed molecule to reach the chemisorption well. Typical adsorption isobars are shown
in (b). The solid lines are equilibrium physisorption and a chemisorption isobar, the dashed line represents
irreversible chemisorption. A maximum coverage of chemisorbed molecules is obtainable at a temperature Tmax.
Below Tmax the chemisorption is irreversible because the rate of desorption becomes negligible.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

65

In addition to the heat of adsorption the activation barrier for chemisorption is also
inuenced by the concentration of charged molecules at the surface. Taking reaction
(14) as an example, the concentration of adsorbed O
2 is given by

dO
2 =dt k ads ns O2  k des O2

19

where ns is the density of electrons at the surface. This is given by the Fermi statistics and
the band bending and is expressed as
ns N C expfeV s EC  EF =kT g

20

where NC is the density of states at the conduction band minimum, VS is the surface
potential given by (12), and EC and EF are the energy levels for the conduction band
and the Fermi energy, respectively. Substituting (12) and using N s O
2 yields
2

ns N C expfEC  EF =kT g expfe2 O


2 =2er e0 D kT g

21

Comparing (17) with (19) and (21) shows that DEA e2 O
2 =2er e0 D and increases
with increasing adsorbed oxygen.
Under realistic conditions more than just one species are adsorbed and these are competing for adsorption sites and are inuencing the heat of adsorption and activation barriers of each other by the above-outlined mechanisms. Consequently, experimental results
on such a complex system cannot be easily interpreted. More surface science studies on
dened surfaces would be desirable. It is however questionable if UHV experiments alone
would give enough insight in the gas sensing mechanism. Controlled high pressure experiments, such as four-point conductivity measurements, combined with UHV sample preparation and characterization appears to be an important component for surface science
studies in this eld [141].

2.3.2. The role of additives in gas sensing materials


Additives are used for sensitizing and increasing the gas response to particular gases.
Generally it is assumed that the metal or metal-oxide additives reside on the surface of
the semiconducting, gas sensing oxide in form of dispersed clusters. However, there are
also some reports that show that in addition to these supported clusters bulk doping of
the semiconducting oxides by some commonly used additives such like Pd and Pt may
have to be taken into account [142144]. In such cases Pd and Pt are assumed to act as
bulk acceptor- or donor-type dopants [145,146]. The diusion of these additives into
SnO2 grains is dependent on the preparation conditions and it was shown, e.g., that oxidation of Pd is a necessary requirement for diusion of Pd ions into SnO2 [147].
For supported clusters, two mechanisms of sensitization are discriminated. These are (i)
catalytic (chemical) sensitization and (ii) electronic sensitization [148].
Chemical sensitization occurs if the supported clusters catalyzes reactions and reaction
products subsequently spill-over from the clusters onto the semiconducting oxide support.
These reaction products then cause the gas sensing response. Atomic hydrogen and oxygen
spill-over is frequently discussed in the literature as an important mechanism but CO spillover has also been suggested. Hydrogen reduces the oxide support and consumes chemisorbed oxygen species at the surface. In the extreme case, Sancier showed that reduction of
the oxide to its base metal is possible at much lower temperatures in the presence of Pdadditive than without [149]. Batley et al. reviewed spill-over of oxygen [150] and Sermon
and Bond reviewed spill-over in catalysis [151].

66

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Electronic sensitization occurs due to the alignment of the Fermi energy of the support with the additive. This is similar to the Schottky barrier when contacting a semiconductor. The Schottky barrier at the interface may be constructed by aligning the Fermi
levels and bending the bands until the work functions of the two materials match up at
the interface. This causes a space charge region in the gas sensing oxide similar to the
space charge regions discussed above as a consequence of chemisorbed, charged molecules. Since these space charge regions are only present in the vicinity of the additives,
the clusters have to be closely spaced in order to inuence the electron transport in the
gas sensing material. Adsorbates-induced shifts in the Fermi energy of supported clusters
or alterations of the work function at the cluster/support interface does inuence the
space charge region in gas sensing materials. Oxidation and reduction of the supported
clusters under varying oxidation potential of the gas phase may exert control over band
bending in the gas sensing material. Silver and palladium are common additives to SnO2
for gas sensing applications. Under atmospheric conditions these additives are present as
Ag2O and PdO [120,132,148,152]. The larger work function of the oxidized additives
compared to the SnO2 support causes an electron depletion zone in the support and thus
a reduced conductivity. Exposure to combustible gases readily reduces the additives to
metals which results in a lower work function and consequently less band bending in
the SnO2 substrate and thus an increased conductivity. Other reports, however, point
out that a complete reduction of PdO is not necessary to detect a large gas response.
In this case the consumption of adsorbed oxygen at the interface between PdO and
SnO2 under reducing condition is believed to be sucient to trigger a gas response
[153]. In Pd sensitized hydrogen sensors the diusion of hydrogen to the Pd/semiconductor interface may cause a variation of the work function that consequently triggers the
gas response [154].
Sensitizing SnO2 for H2S sensing was achieved, for example, by addition of Cu. Under
oxidizing condition Cu is present as CuO, which is p-type and thus a pn junction forms
between the additive and the n-type SnO2. This results in electron depletion at the interface. Exposure to H2S converts the CuO to Cu2S that exhibits metallic character and thus
increases the conductivity of the system [155157]. The formation of Cu2S was achieved in
an atmosphere containing 100 ppm H2S, CuO could be reformed in air at 100300 C
[158].
3. Comparison between stannic and stannous oxide
There are two main oxides of tin: stannic oxide (SnO2) and stannous oxide (SnO). The
existence of these two oxides reects the dual valency of tin, with oxidation states of 2+
and 4+. Stannous oxide is less well characterized than SnO2. For example, its electronic
band gap is not accurately known but lies somewhere in the range of 2.53 eV. Thus
SnO exhibits a smaller band gap than SnO2, which is commonly quoted to be 3.6 eV. Also,
there are no single crystals available that would facilitate more detailed studies of stannous
oxide.
Stannic oxide possesses the rutile structure and stannous oxide has the less common
litharge structure. Stannic oxide is the more abundant form of tin oxide and is the one
of technological signicance in gas sensing applications and oxidation catalysts. In addition to the common rutile (tetragonal) structured SnO2 phase there also exists a slightly
more dense orthorhombic high pressure phase. Suito et al. showed that in a pressure

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

67

temperature diagram the regions of tetragonal and orthorhombic phases can be separated
by a straight line of the equation p (kbar) = 140.0 + 0.022T (C) [159].
Fig. 10 shows the SnO phase diagram for atmospheric pressure [160]. This diagram
indicates the presence of an intermediate tin-oxide phase between SnO and SnO2 at elevated temperature. Sn3O4 is often given for its composition [161] but Sn2O3 [162,163]
has also been considered. In these intermediate oxides Sn is present as a mixture of Sn(II)
and Sn(IV) [161,163]. Also, the SnO2 phase can accommodate a signicant amount of oxygen vacancies. Li-Zi et al. measured the variation of the bulk oxygen vacancy concentration as a function of the oxygen partial pressure by coulometric titration [164]. They found
a relationship of the oxygen vacancy concentration X with the oxygen partial pressure P O2
1=n
via the proportionality X / P O2 with n varying between 5.7 and 8.3 for temperatures
between 990 K and 720 K respectively. In these studies a maximum oxygen deciency of
x = 0.034 in SnO2x at 990 K was observed before metallic Sn is formed. At lower temperatures less oxygen vacancies could be accommodated.
The heats of formation for stannous and stannic oxides at 298 K were determined to
DH = 68 cal/mol and DH = 138 cal/mol, respectively [165]. This results in
DH = 70 cal/mol for the reaction SnO(c) + 1/2 O2(g) ) SnO2(c). Also the disproportionation reaction of SnO(c) ) SnxOy(c) + Sn ) SnO2(c) + Sn has been reported to
occur at elevated temperatures [166]. This disproportionation of SnO into Sn and SnO2
proceeds via the aforementioned intermediate oxides [161,162]. This indicates that stannic
oxide is the thermodynamically most stable form of tin oxide. The oxidation of SnO lms
to SnO2 has been studied by Raman scattering, IR reectivity and X-ray diraction [167].
It was found that the oxidation starts with an internal disproportionation before external
oxygen completes the oxidation to SnO2. More importantly, (0 0 1)-textured SnO layers
convert into (1 0 1)-textured SnO2 lms. The same behavior was observed by Yamazaki
et al. [168]. Geurts et al. [167] explain this by the structural similarities between the tin
matrix of the SnO(0 0 1) plane and that of the SnO2(1 0 1) plane. Because of this structural
similarity essentially only the incorporation of an additional oxygen layer is required to
obtain the nal SnO2 structure. For comparison Fig. 11 shows top views of SnO(0 0 1)
and SnO2(1 0 1). A more detailed discussion of the structure of SnO and SnO2 is given next.

Fig. 10. SnO phase diagram (from Ref. [160]).

68

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 11. Comparison of the structure of the (a) SnO(0 0 1) and (b) SnO2(1 0 1) surfaces. Small, bright balls and
large, dark balls represent tin and oxygen, respectively. The similarities between these two crystal orientation have
been proposed to explain the conversion of SnO(0 0 1) textured lms into SnO2(1 0 1) textures upon oxidation.

3.1. Crystal structure of stannic and stannous oxide


The crystal structure of stannous oxide (SnO) is shown in Fig. 12. It has a tetragonal
unit cell with the litharge structure, isostructural to PbO. The symmetry space group is
and c = 4.8382 A
[169]. Each Sn
P4/nmm and the lattice constants are a = b = 3.8029 A

and O atom is fourfold coordinated with a bond length of 2.23 A. The structure is layered
in the [0 0 1] crystallographic direction with a Sn1/2OSn1/2 sequence and a van-der-Waals
. The positive charge of the Sn2+ ions is
gap between two adjacent Sn planes of 2.52 A
screened by electron charge clouds between the Sn planes, thus reducing the Coulombic
repulsion between adjacent Sn layers [170172]. These charge clouds, or charge hats, arise
from Sn 5s electrons that do not participate in the bonding for Sn(II) and thus can be
described as a lone pair. Galy et al. [173] dene this lone pair as an intermediate state
between an inert spherical s2-type orbital that is centered on the nucleus and a non-bonded
hybridized-orbital lobe that is not spherical but localized far from the atomic nucleus.

Fig. 12. Ball-and-stick model of the litharge structure of SnO.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

69

Thus, the valence shell electronic pair repulsion theory [174] takes this spatial eect into
account to predict local environment symmetries. The local environment can be very different for dierent Sn(II) compound materials and therefore the electronic charge can be
either spherically centered around the Sn atom like for example in the case of SnTe (cubic
NaCl structure) or be strongly directional, forming charge hats, like in the case of SnO
[175]. A more detailed discussion of the bulk electronic structure of SnO is given in Section
5.1.
Stannic oxide (SnO2) is much better characterized than stannous oxide. As a mineral,
stannic oxide is also called Cassiterite. It possesses the same rutile structure as many other
metal oxides, e.g. TiO2, RuO2, GeO2, MnO2, VO2, IrO2, and CrO2. The rutile structure
has a tetragonal unit cell with a space-group symmetry of P42/mnm. The lattice constants
and c = 3.1864 A
[176]. In the bulk all Sn atoms are sixfold coordiare a = b = 4.7374 A
nated to threefold coordinated oxygen atoms. The structure and composition of low index
SnO2 surfaces is discussed in Section 4. A representation of the unit cell can also be found
there (see Fig. 16, below).
3.2. Phonon modes of stannic and stannous oxide
Raman or infrared spectroscopy can be used to dierentiate SnO from SnO2. This has
been utilized for example in order to investigate the oxidation processes of Sn. Phonon
modes that cannot be assigned to either SnO nor SnO2 are sometimes observed in oxidation studies of Sn or SnO and this has been taken as evidence for the intermediate tin oxide
phase with Sn2O3 or Sn3O4 composition [167,177]. For this latter phase Raman peaks at
145 cm1 and 171 cm1 are reported [178].
For SnO group theory gives the following vibrational modes [167]:
C A1g B1g 2Eg A2u Eu 3 acoustic modes

22

Fig. 13 shows the corresponding atomic displacement for the Raman active modes and the
IR modes. In the IR modes the tin sublattice is displaced with respect to the oxygen sublattice. This displacement gives rise to a dipole moment. For the A2u mode this points in
the c-direction (IR polarization Ekc) but for the Eu mode its orientation is perpendicular
to the c-axis (IR polarization E ? c). Raman peaks for SnO at 113 (115) cm1 (B1g) and
211 cm1 (A1g) are reported [167,177].
For SnO2 the six unit cell atoms give rise to 18 vibrational modes. These can be represented by [179181]:
C A1g A2g B1g B2g Eg 2A2u 2B1u 4Eu

23

Of these 18 modes, two are IR active (the single A2u and the triply degenerated Eu mode),
four modes are Raman active (three non-degenerate modes, A1g, B1g, B2g and the doubly
degenerate Eg) and two are silent (A2g and B1u). One A2u and two Eu modes are acoustic.
The atomic displacements for Raman and IR-active modes are illustrated in Fig. 14. In the
Raman active modes the oxygen atoms vibrate while the Sn atoms are at rest. The nondegenerate modes, A1g, B1g, and B2g vibrate in the plane perpendicular to the c-axis while
the doubly degenerate Eg mode vibrates in the direction of the c-axis. The B1g mode consists of rotation of the oxygen atoms around the c-axis, with all six oxygen atoms of the
octahedral participating in the vibration. In the A2g infrared active mode, Sn and oxygen
atoms vibrate in the c-axis direction, and in the Eu mode both Sn and O atoms vibrate on

70

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 13. Schematic representation of the atomic displacements for the Raman and IR-active modes of SnO (from
Ref. [167],  (1984), Elsevier).

Fig. 14. Schematic representation of atomic displacements for the Raman and IR-active modes of SnO2 (from
Ref.[181],  (1980) by the American Physical Society and from Ref. [187],  (1983), Elsevier).

the plane perpendicular to the c-axis. The silent modes correspond to vibrations of the Sn
and O atoms on the direction of the c-axis (B1u) or in the plane perpendicular to this direction (A2g). The dispersion of the phonon modes has been calculated by Sato and Asari

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

71

Table 2
Vibrational modes of SnO2
Ref.

A1g
[cm1]

B2g
[cm1]

B1g
[cm1]

A2g
[cm1]

Eg
[cm1]

A2u (TO)
[cm1]

A2u (LO)
[cm1]

B1u
[cm1]

Eu (TO)
[cm1]

Eu (LO)
[cm1]

[179]

638

782

100a

398a

476

477

705

140a
505a

244
293
618

276
366
710

465

704

243
284
605

273
368
757

[184]

[185]
a

776

123

475

Calculated value.

[182] and Katiyar et al. [179]. The pressure and temperature dependence of the Raman
modes was investigated by Peercy and Morosin [183]. A selected compilation of measured
and calculated vibrational modes for SnO2 is given in Table 2.
Surface optical phonon modes (so-called FuchsKliewer [186] phonons) excited in high
resolution electron energy loss spectroscopy (HREELS) were studied for SnO2(1 1 0) single
crystals [187]. Two single phonon losses were identied at 339 cm1 and 694 cm1. In
HREELS similar to IR spectroscopy only the A2u and Eu modes are active because only
these modes involve the necessary dipole change. Furthermore, in HREELS energy loss
can only occur for modes with dipole changes normal to the surface. This excludes the
A2u mode from contributing in HREELS studies of the (1 1 0) surface, because the atomic
replacements of this mode are parallel to the c-axis, i.e. the [0 0 1] crystallographic direction. Studies on SnO2(1 0 1) crystallites grown on Pt(1 1 1) substrate (see Section 6.3) also
showed similar losses as for the (1 1 0) single crystal with losses at 330 cm1 and 705 cm1
[188]. Also, only two phonon losses at 290 cm1 and 669 cm1 were reported for polycrystalline samples. These phonon losses shifted to 323 cm1 and 726 cm1 for samples that
were doped with 3% Sb [189]. The relative surface insensitivity of these phonon losses with
a probing depth of a several nanometers and their strong intensity make identication of
vibrational losses of adsorbed species dicult.
3.3. Auger- and core level spectroscopy
Discrimination between SnO and SnO2 in photoemission studies is complicated due to
only a small shift in the Sn 3d core level binding energy. A summary of reported values is
given in Table 3. For the O 1s core level, Jimenez et al. reported a shift in the binding
energy of 1 eV for SnO compared to SnO2, i.e. 530.4 eV for SnO and 531.4 for SnO2.
The modied Sn Auger parameter (a 0 = BE Sn 3d5/2 + KE Sn MNN AES) has been
shown to be a more reliable indicator of the charge state of Sn than considering the 3d
core level position alone [190192]. Reported Auger line positions and the Auger parameters are also summarized in Table 3 for dierent oxidation states of tin. For ultrathin
tin oxide lms at the interface to dissimilar materials strong shifts of the binding energy of
Sn states may be observed [192,193]. These are due to (i) reduced nal-state screening within
small islands and/or (ii) initial state charge transfer from the Sn to the substrate. This complicates a reliable determination of the composition of such ultrathin lms even more.

72

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Table 3
Comparison of the reported binding energies of Sn 3d5/2 (EBE), its shift relative to the metallic tin state (DEBE), the
reported kinetic energies of the M5N45N45 or M4N45N45 Auger line (AES EKE), its shift relative to metallic Sn,
and the modied Auger parameter a 0
Phase

EBE Sn 3d5/2
[eV]

Sn
Sn
Sn
Sn
SnO
SnO
SnO
SnO
SnO2
SnO2
SnO2
SnO2
SnO2

483.8
483.8
484.9
484.65
485.6
486.9
486.4
486.5
487.5
486.3
487.1
486.4
486.6

DEBE (Sn)
[eV]

AES EKE M5N45N45


[eV]

AES EKE M4N45N45


[eV]

DEKE
[eV]

a0
[eV]

430

915.4

430

914.9

a0
[eV]

1.8
1.5

426.2

3.8

912.6

432.7
431.0
2.5
2.2
1.75

919.2
918.5

424.2

5.8
5
424.6

911.2
909.7
911.2

Ref.
[193]
[194]
[191]
[195]
[194]
[196]
[191]
[190]
[190]
[194]
[191]
[195]
[196]

Another way to distinguish between SnO and SnO2 in photoemission studies is to characterize the valence band. Only for Sn(II) strong Sn 5s derived states are part of the valence
band maximum and this allows to discriminate it from Sn(IV). This is described in Section
5.1.
4. Surface composition and structure of SnO2 low index surfaces
The surface energies of low index SnO2 surfaces with a termination that maintains the
bulk composition have been calculated [197200]. The results of these calculations are
summarized in Table 4.
For these bulk terminated SnO2 surfaces, i.e. surfaces with surface-tin atoms in their
bulk Sn4+ oxidation state, the (1 1 0) surface exhibits the lowest energy surface followed
by the (1 0 0), (1 0 1), and (0 0 1) surfaces. The fact that the (1 1 0) surface is the lowest
energy surface can be appreciated by investigating the crystallographic orientations of
the bulk termination of single crystals. A photograph of a single crystal grown by a vapor
phase transport technique is shown in Fig. 15. It is obvious that the {1 1 0} surfaces make
up the majority of the surface area followed by the {1 0 1} faces and small {1 0 0} faces. No
{0 0 1} faces are observed. Although this crystal does not necessarily represent an ideal
crystal one would obtain from a Wuls construction, it nevertheless clearly demonstrates
Table 4
Calculated values for surface energies of stoichiometric surfaces in (J m2)
Surface

B3LYP [197]

GGA [198]

LDA [199,200]

GGA [201]

GGA [202]

(1 1 0)
(1 0 0), (0 1 0)
(1 0 1), (0 1 1)
(0 0 1)

1.20
1.27
1.43
1.84

1.04
1.14
1.33
1.72

1.301.40
1.661.65
1.55
2.36

1.01

1.42

1.21
1.29
1.60

The dierent approximations of the DFT functionals are given.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

73

Fig. 15. Photograph of a SnO2 single crystal grown by a vapor phase transport technique (sample grown by
Helbig see Ref. [387]).

the preferential terminations of a SnO2 single crystal. Representations of the bulk terminated surfaces are shown in Fig. 16(b)(d), left side panels. These surfaces are constructed
in an autocompensated way, i.e. the same number of Sn-to-O as O-to-Sn bonds are cut.
This is a rule that often allows assessing the most likely surface termination of ionic

Fig. 16. Ball-and-stick models of SnO2 low index surfaces. The bulk rutile-unit cell is shown in (a). (b), (c), and
(d) show bulk termination of the (1 1 0), (1 0 0), and (1 0 1) surfaces respectively. On the left-hand side
stoichiometric bulk terminations are represented. On the right-hand side surfaces with reduced oxygen
concentration are indicated. These surfaces were obtained by removing twofold coordinated bridging oxygen
rows from the stoichiometric surfaces. It has been shown that the so obtained reduced (1 0 0) and (1 0 1) surfaces
are thermodynamically favored for low oxygen chemical potential. The reduced (1 1 0) surface exhibits a high
surface energy and thus is unlikely to form.

74

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

crystals. This is because by cutting the same number of bonds the overall composition of the
crystal is not changed and the surface charge can be redistributed so that all surface atoms
maintain their preferred oxidation state. Although this method often predicts the surface
termination accurately, in many cases more complex surface reconstruction form that
lower the surface energy or reduce strain in the surface layer. Also, for elements like Sn
that have dual or multivalencies the situation of picking the right bulk termination may
not be that straightforward. For the SnO2(1 0 0) and (1 0 1) surfaces, for instance, one
can cut the bulk crystal along a plane where only Sn-to-O bonds are broken. This is illustrated on the right panels of Fig. 16. By doing so one ends up with threefold coordinated
Sn atoms at the surface that can attain a Sn2+ oxidation state. Consequently, such a surface termination satises the other valency of Sn and therefore may also be considered to
be autocompensated. In Section 4.2 it is shown that both the stoichiometric Sn(IV) and
the reduced Sn(II) bulk terminations are stable and that the surface termination one
obtains on these surfaces depends on the equilibrium conditions with the gas phase. For
the (1 1 0) surface the situation is more complicated. Cutting only SnO bonds results in
a surface illustrated at the right of Fig. 16(b). The composition of such a surface layer
does, however, not permit the surface Sn atoms to obtain a single Sn(II) or Sn(IV) valency
and calculations showed that such a surface exhibits a high surface energy. Alternative,
reduced (1 1 0) surfaces with bulk truncations but lower energies are discussed in Section
4.2.2. In the following section, the more complex surface structures that are found experimentally for the (1 1 0) surface under dierent preparation conditions are addressed.
4.1. Experimental studies of the SnO2(1 1 0) surface
The most frequently studied surface of SnO2 is the (1 1 0) orientation. Unfortunately,
this surface also exhibits the most complex surface reconstructions with structures that
are not well understood.
Depending on the preparation conditions several surface structures were found. In particular, standard vacuum surface preparation techniques consisting of ion sputtering and
annealing does not result in a stoichiometric, bulk terminated surface. In LEED studies
[203] the following sequence is observed with increasing annealing temperature after sputtering: diuse 1 1, faint c(2 2), c(2 2) co-existent with a 4 1, sharp 4 1 with missing ((2n  1)/4, 0) spots indicating a glide plane symmetry, 1 1, and nally a 1 2
pattern. Fig. 17 show LEED patterns and the corresponding STM studies. Surfaces that
exhibit c(2 2) LEED pattern are always quite rough and are not shown in Fig. 17. Coexistence of c(2 2) and 4 1 LEED spots is due to 4 1 domains and small islands with a
c(2 2) structure (Fig. 17(a) and (b)). Annealing of these samples at a temperature of
1000 K result in sharp 4 1 LEED pattern. Such surfaces exhibit extended at terraces
with no preferential step-edge orientation (Fig. 17(c)). The 4 1 unit cell on the terraces
can be easily discerned by STM (Fig. 17(d)). Sometimes, antiphase-domain boundaries
between 4 1 domains shifted by half the 4 1 unit cell are observed on at terraces
(Fig. 17(e)). These antiphase domains are separated by trenches that are close to a monolayer deep and about a unit cell wide, thus these features are unlikely to be just of electronic contrast in STM but are real topographical features. Further annealing results in
decomposition of the 4 1 structure and domains without apparent long range order form
(Fig. 17(f)). At annealing temperatures of about 1200 K all the 4 1 domains have disappeared and only disordered domains remain at the surface (not shown). For such a sur-

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

75

Fig. 17. Overview of surface structures obtained for the SnO2(1 1 0) surface for dierent annealing temperatures
in UHV after ion sputtering. LEED patterns and STM images are shown. The 1 1 unit cell is indicated in the
LEED patterns. (a) and (b) show STM images for low annealing temperatures surfaces are characterized by a
coexistence of 4 1 and c(2 2) domains. The surface exhibits a high density of monolayer high islands that
presumably are the c(2 2) domains while the substrate exhibits mainly 4 1 domains. (c)(f) show 4 1
reconstructed surfaces. In (f) domains of disordered structure appear. (g) and (h) show 1 2 reconstructed
surfaces obtained after high temperature annealing (note that (g) is an AFM image) (from Ref. [209],  (2003),
Elsevier, and from Ref. [204],  (2000) by the American Physical Society).

76

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

face 1 1 LEED spots are observed that originate from the underlying SnO2 bulk structure. In early work this 1 1 LEED pattern was, perhaps erroneously, associated with a
stoichiometric surface. Further vacuum annealing results in a 1 2 reconstruction. In
STM and NC-AFM the surface appears quite corrugated along the 1 1 0 direction with
rows of twice the unit cell periodicity running in the [0 0 1] direction [204]. This reconstruction was suggested to be isostructural to the TiO2(1 1 0)-1 2 reconstruction.
The 4 1 reconstruction is the most frequently studied surface because it is stable over a
wide temperature range and it is the most uniform surface of all the vacuum prepared surfaces. Early models for the 4 1 structure suggested a SnO(1 0 1) overlayer [205] that
would form a coincidence lattice with the SnO2(1 1 0) surface with a 4 1 periodicity if
the SnO(1 0 1) layer was stretched by 10%. However, because SnO is a layered material
with layers in the (1 0 0) planes (see Fig. 12) the existence of a (1 0 1) plane is somewhat
questionable. Later, based on STM studies, a dierent model of periodic, in-plane oxygen
vacancies was proposed [204,206,207]. One of these studies, however, also reported the coexistence of two kinds of 4 1 reconstructions [204]. LEED IV measurements showed a
reasonable well t with the in-plane missing oxygen model [208]. More recent STM studies, however, showed features like antiphase-domain boundaries (Fig. 17(e)) that are
inconsistent with any of the proposed models [209]. Furthermore, experimental observations that the 4 1 reconstruction cannot easily be lifted by exposure to up to at least
100 mbar O2 at room temperature but instead needs to be heated to elevated temperatures
(600700 K), indicates a signicant activation barrier for oxidation. Such an activation
barrier would not be expected if the surface was a bulk termination with oxygen vacancies
only. Consequently, a complex tin oxide overlayer with a structure dierent from the bulk
may be a more likely description of the 4 1 reconstruction. In these authors opinion,
however, a convincing model has yet to be proposed.
Preparation of the (1 1 0) surface in vacuum alone always results in an oxygen decient
surface. In order to prepare stoichiometric bulk terminated surfaces samples have to be
annealed in high pressures of an oxidizing gas (O2 [210,211], N2O [212,213]) or exposed
to an oxygen plasma [214]. Sheet conductance measurements of the SnO2(1 1 0) surface
showed that oxygen decient surfaces exhibit a more than two orders of magnitude
increased conductance [215]. Preparation in fairly low oxygen pressure of 103 mbar at
elevated temperatures and a cool down process of the sample in an oxygen background
results in the 1 1 LEED pattern [209] shown in Fig. 18. This surface also exhibits a by

Fig. 18. LEED pattern and STM images of SnO2(1 1 0) surfaces prepared by (a) annealing in 103 mbar O2 and
(b) after subsequent annealing to 800 K in vacuum. The scan area of the STM images is 100 100 nm2 (from Ref.
[209],  (2003) with permission from Elsevier).

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

77

a factor of eight increased [O]/[Sn] ratio in low energy He+ ion scattering experiments
compared to vacuum prepared surfaces. In STM studies the surface exhibits terraces with
straight step edges oriented preferentially along low index crystallographic orientations of
the bulk. Although there is some order on terraces there are also clear indications of many
defects. Annealing of such an oxidized surface to 800 K in vacuum causes loss of surface
oxygen as is evidenced by He+ ion scattering, but the surface retains a 1 1 LEED pattern
up to 800 K. In STM islands are observed that have heights of non-integer interlayer
separation indicating formation of non-bulk like surface layers [209].
To complicate matters more, SnO2 samples that have been thoroughly oxidized at elevated oxygen pressures and temperatures do not form a 4 1 reconstruction upon vacuum
annealing. Instead, a 1 2 LEED pattern is observed after annealing to around 900 K.
LEED and STM images of this structure are shown in Fig. 19. The surface appears not
well ordered on an atomic scale but troughs and ridges with two times the unit cell periodicity are easily discernable. This structure is similar to the 1 2 structure obtained at
1200 K for samples prepared in vacuum without oxygen treatment (see Fig. 17). The lack
of a 4 1 reconstruction if samples are prepared without ion sputtering suggests that articial oxygen depletion by preferential oxygen sputtering is required for the formation of a
SnO2(1 1 0)-4 1 reconstruction. It is also possible that the 4 1 overlayer formed on a
strongly oxygen depleted surface extends deeper than just the surface layer. Justication
for this assertion comes from measurements of the change in oxygen concentration in
the surface layer in photoemission studies [216]. In one study the Sn 4d/O 2s ratio in the
Cooper minimum of the Sn 4d level was measured. Under most UPS and XPS conditions
the O 2s signal is very weak compared to photoemission from the Sn 4d signal and thus
makes it almost impossible to measure a reliable ratio. However, in the Cooper minimum
[217] for Sn 4d states photoemission is strongly suppressed. This minimum has been calculated for Sn to be around a photon energy of 140 eV [218]. Fig. 20(a) and (b) show an
energy series for a sputtered SnO2 sample and a sample annealed to 1100 K respectively.
Fig. 20(c) shows the Sn 4d/O 2s ratio measured for dierent sample preparation conditions
in the Cooper minimum. Although no information of the surface structure is given it is
apparent that high temperature annealing is necessary to restore some of the oxygen in
the surface layer. More evidence for the reduced nature of the 4 1 structure after vacuum
preparation is given in Section 5.2.1. Thus one may propose that the (1 1 0) surface is

Fig. 19. LEED and STM images of a 1 2 reconstructed surface obtained by vacuum annealing of a thoroughly
oxidized SnO2(1 1 0) sample to 950 K in vacuum.

78

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 20. Photoelectron spectra for (a) sputtered SnO2 and (b) vacuum annealed samples to 1100 K for dierent
photon energies close to the Cooper minimum for Sn. The Sn 4d/O 2s ratio measured at the Cooper minimum is
shown in (c) for dierently prepared samples. The large oxygen deciency after sputtering of the samples is
apparent (from Ref. [216]).

covered by a reduced overlayer and this overlayer is stable until it starts to decompose (see
Fig. 17(f)). This decomposition may involve a similar disproportionation reaction as
described above for SnO into SnO2 and metallic Sn. At the elevated temperatures at which
the 4 1 reconstruction starts to decompose (1100 K) metallic Sn evaporates from the
surface. In such a scenario the temperature necessary to trigger the decomposition of
the 4 1 overlayer limits the formation of the 1 2 reconstruction. For samples that do
not form a 4 1 structure this 1 2 reconstruction can form at lower temperatures as is
observed for initially fully oxidized surfaces. It is worth pointing out that if the above
assertions are correct that the 4 1 reconstruction will not be encountered in any realistic
applications of SnO2 as gas sensors, or catalysts and reduced surface structures with a
1 1 periodicity and the 1 2 reconstruction are more signicant. Nevertheless, the
4 1 overlayer is an intriguing structure that is far from being understood.
The above discussion of the (1 1 0) surface demonstrates that the surface composition of
SnO2 is a delicate matter. Fundamental aspects of the variation of the surface composition

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

79

as a function of oxygen chemical potential, i.e. oxygen partial pressure and temperature, is
discussed in the next subsection for the (1 1 0) surface as well as for the less frequently studied (but apparently less complex) (1 0 1) and (1 0 0) surfaces.
4.2. Oxygen chemical potential dependent properties of SnO2(1 1 0), (1 0 0),
and (1 0 1) low index surfaces
As mentioned above many oxide surfaces may exhibit dierent surface compositions
compared to the bulk. This surface composition is often variable and depends on temperature and gas pressure. Thus it is important to keep in mind that the surface properties of
model systems under idealized conditions may be dierent from those under realistic operation conditions of catalysts. As pointed out above, this is one manifestation of the often
discussed pressure gap in surface science studies of catalysts. As is discussed in detail
below, tin oxide, because of its dual valency, is susceptible to forming surface phases with
dierent oxygen composition. Because of the obvious importance of the surface composition for their chemical and electronic properties we briey discuss the thermodynamic
description of surfaces under varying oxidation potential and its implications for surface
science studies.
In thermodynamic equilibrium the chemical potentials of the bulk, surface, and the gas
phase are equal. Thus a change in the gas phase chemical potential has to result in an
equilibration of the surface chemical potential, which in turn may cause compositional
and structural changes at the surface. For instance, the dierence in oxygen chemical
potential of a pure element in contact with an oxidizing atmosphere drives the corrosion
of surfaces, i.e. incorporation of oxygen into the lattice. In order to investigate the surfaces
of clean elements, conditions had to be developed that eliminated the interaction of the gas
phase with the surface. Ultrahigh vacuum conditions reduce the oxidation potential of the
environment and allow the investigation of pure elements without presence of adsorbates
or formation of an oxide layer. However, if an oxide is studied, then the oxygen chemical
potential of the stoichiometric oxide surface may be higher than that of the vacuum. This
gradient in the chemical potential will be lowered in the absence of signicant kinetic barriers by removing oxygen from the surface and thus bringing the system closer to equilibrium. The tendency of oxide surfaces to reduce is a fundamental setback for preparation of
some oxide surfaces under UHV conditions. Such preparation conditions may result in
surface compositions and structures that dier from that of surfaces at elevated pressure
conditions. This is a dilemma if one is interested in UHV studies of surfaces that simulate
oxides at atmospheric conditions. On the other hand if one is aware of this situation, different surface structures and compositions can be obtained by tuning the preparation conditions and new insight can be gained on surface properties at dierent oxidation chemical
potentials of the gas phase. Ultimately this will enhance our fundamental understanding of
oxidation catalysts and metal oxide gas sensors.
Recent ab initio atomic thermodynamics calculations have highlighted the issue of compositional variations of oxide surfaces under various chemical potentials [1,219227]. In
these calculations the surface energies are computed for dierent surface structures and
compositions. These surface energies are functions of the chemical potential, i.e. partial
pressures and temperature. The surface energy can be expressed as
h
i.
X
cli G 
N i li T ; pi A.
24

80

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

with G, the Gibbs free energy of the solid (slab) including the surfaces, li(T, pi) the chemical potential of species i which is a function of temperature T and partial pressure pi, and
Ni is the number of particles of the ith species. The expression for the surface energy is
normalized to the surface area A. For the special case of SnO2 surfaces with only tin
and oxygen forming the thermodynamic reservoirs, i.e. no adsorbates or contaminations
are considered, the expression for the surface energy simplies to
clSn ; lO 1=AG  N Sn lSn  N O lO

25

where lO 1=2 lO2 . Furthermore, it can be assumed that the chemical potentials of tin
and oxygen are not independent, but rather that the SnO2 bulk is in thermodynamic
equilibrium with the O2 gas phase. This yields the simple relationship
lSn 2lO gSnO2

26

where gSnO2 is the Gibbs free energy for the bulk oxide. Obtaining lSn from (26) and substituting it into (25) gives an expression for the surface energy that is only dependent on the
oxygen chemical potential
clO 1=AG  N Sn gSnO2 f2N Sn  N O glO

27

If we are considering a solid for which stoichiometry is conserved throughout the crystal,
i.e. the surface layer has the same SnO2 stoichiometry as the bulk, then NO = 2NSn and the
surface energy c becomes independent of the oxygen chemical potential. If the considered
surface layer has a dierent composition than the bulk, the surface energy becomes a linear
function of the oxygen chemical potential with a positive slope for surfaces with oxygen
deciency relative to the bulk and a negative slope for oxygen enriched surfaces. Therefore, dierent surface compositions have dierent surface energies for dierent oxygen
chemical potentials. Obviously, the surface with the lowest energy at a given oxygen chemical potential is the thermodynamic stable surface. Consequently, oxide surfaces may
exhibit compositional and structural phase transitions as a function of the gas phase oxygen
chemical potential. Regions of compositional changes in the surface-phase diagrams may
play a signicant role for oxidation reactions at surfaces, because lattice oxygen can be
easily consumed by the reactant and at the same time replenished by the gas phase oxygen.
For a successful theoretical study of oxide surface phases relevant surface structures
have to be identied. DFT calculations allow excellent assessment of the stability of surfaces, but without experimental input the actual structures of often complex surface reconstructions cannot always be predicted. Therefore it is necessary to have experimental
justication for considering the right surface structure. Vice versa experimental studies
are often not unambiguous in determining structures and DFT calculations are important
to clarify the proposed models. In the next three subsections experiments and ab initio
thermodynamics calculations are summarized that elucidate the oxygen chemical-potential
dependent surface composition and structure of SnO2. First, in Section 4.2.1 we describe
experimental studies that determine phase transition temperatures under UHV conditions
and determinations of surface structures. Second, in Section 4.2.2 theoretical ab initio
thermodynamics calculations of the thermodynamic stability of surface structures and
composition as a function of oxygen chemical potentials are discussed and compared to
experimental results. Third, in Section 4.2.3 the dierences between the thermodynamic
approach in the theoretical treatment and the experimental approach that measures
desorption barriers is discussed.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

81

4.2.1. Surface structure and composition I: Experimental


In order to determine surface phase transitions experimentally the composition of the
surface layer has been probed as a function of sample temperature. He+ ion scattering
has been utilized for this purpose. Its extreme surface sensitivity allows identication of
atomic species in the topmost surface layer. A quantitative determination of the surface
composition is however in most cases not possible because of unknown scattering crosssections and neutralization probabilities for the He+ ions interacting with dierent target
atoms and atomic shadowing eects. Nevertheless, a decrease in the oxygen signal concurrent with an increase in the tin signal is an unambiguous sign for oxygen depletion from
the surface layer. Continuous monitoring of the surface composition with LEIS while the
sample temperature is ramped allows determination of critical temperatures at which compositional changes occur at the surface. The temperature programmed LEIS measurements as well as structural studies by LEED and STM are discussed in the following
for the three low index SnO2 surfaces investigated [228].
4.2.1.1. The SnO2(1 1 0) surface. The structure of the (1 1 0) surface after UHV preparation
was extensively studied previously (see Section 4.1). Cox et al. were the rst who evaluated
the surface composition of SnO2(1 1 0) surfaces by LEIS [210]. Temperature programmed
LEIS studies for an initially fully oxidized (1 1 0) sample are shown in Fig. 21(a) [228]. Initially, at low temperatures, the oxygen signal is stronger than the Sn scattering peak. At
higher temperature this is inverted. A transition occurs in a temperature range of 440
520 K. This transition is somewhat clearer dened for the oxygen scattering peak, which
stays roughly constant up to 440 K then up to 520 K the peak intensity sharply drops.
For even higher temperatures the oxygen peak continues to decrease (and the tin peak
increases), but with a much smaller rate. The continuous change in the Sn/O ratio indicates further loss of oxygen from the surface at higher temperature. In qualitative LEED
studies no transition is discerned in this temperature range and only a 1 1 pattern is
observed. STM studies on oxidized surfaces discussed in Section 4.1 (see Fig. 18) shows
some structural changes upon annealing but atomic-scale models could not be derived
from these studies.
4.2.1.2. The SnO2(1 0 0) surface. Only one study of SnO2(1 0 0) surface has been reported
so far [228]. However, the large bulk-terminated terraces that are obtainable on this surface orientation may make this a good model surface for future studies.
Preparation by sputtering and annealing in a range of 6001000 K results in a 1 1
LEED pattern. The sample retains its 1 1 LEED pattern even after exposure of the surface to 10 mbar O2 at room temperature or elevated temperatures. Although no changes in
qualitative LEED can be identied the oxygen to tin peak ratios measured in LEIS
increases signicantly after high pressure oxygen exposure. The evolution of the oxygen
and tin peaks in LEIS measurements with temperature of an initially oxidized surface is
shown in Fig. 21(b). From this representation it is apparent that the surface composition
changes in a temperature range of 610660 K. The small decrease of the oxygen signal with
a concurrent increase in the tin signal indicates a depletion of surface oxygen. Since the
dimensions of the unit cell remain unchanged it is proposed that this transition is a consequence of the removal of bridging oxygen atoms from the stoichiometric SnO2 bulk termination. Ball-and-stick models of the stoichiometric and reduced surfaces are shown in
Fig. 16(c). Reduction leaves a surface with threefold coordinated Sn atoms and a surface

82

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 21. Evolution of the surface composition measured by He ion scattering spectroscopy with increasing sample
temperature under UHV conditions for initially fully oxidized (a) SnO2(1 1 0), (b) SnO2(1 0 0), and (c) SnO2(1 0 1)
surfaces. The left panel shows the peak intensity for Sn and O versus temperature. The right panel shows the HeISS spectra for the oxidized surface and after vacuum annealing (from Ref. [228],  (2005) by the American
Physical Society).

layer with a SnO composition. Justication for this assignment also comes from STM
measurements and DFT calculations that are discussed below.
The surface structure of vacuum prepared surfaces, i.e. a reduced surface layer, was
studied by STM. The atomic corrugation on terraces corresponds to the surface unit cell
of SnO2(1 0 0)-1 1. Fig. 22(a) shows a row structure with the rows running along the
direction of the close packed Sn atoms, i.e. the [0 0 1] direction (see Fig. 16(c)).
Fig. 22(b) shows an atomically resolved STM image that allows to clearly discerning
the unit cell. The measured dimensions of the unit cell agree well with the dimensions
4.74 A
.
of a 1 1 SnO2 surface, i.e. 3.19 A

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

83

Fig. 22. STM images of a vacuum prepared SnO2(1 0 0) surface. The scan sizes for the STM images are (a)
20 20 nm2, (b) 4.7 4.7 nm2, (c) 200 200 nm2, and (d) 45 45 nm2. The surface unit cell is indicated in (b).
Two kinds of steps are dierentiated in (c). Step I is a full unit cell high and step II consists of two steps each of
half-unit cell height (from Ref. [228],  (2005) by the American Physical Society).

Two kinds of steps have been identied in STM. This is apparent in Fig. 22(c): Type-I
steps are straight and are oriented along the [0 0 1] crystallographic direction. Type-II steps
are less well oriented but are mainly aligned along the [0 1 0] direction. The measured step
. It can be seen that step II consists of a
height of step I is close to unit cell height, i.e. 4.7 A
double step, i.e. two steps with half the unit cell height each. Furthermore, terraces are also
traversed by half-unit cell height steps. These steps dene elongated holes in the terraces
with long axes of the holes along the [0 0 1] direction. Often these elongated holes terminate
at full unit cell height steps (step I). These elongated holes, and pairs of steps predominantly along the [0 1 0] direction, give the surface a zebra pattern. This step morphology
can be understood in terms of autocompensated step edges, i.e. step edges that exhibit
atom constellations that satises one of the two preferred valencies of Sn. Fig. 23(a)
and (b) shows ball-and-stick models for half-unit cell height step edges along the [0 1 0]
and [0 0 1] directions, respectively. Formation of such step edges results in vefold coordinated Sn atoms on the lower terrace of the steps for both step edge orientations. For the

84

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 23. Ball-and-stick models of step edges on the (1 0 0) surface. Small, white balls represent Sn- and large, dark
balls represent O-atoms. (a) Step edges oriented along the [0 1 0] crystallographic direction. Autocompensated step
edges form half-unit cell height step edges with vefold coordinated Sn4+ and twofold coordinated O2 atoms at
the step edge. (b) Step edges oriented along the [0 0 1] crystallographic direction with half-unit cell height steps.
These steps are not autocompensated in this direction and therefore are not observed experimentally. (c) Full-unit
cell height steps along the [0 0 1] direction. These steps form (0 1 0) microfacets and resemble the structure of the
most stable reduced (1 1 0) surface (see Fig. 25b).

[0 1 0] step edges these vefold coordinated Sn atoms are bonded to one twofold coordinated O atom. In other words, the same number of Sn to O bonds as O to Sn bonds were
broken to form this step edge and therefore both O as well as Sn can maintain their oxidation state. For the [0 0 1] step edge this is not the case, only Sn to O bonds are broken
and all oxygen atoms at the step edge are still fully threefold coordinated. Therefore the
oxidation state of the vefold coordinated Sn atom at the [0 0 1] step edge is not compensated, i.e. from a charge counting point of view the step-edge Sn atoms cannot attain a
nominal 2+ or 4+ oxidation state if all the oxygen atoms remain 2. Consequently, most
step edges of half a unit cell height are oriented along the [0 0 1] direction. It has been speculated [202] that the elongated holes form by the decomposition, i.e. desorption of oxygen
and tin, from the surface layer at elevated temperatures and that because of the unfavorable constellation of [0 0 1] step edges such a decomposition mainly propagates normal to
this direction and thus causing the zebra-pattern at the surface.
The situation is dierent for steps of a full unit cell height (step-I type). Here step edges
along the [0 0 1] direction are clearly favored. A model for such a step edge is illustrated in
Fig. 23(c). This step edge can be viewed as a (0 1 0) microfacet, i.e. a face identical to the
(1 0 0) surface. It is also intriguing that this step edge exhibits the same structure as a unit
cell of the lowest energy reduced (1 1 0) surface (see Section 4.2.2) and this may explain its
apparent stability. The unique feature of this step edge is that there are no dangling bonds
if we accept that the surface prefers to be Sn2+, i.e. all Sn and O atoms at the step edge
have the same coordination as atoms on terraces. The fact that this step edge maintains

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

85

a SnO stoichiometry rationalizes this favored step edge orientation. It also explains the
experimental observation that always full unit cell height steps are favored along the
[0 0 1] direction although half-unit cell steps would give a similar terrace termination.
4.2.1.3. The SnO2(1 0 1) surface. The (1 0 1) surface has also not been very often investigated. There are only two experimental studies of clean, single crystal (1 0 1) surfaces
[202,228]. The (1 0 1) surface is however also the termination obtained for epitaxial lms
grown on Pt(1 1 1) and Al2 O3 
1 0 1 2 substrates (see Section 6) and is the dominant surface
for SnO2 nanobelts (see Section 9).
Preparation of single crystal (1 0 1) surfaces by sputtering and annealing in a range of
6001000 K results in a 1 1 LEED pattern. Fig. 24 shows LEED and STM measurements. In high resolution STM images shown in Fig. 24(a) the unit cell of the (1 0 1) surface
can be easily discerned. It consists of bright protrusions in the corner of a rectangle and a
bright protrusion in its center. The center protrusion appears however asymmetric with
respect to the symmetry axis of the rectangle described by the corner protrusions. The protrusions are consistent with positions of Sn atoms in an unreconstructed 1 1 bulk termination of the SnO2(1 0 1) surface. The asymmetry of the center protrusion is due to the

Fig. 24. Surface structure of the SnO2(1 0 1) surface. (a) Atomically resolved STM image with the unit cell
indicated. (b) 1 1 LEED-pattern of SnO2(1 0 1). Note the missing spots indicating the glide-plane symmetry of
the unit cell. (c) Ball-and-stick model of the surface with the unit cell indicated. (d) Large scale STM image of the
surface showing the high density of step edges with preferential step edge orientation along the [1 0 1] direction.

86

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

asymmetric coordination of the centered Sn atom. Models for the atomic structures of a
stoichiometric surface and a surface with a SnO composition are displayed in Fig. 16(d).
Exposure of this surface to 10 mbar O2 does not result in a change of the qualitative
1 1 LEED pattern under most conditions. Sometimes, however, a 2 1 LEED pattern
was observed, that converts in a 1 1 pattern upon vacuum annealing. Attempts to form
this 2 1 pattern reproducibly were unsuccessful and thus the experimental conditions
necessary for the formation of this structure are unclear.
In LEIS the O/Sn ratio increases after high pressure oxygen exposure. Fig. 21(c) shows
temperature dependent LEIS of an initially oxidized sample. An increase of the Sn peak is
observed in a temperature range of 550650 K. This indicates desorption of oxygen from
the surface.
On vacuum prepared (1 0 1) surfaces a much higher step-edge density is observed compared to the (1 0 0) or the vacuum prepared (1 1 0)-4 1 surfaces. Often very narrow terraces of only a couple of unit-cell width are observed (Fig. 24(d)). The preferential
orientation of the step edges is along the 
1 0 1 direction.
4.2.2. Surface structure and composition II: Theoretical
Ab initio atomistic thermodynamic calculations were performed to evaluate the surface
energy of dierent surface structures as a function of the oxygen chemical potential. For
all surface orientations studied, the stoichiometric, i.e. SnO2, bulk termination was the
energetically favored surface only at high oxygen chemical potentials (see Table 4 for
the surface energies of stoichiometric surfaces). At lower oxygen chemical potentials oxygen depleted surface phases become energetically favored. This is in agreement with the
experimental LEIS studies that showed compositional transitions of the surfaces upon
vacuum annealing. In the following the structures and thermodynamic stability of dierent
surface phases of the (1 1 0), (1 0 1) and (1 0 0) crystallographic surface orientations are
presented.
4.2.2.1. SnO2(1 1 0). The SnO2(1 1 0) shows the most complex behavior of the three surface
orientations studied. Experimentally it has been found that it reduces at the lowest temperature under vacuum but a high degree of disorder observed in STM studies made it impossible to derive a reliable surface model [209]. In order to nd a likely surface structure for
low oxygen chemical potential dierent reduced surface structures were evaluated by DFT
calculations. These structures included previously proposed structures for 4 1 reconstruction with periodically removed in-plane oxygen atoms [206], various dierent
arrangement of removed surface oxygen atoms, as well as constellations of threefold coordinated Sn atoms in non-bulk positions. Under all these considered models the two structures that showed lowest surface energies in some ranges of the oxygen chemical potential
are the stoichiometric surface at high oxygen potential and a surface that in addition to
removed bridging oxygen atoms also has every other row of in-plane oxygen atoms
removed for low oxygen chemical potential. This structure was found independently by
two research groups [201,228]. These two structures and the dependence of their surface
energy on the oxygen chemical potential is shown in Fig. 25. The reduced surface structure
leaves threefold coordinated tin and oxygen atoms at the surface and a surface layer with
SnO composition. Thus Sn is attaining a 2+ oxidation state. Furthermore, such a surface
structure still exhibits a 1 1 unit cell. This surface phase becomes favored over the stoichiometric surface at a chemical potential 2.4 eV. It is worth pointing out that the sur-

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

87

Fig. 25. Surface structure of (a) the stoichiometric and (b) the reduced SnO2(1 1 0) surfaces. The dependence of
the surface energy on the oxygen chemical potential for the dierent bulk termination is shown in (c).

face energy is lowest if every second row of in-plane surface oxygen atoms is removed
instead of alternatingly removing oxygen atoms from adjacent rows. This thermodynamic
result is consistent with vacancy formation energies calculated by Oviedo and Gillan [229].
They showed that the energy for oxygen vacancy formation is lowest for in-plane oxygen
atoms if an entire oxygen row is removed. For such geometries the formation energy was
comparable to the energy it takes to remove bridging oxygen atoms. Thus the formation of
missing in-plane oxygen rows may contrary to an intuitive feeling not have a very high formation energy.
A half reduced surface termination with only slightly higher surface energy than the
fully reduced surface is also possible. The small energy dierence between the two surfaces
makes it possible that this surface is thermodynamically accessible. This surface has every
other in-plane oxygen row removed but retains its bridging oxygen rows. Thus this surface
consists of rows of threefold coordinated Sn2+ and rows of sixfold coordinated Sn4+.

88

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

In other communications a half reduced surface with only bridging oxygen atoms removed
was often assumed [205]. Such a structure, however, has a much higher surface energy and
thus appears to be thermodynamically unreasonable [228].
For annealing temperatures above 1000 K a 1 2 reconstruction forms. STM images
show that this structure is extremely defective (see Figs. 19 and 17(g) and (h)) and no
atomic scale features could be deduced from these images that would allow deriving a
model for this surface phase. It has been speculated before that this structure is isostructural to the 1 2 reconstruction observed on the rutile TiO2(1 1 0) surface. There are two
models for such a 1 2 structure, one model suggest a reconstruction consisting of missing
bulk units and removed bridging oxygen atoms [230], the other model has a Ti2O3 composition [231]. From the high density of defects in the 1 2 row structure it is evident that
neither model describes the SnO2(1 1 0)-1 2 reconstruction in its entire complexity. The
rst model consisting of missing bulk units and missing bridging oxygen atoms is unreasonable for SnO2 because we know that removal of bridging oxygen atoms alone is not
thermodynamically favored. Calculation for a SnO2(1 1 0)-1 2 structure, isostructural
to the Ti2O3 model, showed that the surface energy of this structure is high and does
not even drop below the surface energy for the stoichiometric surface for any oxygen
chemical potential. Therefore this structure should not form. This is in agreement with
the notion that only surfaces that satisfy either a 2+ or 4+ oxidation state for tin is stable.
Consequently, another surface structure for the 1 2 reconstruction has to exist that has
not yet been considered.
The c(2 2) and 4 1 structures were not considered in these calculations because, as
pointed out above, these structures form after oxygen depletion of the surface layer by
ion sputtering and are not thermodynamic equilibrium structures.
4.2.2.2. SnO2(1 0 0) and (1 0 1). The basic argument that surface structures that exhibit
either a SnO or a SnO2 composition are favored, also applies to the (1 0 0) and (1 0 1) surfaces. The dierence between these surfaces and the (1 1 0) surface is, however, that
removal of the twofold bridging oxygen atoms leaves threefold coordinated Sn atoms
and a surface layer of SnO composition. Thus, as shown in Fig. 16(c) and (d), simple bulk
terminations of the crystal exist that have surface tin atoms in either the 4+ or 2+ oxidation state. Consequently, these two surfaces can convert from a SnO2 stoichiometry to a
SnO composition by removal of bridging oxygen atoms. The absence of signicant surface
reconstructions has been veried experimentally by LEED and STM that show the same
unit cell for reduced and oxidized surfaces. Relaxation of the two surface terminations for
the (1 0 0) and (1 0 1) surfaces have been calculated by DFT calculations. Somewhat surprisingly the reduced surfaces show less surface relaxations compared to the stoichiometric
bulk terminations. The vertical relaxations compared to the bulk planes are given in Table
5. Ab initio atomistic thermodynamic calculations of fully relaxed surfaces conrm the
experimental observation of a dual surface termination. Fig. 26 shows that the oxygen-rich
surfaces are preferred at high oxygen chemical potential while the reduced surfaces become
favored if the oxygen chemical potential is reduced. For the (1 0 1) surface an intermediate
structure is found with a 2 1 superstructure consisting of alternating rows of bridging
oxygen and rows with the bridging oxygen atoms removed, i.e. alternating double rows
of Sn4+ and Sn2+. Such a structure has the lowest surface energy in a narrow range of oxygen chemical potentials and this is consistent with the infrequent observation of a 2 1
LEED pattern after high pressure treatment. The phase transition chemical potentials

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

89

Table 5
Calculated relaxations of the atomic planes normal to the surface compared to the bulk position for the
stoichiometric and reduced (1 0 0) and (1 0 1) surfaces (adapted from [202,228])
(1 0 0)

D12
D23
D34
D45
D56
D67
D78
D89
D910

(1 0 1)

Stoich. (%)

Red. (%)

Stoich. (%)

Red. (%)

13.9
6.9
24.6
3.7
1.8
5.7
1.5
0.3
0.2

6.3
30.3
3.1
3.2
8.9
1.8
0.4
0.09

13
13.4
12.7
9.6
1.3
0.2
0.4

4.5
5.2
0.2
0.7
1.1
1.0

Fig. 26. Surface energy as a function of the oxygen chemical potential for stoichiometric and reduced surfaces for
(a) the (1 0 0) and (b) the (1 0 1) surfaces of SnO2.

can be read o in Fig. 26. According to this gure the (1 0 1) surface reduces at somewhat
higher chemical potential (1.8 eV) than the (1 0 0) surface (2.1 eV) but both surfaces
reduce at higher oxygen chemical potentials than the (1 1 0) surface (Fig. 25(c)).
4.2.3. Surface structure and composition III: comparison of experimentally
and theoretically determined phase transitions
Ideally one would hope that the experimentally determined reduction temperatures
could be related to calculations that determine the thermodynamic equilibrium structure.
However, the temperature programmed ion scattering experiments measure the kinetic
desorption barriers, similar to the more commonly employed temperature programmed
desorption (TPD) studies. Thus it is unsurprising that the experimentally determined
transition temperature for the (1 1 0), (1 0 1), and (1 0 0) surface follow not the same trend
as the transition temperatures deduced form ab initio thermodynamics calculations.
Although stoichiometric surface phases prepared at elevated pressures are stable (kinetically trapped) under UHV conditions, intermediated (half reduced) phases, which are
stable in a narrow regime of oxygen chemical potential cannot be accessed by vacuum
annealing of the stoichiometric surface. For instance this is the case for the
SnO2(1 0 1)-2 1 surface which could not be prepared by annealing of a fully oxidized
SnO2(1 0 1)-1 1 surface under UHV conditions. Instead the fully reduced surface has
been always obtained. For measuring thermodynamic transition conditions experiments

90

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

in a dynamic equilibrium with the gas phase at higher oxygen partial pressures need to be
conducted. Under such conditions the standard UHV surface science techniques are,
however, not employable. Alternatives to UHV experiments may be surface X-ray diffraction or high pressure STM studies. Such experiments have yet to be reported for
SnO2 surfaces.
4.3. Conclusions for surface structure and composition
All three investigated low index SnO2 surfaces experience loss of surface lattice oxygen
under reducing conditions. Surfaces with all surface Sn atoms reduced to a Sn(II) state
were found to be the preferred bulk termination for low oxygen chemical potential. For
the (1 0 0) and (1 0 1) surface STM and LEED showed that such reduced surfaces are
formed by removal of bridging oxygen atoms from the stoichiometric bulk termination.
All atoms remain in their bulk positions with some relaxations normal to the surface.
On the (1 1 0) surface, on the other hand, just removal of the bridging oxygen rows results
in a surface of relatively high surface energy for all oxygen chemical potential and is thus
rendered to be thermodynamically inaccessible. This nding contradicts early interpretations that assumed that such surfaces could be prepared. The most stable (1 1 0) reduced
surface was found to be a surface with every other row of in-plane oxygen atoms removed
in addition to the bridging oxygen rows. This structure satises the condition for every surface Sn atom to be in a 2+ charge state. From STM studies of the (1 1 0) surface it is
known, however, that such idealized surfaces do rarely form and surfaces with high defect
concentrations are observed. Furthermore, at high annealing temperatures a 1 2 reconstruction forms, for which no satisfying structural model has been found yet. The often
studied 4 1 reconstruction can be prepared uniformly with large terraces by sputtering
and annealing cycles. This structure is highly oxygen decient and cannot be formed without sputtering procedures implying that some articial oxygen depletion is necessary to
form the 4 1 reconstruction.
The variation of the lattice-oxygen concentration at SnO2 surfaces has implications for
the use of this material as an oxidation catalyst [106], and as a sensing material for oxidizing and reducing gases. The consumption of lattice oxygen in oxidation processes on oxide
catalysts is a well-established mechanism (Marsvan Krevelen mechanism). For CO oxidation over RuO2 catalysts it has been shown that the best results should be obtainable
when the RuO2(1 1 0) surface is close to the phase boundary of an oxygen rich and poor
surface [223,224]. Under such conditions CO can easily adsorb and react with RuO2 surface oxygen atoms and at the same time the surface lattice oxygen can be replenished by
gas-phase oxygen. Similar mechanisms are likely to be at work for catalytic oxidation reactions on SnO2. Since the phase-transition conditions are dierent for the dierent surfaces
investigated one would expect that dierent crystallographic faces are catalytically active
under dierent conditions.
The catalytic activity of SnO2 may be closely related to its gas sensing properties for
reducing and oxidizing gases [232]. A change in the oxidation potential of the gas phase
results in a shift in the surface phase diagram towards an oxygen rich or poor surface.
These surface phases have dierent properties resulting in altered moleculesurface interactions (see e.g. Section 8.2.10). Manifestations of changed physical surface properties for
dierent surface phases, such like altered work function and the formation of electronic
surface states, are discussed in the next chapter.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

91

5. Electronic structure of tin oxide


A change in the surface electronic structure may be expected for a variation in the surface composition. As pointed out above reduction of the surface converts the surface Sn
cations from a (IV) to a (II) valence state. Thus the Sn 5s states that are predominantly
empty for Sn(IV) are becoming lled for Sn(II) and form part of the valence band. As
pointed out in Section 3.1 Sn 5s lone pair charge clouds are present in stannous oxide
and thus some similarities between the surface electronic structure of reduced SnO2 surfaces and bulk stannous oxide may be anticipated with Sn 5s lone pair electrons forming
states conned to the surface layer. Therefore a comparison of the bulk electronic structure
of stannous and stannic oxide may be instructive in order to understand how the Sn 5s
electrons, which do not contribute to the bonding in stannous oxide, inuence the valence
band. The bulk electronic structure of SnO is compared to that of SnO2 in the next section.
Thereafter, the surface electronic structure for dierent reduction states and surface orientation of SnO2 are presented. Photoemission studies are compared to band structure calculations for each surface orientation. Some general conclusions for the surface electronic
structure of SnO2 are given at the end of this section.
5.1. Comparison between the bulk electronic structure of SnO and SnO2
Comparison between SnO and SnO2 is complicated by the reliable preparation of SnO
and diculties in discriminating it from SnO2. Chemical shifts of Sn core levels between
Sn2+ and Sn4+ are small and often it is assumed that both valence states are indistinguishable by core level photoemission (see also Table 3). Lau and Wertheim [233] proposed that
such a negligible chemical shift is observed because the change in the free ion potential
between Sn2+ and Sn4+ is cancelled by the change in Madelung potential at tin sites in
SnO and SnO2. However, there also exist reports of a sizeable shift of 0.50.7 eV of the
Sn 3d core level between the two tin oxides (see Section 3.3). Nevertheless, the probably
most reliable way to distinguish SnO from SnO2 is by comparing their valence band spectra or by measuring the energy separation between the Sn 4d5/2 peak and the leading edge
of valence band [194]. For SnO2 a separation of 21.121.4 eV and for SnO of 22.423.7 eV
is reported. This increase in energy separation for SnO is mainly due to the smaller band
gap of SnO compared to SnO2. Fig. 27 shows XPS spectra of the valence band and the
Sn 4d core level for a SnO sample and for two SnO2 samples. The closing of the band
gap for SnO and the associated increase in the separation between the valence band leading edge and the Sn 4d core level are clearly observed. The additional state at the valence
band maximum for SnO is attributed to lone pair Sn 5s states. This is conrmed by density
of states calculations [234]. The total density of states of SnO is characterized by four
peaks labeled AD in Fig. 28. These four structures are mainly O 2p states, which are
hybridized with Sn-p and Sn-s states (note the dierence in scale in the partial DOS and
the almost negligible contribution of t2g-Sn, eg-Sn, and 2s-O curves). Structure A corresponds mainly to hybridization with Sn 5s states, structure B also has some 5s character,
while structure C results from hybridization between O 2p and Sn 5p states. The last state
D has signicant 5s character.
In contrast to SnO, SnO2 does not exhibit any Sn 5s character at the VBM. Fig. 29(a)
shows DOS calculations for SnO2 [235,236]. Three energy regions can be dierentiated.
The VBM is mainly of O 2p character, while the center region results form hybridization

92

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 27. Photoelectron spectra of the Sn 4d core level and the valence band of SnO and two SnO2 samples. A
chemical shift of 0.5 eV is observed for the Sn 4d core level. For SnO an additional state at the valence band
maximum is observed (from Ref. [194],  (1992) American Physical Society).

Fig. 28. Density of states calculations for SnO (from Ref. [234],  (1999) American Physical Society).

of Sn 5p with O 2p, and only the bottom of the valence band has some Sn 5s character.
Experimental results for a SnO2(1 1 0) surface are also shown in Fig. 29(a) for comparison.
Fig. 29(b) shows the orbital character of the valence band and conduction band more
schematically. It can be seen from this representation that for SnO2 the empty Sn 5s orbital
make up the bottom of the conduction band. As has been outlined in Section 2.1, this
s-character of the conduction band is one of the reasons why SnO2 makes a good transparent
conductor. For a more detailed band structure calculation for SnO2, see also Fig. 3(a).
With these general properties for the bulk electronic structure in mind we now turn to
the electronic structure of stoichiometric and reduced low index SnO2 surfaces.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

93

Fig. 29. Density of states for SnO2. (a) Valence band calculations for bulk SnO2 (solid line) and convoluted with
a Gaussian of HWHM 0.53 eV (dashed line). Middle line indicates the local density of states of a stoichiometric
(1 1 0) surface. The top curve represents an experimental photoelectron spectrum. (b) Schematic illustration of the
contribution of the hybridization of Sn and O molecular orbitals for the valence band and the conduction band
(from Ref. [236,235],  (1990) and (1983) by the American Physical Society).

5.2. Surface electronic structure of SnO2


As pointed out already, most experimental as well as theoretical studies of the electronic
structure of SnO2 surfaces have been focused on the (1 1 0) surface. Experimentally, most
photoemission studies were conducted on samples prepared by vacuum preparation only
and consequently on reduced surfaces. Oxidation procedures in order to obtain stoichiometric surfaces were only included in studies by Cox and co-workers for the (1 1 0) surface
and recent synchrotron radiation studies by the authors of this review for (1 1 0), (1 0 1),
and (1 0 0) surfaces [228]. In particular for the (1 0 1) and (1 0 0) surfaces sample preparation
procedures allowed comparison between the stoichiometric Sn(IV) and reduced Sn(II) surfaces. The expected Sn-derived surface states have been identied for all reduced surfaces.
For the (1 1 0) surface the electronic structure is strongly dependent on the preparation
conditions, conrming the observations in STM that complex defect structures may exist
on the (1 1 0) surface.
In the following we summarize experimental and theoretical results for the electronic
structure for the three investigated low index surfaces. For each surface we present
photoemission studies followed by band structure calculations.
5.2.1. SnO2(1 1 0)
In Section 4 it has been shown that the (1 1 0) surface exhibits intricate surface reconstructions and surfaces that exhibit a 1 1 diraction pattern can have a high density of
defects. This complexity of the (1 1 0) surface is also observed for its electronic structure.

94

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Dierent valence band structures are obtained depending on the preparation conditions,
even for apparently the same 4 1 surface reconstruction the photoemission may be different depending on the preparation route. The surface electronic structures for dierent
preparation procedures and the band structures calculated for the most stable surface
structures discussed in Section 4 are summarized in this section.
5.2.1.1. Preparation dependent valence band structure. Early photoemission studies on
SnO2(1 1 0) were performed by Themlin et al. [236]. However, the SnO2 samples were prepared by vacuum procedures and the 1 1 structure obtained after high temperature
annealing has been shown in recent STM studies to be more likely a reduced, defective surface than a bulk termination [209]. The studies by Themlin et al. and more recent photoemission studies [228,237] of the vacuum prepared 4 1 reconstruction as well as highly
defective surfaces prepared by ion sputtering did only show weak photoemission from
within the 3.6 eVband gap of SnO2 with a weak shoulder close to the bulk valence band
maximum. This may be surprising since it is known from LEIS studies that the 4 1 structure has little surface oxygen and sputtered surfaces are heavily oxygen depleted. For such
strongly reduced surfaces one may expect stronger states within the band gap due to occupied Sn 5s levels, similar to SnO.
The degree of reduction of the SnO2 surfaces was judged form the extent of the shoulder
at the VBM using a He(I) discharge lamp by Egdell et al. [238,239] and Cox et al.
[240,241]. Cox showed that the shoulder could be eliminated from the band gap by oxidizing the sample in 1 Torr of O2. Surprisingly, synchrotron UPS studies showed that strong
photoemission intensity from the bulk band-gap region can be generated by mildly oxidizing a (1 1 0) sample or by annealing an oxidized surface in UHV. This implies that the
defects present on an incompletely oxidized surface or a surface reduced by vacuum
annealing are dierent from defects generated by ion sputtering or on a well-ordered
4 1 reconstructed surface. Fig. 30 shows the electronic structure obtained for various
preparation conditions. It is apparent that one can obtain very dierent results depending
upon the sample treatment. In Fig. 30(a) the sample was prepared by standard vacuum
preparation, i.e. several cycles of sputtering and annealing to 700 C. Consequently
the sample exhibited a 4 1 LEED pattern. In Fig. 30(b) the sample was exposed to
20 mbar O2 at a sample temperature around 300 C. This sample exhibited a 1 1 LEED
pattern. Subsequent annealing in UHV to 600700 C resulted in the reformation of a
4 1 LEED pattern and the corresponding PES is shown in Fig. 30(c). Oxidation at oxygen pressures of 100 mbar and at 300 C sample temperature caused a disordering of the
surface that resulted in the disappearance of any LEED pattern. This disordering is also
apparent from the broad, poorly dened valence band spectra shown in Fig. 30(d).
The most obvious changes in the valence band spectra of these surfaces are observed in
the band gap region. Only surfaces prepared at high oxygen pressures (Fig. 30(d)) exhibit a
bulk-like band gap with only residual photoemission from within the band gap itself. This
residual photoemission intensity may result from surface defects or disorder. The observation that a bulk-band-gap-like surface forms only after exposure to high oxygen pressures
is in agreement with UPS studies by Cox et al. [210].
Surfaces oxidized under milder conditions at somewhat lower oxygen pressures exhibit
photoemission from the band-gap region (Fig. 30(b)). The maximum emission of the band
gap states is around 2.3 eV and extends up to 1 eV below the Fermi edge. One can only
speculate what exactly the surface structure is that causes these strong band-gap emissions.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

95

Fig. 30. Photoemission spectra for dierently prepared SnO2(1 1 0) surfaces acquired with photon energies of
20 eV (left panel) and 29 eV (right panel). (a) Surface prepared by cycles of sputtering and vacuum annealing to
700 C. This surface exhibits a 4 1 LEED pattern. (b) Surface oxidized in 20 mbar O2. A 1 1 LEED pattern
is observed. (c) Surface in (b) annealed to 700 C in vacuum. This annealing procedure reformed a 4 1 LEED
pattern. (d) Surface oxidized at above 100 mbar O2 pressure (from Ref. [228],  (2005) by the American Physical
Society).

It is clear that it is not a stoichiometric surface and the surface structure that is assumed to
be the most stable reduced surface consists of missing bridging and in-plane oxygen rows
(see Section 4.2.2). Such a surface has a SnO surface composition and as such one could
intuitively expect surface states similar to states of bulk tin monoxide, i.e. with a Sn 5s
derived state around 2 eV below the CBM.
5.2.1.2. 4 1 reconstructed surfaces. It is intriguing that the photoemission intensity from
the band-gap region can be very dierent for surfaces apparently exhibiting the same 4 1
reconstruction. Samples prepared by sputtering and annealing alone (Fig. 30(a)) exhibit
weak emission from the band-gap region. This has also been observed by Themlin et al.
and Sinner-Hettenbach et al. [236,237]. The surface can be described as almost metallic
with states occupying nearly the entire band gap close to the Fermi edge. SnO2(1 1 0)4 1 reconstructed surfaces obtained by vacuum annealing of an oxidized sample
(Fig. 30(c)), on the other hand, show a much stronger emission from the bulk band-gap
region, but a small band gap of about 0.5 eV remains.
Dierences in the Sn/O ratio for the dierently prepared surfaces have been discussed in
Section 4.1 (see also Fig. 20). Fig. 31 shows PES for the binding energy region from 33 eV
to the Fermi edge for the same samples used for obtaining the spectra shown in Fig. 30(a)
(c). These spectra include the Sn 4d and weak O 2s core levels [216], as well as the VB
region. The spectra were normalized to the maximum intensity of the mainly O 2p derived
VB. Strong variations in the intensity of the Sn 4d level are observed. The samples sputtered and annealed in vacuum show the strongest Sn 4d emission (Fig. 31(a)) consistent
with a strong depletion of surface oxygen. SnO2(1 1 0)-4 1 reconstructed samples

96

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 31. Wide scan photoemission spectra acquired with a photon energy of 65 eV for three dierently prepared
samples: (a) 4 1 reconstructed surface, prepared by sputtering and annealing (see also Fig. 6(a)). (b) 4 1
reconstructed surface, prepared by vacuum annealing of an oxidized surface (see also Fig. 6(c)). (c) 1 1 surface
prepared by oxidizing in 20 mbar O2. The spectra are oset for clarity and normalized to the valence band
maximum (from Ref. [228],  (2005) by the American Physical Society).

prepared by vacuum annealing of an oxidized surface show a much reduced Sn 4d emission, indicating that although it shares the same reconstruction as the sputtered and
annealed surface it possesses a lower Sn/O ratio. It is somewhat counterintuitive, though,
that the more reduced 4 1 surface (Fig. 31(a) and Fig. 30(a)) exhibits less emission from
the band-gap region than the less reduced surface shown in Figs. 31(b) and 30(c). These
dierences in the Sn/O ratio for dierently prepared 4 1 reconstructed surfaces may indicate that the 4 1 structure is not just a reconstruction conned to the surface layer but
may involve underlying layers as well. As expected, the Sn 4d emission is smallest for
the oxidized sample (Fig. 25(c)). Additionally, a small shift of 0.25 eV for the Sn 4d peak
to higher binding energy is observed. This may be explained by a small chemical shift in
the 4d core levels as is expected for Sn in a 4+ oxidation state compared to a 2+ oxidation
state.
5.2.1.3. Resonant photoemission studies. Using resonant photoemission, i.e. swiping the
photon energy through the Sn 4d ) 5p excitation threshold, Sn-derived states in the
valence band can be identied. Fig. 32 illustrates this resonance mechanism. The Sn 5s
and Sn 5p states in the bulk valence band were identied in agreement with the theoretical
predictions shown in Fig. 29. Additionally, the weak shoulder at the VBM on the sputtered surface was associated with a Sn-derived state, suggesting that it is of Sn 5s character
[236]. In a similar study the VB spectra for photon energies in a range from 20 eV to 35 eV
were acquired for 4 1 reconstructed samples prepared by repeated cycles of sputtering
and annealing (Fig. 33(a)), and for samples prepared by annealing in 20 mbar O2
(Fig. 33(b)). The photoemission intensity for normalized VB spectra are plotted in
Fig. 33(c) and (d) for 4.7 eV and 3 eV binding energy. For both samples the photoemission
intensity of the band-gap states shows an increase for photon energies between 28 eV and
34 eV. Such an increase is consistent with a resonant behavior involving the Sn 4d ) Sn 5p

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

97

Fig. 32. Schematic energy-level diagram for SnO2 and illustration of the resonant photoemission processes. (a)
Absorption, dipole allowed transition Sn 4d ) 5p. (b) Direct recombination: an electron from tin 5s levels, for
example, has been ejected. (c) Direct photoemission from tin 5s levels. Note that the nal energy is the same for
the latter two processes. (d) Recombination by Auger decay (from Ref. [236],  (1990) by the American Physical
Society).

transition described above. Thus it is concluded that the band-gap states on the (1 1 0) surface are Sn-derived. The intensity plots for the VBM, i.e. a binding energy of 4.7 eV, do
not show an enhanced emission in the region of a possible resonance for Sn states. Consequently, the states at this binding energy appear to be mainly O 2p derived in accord with
the bulk DOS calculations (see Fig. 29).
One interesting observation is a strong resonance for the 4 1 reconstructed surface for
photon energies between 28 eV and 31 eV at a binding energy of 9.6 eV (indicated by
arrows in Fig. 33(a)). This resonance is only seen if the surface exhibits a 4 1 reconstruction (see also Fig. 30(a) and (c)) and has not been observed for the SnO2(1 0 1) or (1 0 0)
surfaces (see below) [228]. Similar observations are reported in another resonant photoemission study using photon energies sweeping through the dipole allowed Sn 4p ) Sn 5s
threshold [237]. In this study strong resonances were observed for the 4 1 reconstructed
surface, but not on a sputtered sample. This is shown in Fig. 34. It is interesting to point
out that the proposed resonance mechanism for the latter study only holds if there are
Sn 5s states in the conduction band, because of the dipole selection rules of the initial excitation process. This implies that the 4 1 structure has empty Sn 5s states and thus cannot
be SnO-like (see Fig. 28). It was proposed that the observed resonances may originate
from the SnO2 substrate, but this is unlikely, since it is only observed on (1 1 0)-4 1 reconstructed samples, and neither on the sputtered nor the oxidized samples or SnO2 samples
with dierent surface orientation. It is also possible that the resonances observed on the
4 1 reconstructed surface have a dierent origin that has not been well explained yet.
5.2.1.4. Band structure calculations. Electronic structure calculations have been performed
for stoichiometric and reduced (1 1 0) surfaces [235,242]. For the stoichiometric surface, i.e.
a bulk terminated surface including bridging oxygen atoms, surface states at the top of the
valence band have been predicted by calculations by Oviedo and Gillan [229] as well as by
Maki-Jaskari and Rantala [242,243]. It has been proposed that this surface state is formed
by p-orbital electrons of the bridging oxygen atoms and results in a at band about 0.8 eV

98

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 33. Photon-energy dependence of valence band spectra for an energy range between 20 eV and 35 eV for
(1 1 0) surfaces. The spectra are normalized to the valence band area. In (a) valence band spectra for a 4 1
reconstructed surface, obtained by sputtering and annealing, are shown. In (b) the valence band spectra of a
sample prepared by oxidation in 20 mbar O2 are displayed. (c) and (d) show the photoemission intensity as a
function of photon energy for the binding energies indicated by the colored lines in (a) and (b) respectively (from
Ref. [228],  (2005) by the American Physical Society). (For interpretation of colours in this gure legend, the
reader is referred to the web version of this article.)

above the bulk valence band maximum. Removal of bridging oxygen atoms causes formation of defect states within the band gap. At rst for low densities of bridging oxygen
vacancies a separate peak at the bottom of the band gap appears. Eventually, with all
the bridging oxygen atoms removed, a broad distribution of defect states lls the entire
band gap associated with fourfold Sn2+ ions with in-plane oxygen p contribution
[243,244]. Further reduction of the surface by formation of in-plane oxygen vacancies

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

99

Fig. 34. Resonant photoemission from (a) a SnO2(1 1 0)-4 1 and (b) sputtered SnO2(1 1 0) surface for photon
energies between 60 and 130 eV. Resonances are observed for the 4 1 reconstructed surface but not for the
sputtered surface. The photoemission intensity for binding energies marked AD are plotted in (c) and (d) for the
4 1-reconstructed and sputtered surface respectively (from Ref. [237],  (2002) Elsevier).

has also been investigated. In this case, surface states with Sn2+ charge hats are formed
[243], these states have a tendency to lie near the middle of the band gap [244].
Results from band structure calculations for three dierent bulk terminations are presented in Fig. 35 [228]. The stoichiometric SnO2(1 1 0) surface exhibits surface states within
the bulk band-gap region (Fig. 35(a)), in agreement with the above-mentioned calculations. These two states are nearly at, with one just above and the other 0.20.3 eV above
the bulk VBM. In the experimental studies described in the previous section and in those
by Cox et al. [210] no such surface state for an oxidized (1 1 0) surface was observed. It
should be noted, however, that the fully oxidized surface prepared in this study was very
disordered, as indicated by broad, poorly dened valence band peaks. In addition the surface band is predicted to be separated from the VBM by a small amount only and hence
may be dicult to identify in the experimental studies.
Fig. 35(b) shows the band structure calculations for a surface with removed bridging
oxygen atoms only. Such a surface would have a strongly dispersed surface state that
extends 2 eV above the bulk VBM at the M-point of the surface Brillouin zone. Other
reported DFT calculations for the same surface structure show the same features [242].
However, the high surface energy of this structure makes it an unlikely candidate (see Section 4.2.2). Also in ARUPS studies no dispersion of band-gap states was observed [228],
clearly indicating that such a surface structure has not been prepared even for surfaces that
exhibited a (1 1) LEED pattern.
The electronic structure of the surface with a Sn2+O2 composition that has been identied as the most stable under reducing conditions (see Section 4.2.2) is shown in
Fig. 35(c). This structure also exhibits surface bands in the bulk band-gap region, but they
only extend less than 1 eV into the bulk band gap from the VBM. Thus there is some
resemblance between the electronic structure calculations for this surface and the PES
of the sample prepared at 20 mbar O2 (Figs. 30(b) and 33(b)). Consequently, the agreement between the measurements and the electronic structure calculations suggests that this
structure may be indeed signicant for the reduced SnO2(1 1 0) surface.

100

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 35. Surface band structure calculations for bulk terminations indicated by ball-and-stick models on the right.
Gray shaded area represents the projection of the bulk bands on the surface and red-dotted lines indicate the
surface bands (from Ref. [228],  (2005) by the American Physical Society).

5.2.2. SnO2(1 0 1)
Secondary electron cuto measurements show that the work function of the reduced
(1 0 1) surface is 1.0 eV lower than for the stoichiometric (1 0 1) surface. The experimental
values for the reduced and stoichiometric surfaces are 4.7 eV and 5.7 eV respectively.
These values are in good agreement with values calculated by DFT of 4.7 eV and 6.1 eV
respectively [228]. Such a pronounced dierence in the work function is consistent with

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

101

surfaces that exhibit two dierent bulk terminations with dierent oxygen contents. The
somewhat lower experimental value for the stoichiometric surface compared to the
DFT results may indicate the presence of oxygen vacancies at the surface, i.e. an incompletely oxidized surface.
Normal emission and angle-resolved photoemission data were collected for the VB for
reduced and stoichiometric surfaces in order to identify surface states and to characterize
their band dispersion. These data are presented in the next two subsections, followed by a
comparison of the experimental results with theoretical band structure calculations.
5.2.2.1. Resonant photoemission. Fig. 36(a) shows normal emission PES for photon energies between 24 eV and 40 eV. The spectra were normalized to the integrated area of
the photoemission intensity in the VB region with a binding energy between 0 eV and
12 eV. Comparison of the VB spectra of the reduced and stoichiometric SnO2(1 0 1) surfaces show signicant dierences at the top of the VB. The reduced surface exhibits a state
within the bulk band gap of SnO2 that extends up to a binding energy of 2 eV with a maximum at 2.8 eV. This band gap state does not disperse as a function of photon energy,
which is indicative of a surface state. For the oxidized surface only a very small intensity is
detected in this range. Additionally, there appears to be a higher photoemission intensity
at a binding energy of 4.7 eV, i.e. at the bulk VB maximum (VBM), for the reduced surface compared to the stoichiometric surface. Resonant photoemission has been used to
identify the origin of the increased intensities for the reduced surface in the band gap
and VBM regions. Fig. 36(b) and (c) show plots of the photoemission intensities at binding
energies of 4.7 eV and 2.8 eV, respectively, as a function of photon energy. Increased photoemission is observed for photon energies between 30 eV and 36 eV for the reduced surface for both binding energies. The residual intensity in the band-gap region at 2.8 eV for
the stoichiometric surface also shows an increased intensity for the same photon energies
(Fig. 36(c)), which suggests that this residual emission has the same origin as the photoemission of the reduced surface and thus suggests an incomplete oxidized surface.
The increased intensities at 3036 eV can be explained by resonant photoemission of
Sn-derived states as the photon energy is ramped through the threshold value for the
dipole-allowed Sn 4d ) Sn 5p transition, as has been illustrated in Fig. 32. Thus resonant
PES implies the presence of two Sn-derived surface states at the top of the VB for the
reduced (1 0 1) surface, which are not present for the stoichiometric surface.
5.2.2.2. Surface band dispersion measurements. In order to characterize the dispersion of
the two Sn-derived surface states for the reduced SnO2(1 0 1) surface, angle-resolved
PES were acquired along the CX, CY, and CM directions of the rectangular surface
Brillouin zone. For comparison the same angle-resolved data were collected for the oxidized surface. Fig. 37 shows angle-resolved data along the three low symmetry directions
for photon energies of 31.5 eV for the entire valence band region, and at 20 eV for the
band-gap region alone [228]. PES for stoichiometric as well as for reduced surfaces are
displayed.
The band dispersion obtained from the angle dependent photoemission is plotted on the
right panel of Fig. 37. In order to t the symmetry of the band dispersion to the Brillouin
zone dimensions, an eective electron mass of 0.7 times the free electron mass was used. A
reduced zone scheme is also shown on the right side together with theoretical predictions
of surface states that are discussed next.

102

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 36. Dependence of valence band spectra of the SnO2(1 0 1) surface on photon energy. (a) VB spectra for
photon energies in a range of 2440 eV for stoichiometric (dashed-red lines) and reduced (solid-black lines)
surfaces. All spectra are normalized to the valence band area in a range between 0 and 12 eV binding energies. (b)
Comparison of photoemission intensities for the reduced (black squares) and stoichiometric surface (red
triangles) at 4.7 eV binding energy. An increased photoemission intensity is observed for the reduced surfaces
compared to the stoichiometric surface at photon energies between 30 and 36 eV. This is explained by a resonance
with the Sn 4d core level. (c) Similar to (b) but for a binding energy of 2.8 eV (from Ref. [228],  (2005) by the
American Physical Society).

5.2.2.3. Comparison between experiment and ab initio calculations. The only reported band
structure calculations reported for the (1 0 1) surface is shown in Fig. 38(a) and (b) for the
reduced and stoichiometric surfaces, respectively [228]. The binding energy in the calculations is referenced to the VBM, which is set to 0 eV. The bulk bands projected onto the
(1 0 1) surface are indicated by the gray shaded area and surface states and resonances
by the dotted lines. The calculated surface states for the reduced structure shown in
Fig. 38(a) are in good agreement with the experimental results as indicated by the solid

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

103

Fig. 37. Angle-resolved photoemission measurements for the reduced (solid lines) and stoichiometric (dashed
lines) SnO2(1 0 1) surfaces along low symmetry directions of the surface Brillouin zone. On the right the dispersion
of surface states for the reduced (1 0 1) surface are plotted. An eective electron mass of 0.7 times the free electron
mass has been assumed to t the symmetry of the surface Brillouin zone. A reduced zone scheme is shown on the
plots on the far right. The solid lines represent DFT-calculated surface states (from Ref. [228],  (2005) by the
American Physical Society).

lines in the right-hand panel of Fig. 37. The binding energies of the calculated bands are
shifted in order to match the experimental data points, as the band gap is not accurately
reproduced in DFT. The same shift has been used for all three symmetry directions.
Agreement between theory and experiment for the second state is not as good as for
the band-gap state, which may be explained by diculties in tracking this state in the
experiments due to overlap with strong bulk states.
Similar to the band structure calculations for the stoichiometric (1 1 0) surface calculations also predict surface states distinct from the bulk-valence band for the stoichiometric

104

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 38. Band structure calculations for (a) the reduced and (b) the stoichiometric SnO2(1 0 1) surface (from Ref.
[228],  (2005) by the American Physical Society).

(1 0 1) surface. But again, as for the (1 1 0) surface, these predicted surface states have not
been unambiguously identied in experiments.
5.2.3. SnO2(1 0 0)
In Section 4.2 we saw that the (1 0 0) surface can, similarly to the (1 0 1) surface, convert
from a stoichiometric 1 1 surface termination to a reduced 1 1 surface by removal of
twofold coordinated bridging oxygen atoms under reducing conditions. The reduced
(1 0 0) surface also attains a Sn2+O2 composition. As for the (1 0 1) surface, the reduced
structures are formed after annealing in UHV and the stoichiometric surface termination
is obtained by high-pressure oxygen treatment.
Fig. 39(a) shows normal emission PES for the stoichiometric and reduced (1 0 0) surfaces at a photon energy of 26 eV [228]. There is a pronounced increase in photoelectron
emission for the reduced surface in the region of the valence band maximum. This additional state at the VBM for the reduced surface was observed for all photon energies investigated, i.e. in the range of 2084 eV. This VBM state did not disperse with photon energy
in normal emission spectra, implying that it is a surface state. Polarization-dependent photoemission studies verify the two-dimensional nature of this state. Fig. 39(b) shows normal
emission spectra for the same reduced (1 0 0) surface acquired with incident light at 70 and
40, i.e. for s and p polarized light. For a 70 incidence angle with the vector potential A of
the incident light almost normal to the sample surface, emission from the state at the VBM
is almost completely suppressed, as is expected for a surface state. Thus it is possible to

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

105

Fig. 39. Normal-emission PES of the SnO2(1 0 0) surface. (a) Comparison of photoemission from a stoichiometric
(red circles) and a reduced (black squares) surface, measured with a photon energy of 26 eV and a photon
incidence angle of 45. (b) Photoemission from a reduced surface with predominately s-polarized (open circles)
and p-polarized (black squares) light (from Ref. [228],  (2005) by the American Physical Society).

observe a surface state located at the VBM, as for the (1 0 1) surface. Unlike the (1 0 1),
however, no second surface state higher up in the band-gap region could be identied.
No band structure calculations for this surface have been reported.
5.2.4. Homogeneity of surface electronic structure
STM measurements of all single crystal surfaces revealed an apparent modulation in the
surface layer on the nanometer scale. This is visible in almost all the STM images reported
in Section 4 and is particular apparent in Fig. 22. Because STM is probing the electronic
structure these modulations may reect local variations in the empty DOS (note that only
empty state imaging give stable tunneling conditions on n-type SnO2). The association of
local modulation in STM images with variations in the surface electronic structure is demonstrated in Fig. 40. In addition to STM, IV curves were recorded at each data point and
subsequently the current recorded at 1.4 V has been plotted in a 2D current map shown
in Fig. 40(b) [209]. With this method regions were identied that exhibit dierent tunneling
conditions. If the surface modulations had a topographic origin no modulations in the 2D
current map would be expected. Consequently there exist nanoscopic changes in the density of empty states that give rise to the modulations in STM. It is known that, for
instance, subsurface, charged dopant atoms cause local band bending in semiconductors
that result in contrast in STM images [245247]. Thus one may speculate that a similar
mechanism is at work on SnO2. A local downward band bending in the vicinity of a positive subsurface charge for instance would pull the conduction band down and increase the
tunneling probability which consequently results in an apparent upward bump in constant current STM images. The origin of such subsurface charges is however not revealed;
ionized oxygen vacancies or Sn interstitials may give rise to an electrostatic eld causing
the band bending. Local variations of the concentration of these intrinsic defects could

106

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 40. STM image (a) and 2D current map (b) of a 40 40 nm2 surface area of SnO2(1 1 0)-1 1 surface. The
current map is obtained by acquiring, simultaneous to the STM image, IV curves at each data point of the STM
image and subsequently plotting the measured current for a given voltage of the IV curves. The 2D current map
shown in (b) represents the tunneling current measured at 1.4 V. For illustration averaged IV curves at points
marked A, B, and C in the current map are also displayed. The variations in the current map and the IV curves
illustrate local variation in the surface electronic structure of SnO2 (from Ref. [209],  (2003), Elsevier).

thus explain the lateral changes in the electronic structure observed in STM. An alternative explanation for the bumpy appearance of atomically at terraces in STM images is
tip-induced band bending. The electrostatic eld of the tip causes an upward band bending
in the substrate under empty state imaging conditions. Strong band bending eects of up
to 0.7 eV have been observed due the eld of charged adsorbates on SnO2 (see Section
8.2.1). Thus a similar large band bending by the inuence of the eld of a scanning tunneling microscope tip may be expected. In this scenario the bumpiness and the variations in IV curves arises from local variations of tip-induced band bending. Pinning of
the Fermi level at defect levels in the band-gap region, for example defect states induced
by bulk-oxygen vacancies, may limit the amount of tip-induced band bending. Thus the
observation of variations in the band bending would also suggest local variations of the
defect concentration in the substrate.
In either case STM shows a variation of the electronic structure of SnO2 single crystal
surfaces on a nanometer scale. This implies variations in the surface reactivity and gas sensitivity of SnO2 on the same length scale.
5.3. Conclusions for the surface electronic structure of SnO2
For all three low index surfaces investigated, Sn-derived surface states were identied
for reduced surfaces with a Sn2+O2 composition. It was hypothesized that the Sn2+O2

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

107

composition of the reduced surfaces would exhibit electronic characteristics similar to that
of tin monoxide with a Sn 5s derived state at a binding energy of 2 eV. The greatest similarity was found for the reduced (1 1 0) surface with a Sn-derived state with a maximum at
2.5 eV and extending up to a binding energy of 2 eV. For the reduced (1 0 1) surface a
band-gap state was also found but at slightly higher binding energies of 2.8 eV. The
(1 0 0) surface exhibited surface states close to the bulk VBM of SnO2. Thus the surface
electronic structure clearly depends on the crystallographic face with surface states forming at dierent binding energies. Nevertheless for all three reduced surfaces, Sn-derived
states were found experimentally and in band structure calculations at the top of the
VB or within the bulk band gap of SnO2 and thus bear some resemblance to the band
structure of SnO.
There exist however a discrepancy between theory and experiment for the stoichiometric surfaces. The calculations indicate surface states for the stoichiometric (1 1 0) and
(1 0 1) surfaces that are split o from the bulk VB, but such states could not be identied
experimentally. The theoretically predicted states are due to the two lone pairs of the surface oxygen atoms, which are twofold coordinated. At this point it is not clear if experimental problems in preparing well-ordered stoichiometric surfaces are to blame for the
discrepancy or if the calculated energy eigenvalues for these oxygen lone pair states are
simply too high.
The surface states for reduced SnO2 surfaces and defect states at oxygen vacancies have
been observed to lie low within the band gap for all three surfaces investigated. This is different for example to the well-studied TiO2(1 1 0) surface [238]. For this surface it is well
known that oxygen vacancies at the surface introduce defect states close to the Fermi edge.
Although SnO2 shares the same rutile structure of TiO2 the dierences in the defect levels
is not surprising because of the very dierent electronic structure. For TiO2 the bottom of
the conduction band has a strong peak with Ti-3d character. Creation of an oxygen
vacancy causes a downward shift of Ti-d states into the band gap. For SnO2 the conduction band consists of a strongly dispersing Sn 5s level which gives rise to a rather at density of states in the conduction band. Munnix and Schmeits [248] pointed out that in this
situation the cation interactions are not sucient to move any states out of the conduction
band into the band gap and consequently no high lying defect states are observed for
SnO2.
For reduced SnO2 surfaces band-gap states cannot contribute directly to the
conductivity of the sample because there are no states close to the Fermi edge.
Consequently removal of lattice oxygen from the surface, for example in a reducing atmosphere, does not trigger a change in the conductivity of the sample and thus does not contribute to a gas response directly. In this context it is also important to point out that only
small band bending at the surface has been observed due to the formation of surface states
at reduced surfaces [205,210]. This indicates that charge neutrality of the surface is preserved by changing the surface from Sn(IV) to Sn(II). An indirect contribution of the
defect states created by surface reduction to the gas sensing mechanism can, however,
not be excluded. Although the Sn 5s lone pair electrons are fairly inactive to most adsorbed
molecules the changed electronic structure of the surface inuences the surface reactivity
(see Section 8). Since charge transfer to adsorbed molecules is responsible for the gas
response these changes of the surface electronic structure may play an intricate role in
the gas sensing mechanism.

108

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

6. Epitaxial SnO2 lms


There have been several eorts to grow heteroepitaxial tin oxide lms on various substrates. In all these studies stannous oxide (SnO2) was grown. Techniques employed
include pulsed laser deposition from a pressed SnO2 target [249252], sputter deposition
of a tin target in an rf-plasma containing a 1:1 oxygenargon gas mixture [253,254],
(atomic layer) chemical vapor deposition using SnI4 [255] or SnCl4 [256258] as tin precursors and H2O [259], H2O2 [260], or O2 [261] as the oxygen source, thermal evaporation of
Sn in an NO2 or ozone atmosphere [188], and oxygen plasma assisted molecular beam epitaxy (MBE) [262]. Thermal evaporation of Sn in O2 usually results in an only incompletely
oxidized lm. The crystallographic orientation of the grown SnO2 lms can be controlled
by the selection of appropriate substrates. Gas response measurements of the lm conductivity for heteroepitaxial tin oxide lms on insulating substrates have been performed on
clean surfaces [263] and surface modied by Pd adsorption [264]. Three main substrates
have been used to grow SnO2 lms, (i) isostructural growth on rutile TiO2 surfaces with
dierent orientations, (ii) growth on a-Al2O3, and (iii) growth on Pt(1 1 1). The growth
on these substrates is summarized in the following.
6.1. SnO2 on TiO2
Tin oxide was grown on the (1 1 0) [254,258,259], (1 0 0) [257,249,250] and (0 0 1) [259]
surfaces of rutile TiO2 substrates. On TiO2(1 1 0) and (1 0 0) substrates the SnO2 lms grew
with the same crystallographic orientations as the substrate for low enough deposition
rates and growth temperatures around 400500 C. Although TiO2 is isostructural to
SnO2 the lattice mismatch is substantial (aSnO2  aTiO2 =aTiO2 0:473  0:461=
0:461 2:6% and cSnO2  cTiO2 =cTiO2 0:320  0:297=0:297 7:7%), and consequently the interface between substrate and lm has to accommodate this lattice mismatch. For TiO2(1 0 0) substrates this has been investigated by high resolution TEM
[249,250]. From these investigations it was concluded that introduction of partial step dislocation at the interface with Burgers vectors of 1/2 [1 0 1] and 1/2 [0 1 1] provide the relaxation of the lattice mist. These partial dislocations form planar defects in the SnO2 lm
that correspond to crystallographic shear planes.
TiO2 and SnO2 form solid solutions over a wide range of composition [265267].
Therefore intermixing at the interface particularly at elevated growth temperature is a
possibility. This could be advantageous to reduce the lattice mismatch at the interface
but on the other hand may result in a contamination of the SnO2 lm with Ti. There
are no studies reported that address the extent of intermixing at the interface. It is the
experience of the authors of this review that the growth of SnO2 by oxygen plasma
assisted MBE on reduced TiO2 substrates at 550 C results in oxygen diusion from
the tin oxide lm into the TiO2 substrate. This is evident from the change in color of
the TiO2 substrate from dark blue to almost completely clear. Furthermore XPS spectra
collected on the grown lms show a strong metallic Sn component indicating that oxygen
from the tin oxide lm has been depleted. Oxygen diusion from a SnO2 lm into a
reduced TiO2 substrate indicates a gradient in the oxygen chemical potential that drives
the directional diusion of oxygen. This may be rationalized by the much higher heat of
formation for TiO2 (DfH(298.15 K) = 944 kJ/mol) than for SnO2 (DfH(298.15 K) =
577.6 kJ/mol).

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

109

6.2. SnO2 on Al2O3


Tin oxide lms have been grown on Al2O3(0 0 0 1) and on Al2 O3 1 0 1 2 (r-axis cut) substrates. In the rst case predominantly (1 0 0) oriented SnO2 lms were obtained [254] while
in the latter case SnO2(1 0 1) lms were grown [251253,255,256,268]. These lms were
grown by variety of techniques, including reactive sputter deposition [253], pulsed laser
deposition [251,252], pulsed-electron-beam deposition [268], atomic layer chemical vapor
deposition [255,256], and oxygen plasma assisted MBE [262]. For the SnO2(1 0 1) lms on
Al2 O3 
1 0 1 2 an epitaxial relationship of 
1 0 1 SnO2 k1 0 1 1 Al2 O3 and 0 1 0 SnO2 k1 2 
1 0 Al2 O3 has been established. Ball-and-stick models for the Al2O3 substrate and SnO2 lm
are shown in Fig. 41. From this representation the lattice match between the two crystal
structures is apparent. The lattice mismatch along the 0 1 0 SnO2 direction is only 0.4%
while it is large in the 
1 0 1 SnO2 direction where it amounts to almost 12%. The dierent

Fig. 41. Ball-and-stick models of SnO2 and Al2O3 illustrating the epitaxial relationship and the lattice mismatch.
The high resolution TEM image (viewed in the SnO2[0 1 0] direction) at the bottom of the gure shows the
formation of crystallographic shear planes to accomodate the large lattice mismatch (from Ref. [251,262], 
(2005), Elsevier).

110

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

mismatch between the two low symmetry crystallographic direction is also observed in
TEM studies of these lms. In high resolution TEM images of the lm viewed in the
1 0 1 SnO2 direction almost no interface defects are observed (not shown) while for lms
viewed in the 0 1 0 SnO2 direction a high density of interface dislocation are observed
(see Fig. 41) [251]. Similar to the case for tin oxide lms on TiO2, crystallographic shear
planes (or antiphase boundaries) are formed that accommodates the lattice mist at the
interface. These antiphase boundaries lie on the 1 0 1 planes with a displacement vector
of 1/2 [1 0 1]. The spacing between shear planes increases with lm thickness because antiphase boundaries tend to terminate inside the lms as the lm thickness increases. Ambient AFM and STM [269] studies of these lms showed that the surfaces exhibit small
atomically at terraces. Also LEED studies showed (1 0 1) spots, indicating the atomic
order of the surface, however the spots were broadened suggesting that there is some
length scale of the order of the coherence length of the low energy electrons, possibly associated with antiphase boundaries in the lm. Comparative X-ray photoelectron diraction
studies of a (1 0 1) oriented single crystal and a SnO2 lm grown by oxygen plasma assisted
MBE on r-plane Al2O3 is shown in Fig. 42. The polar scan along the 1 0 1 azimuth of
SnO2(1 0 1) shows an excellent agreement between the two samples [262].

Fig. 42. (a) Ball-and-stick model of a cross-section through a SnO2(1 0 1) surface viewed along the [0 1 0]
direction. Oxygen atoms are indicated as large balls and Sn atoms as small balls. Oxygen atoms in one (0 1 0)
plane are marked in green and Sn atoms in another plane are indicated as dark gray. Forward focusing directions
for photoelectrons in this plane are indicated by arrows. (b) Ball-and-stick model of the Al2O3 substrate viewed in
the 1 2 
1 0 direction and drawn at the same scale as the SnO2 model in (a). (c) and (d) show XPD measurements
for Sn 4d and O 1s core levels along the 1 0 1 and the 1 0 1 azimuths for a 13-nm thick SnO2 lm and a
SnO2(1 0 1) single crystal, respectively (from Ref. [262],  (2005), Elsevier).

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

111

6.3. SnO2 on Pt(1 1 1)


Oxidation of bulk PtSn and SnPt(1 1 1) surface alloys results in the formation of complex SnOx surface layers that depending on the preparation conditions result in several different ordered oxide overlayers [270274]. These monolayer thick lms are proposed to
have dierent stoichiometries but generally x is less than 2. Also Sn alloyed with the Pt
substrate is always present and may play an important role in the formation of some of
the monolayer oxides. Deposition of multilayer-Sn (34 monolayers) in an oxidizing atmosphere on Pt(1 1 1) substrates at 600 K and subsequent annealing to 900 K results in the
formation of SnO2 crystallites [188]. STM investigations shown in Fig. 43(a) illustrate that

Fig. 43. SnO2 growth on Pt(1 1 1). (a) Large scale STM image of SnO2(1 0 1) crystallites with preferential
orientation along the three low symmetry directions of Pt(1 1 1). (b) STM image and LEED pattern of the 4 4
wetting layer. The 4 4 unit cell and a 1 1 periodicity net are superimposed on the STM image. The 1 1 unit
cell is indicated in the LEED pattern. (c) Close-up STM image of a SnO2(1 0 1) crystallite and atomic resolution
image. The epitaxial relationship of the (1 0 1) unit cell with respect to the Pt(1 1 1) lattice is also indicated (from
Ref. [188],  (2004) by the American Physical Society).

112

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

these crystallites are aligned along the three {1 1 0} directions of the Pt(1 1 1) substrate. The
Pt substrate between the crystallites is covered with a SnOx wetting layer that exhibits a
4 4 periodicity with respect to the Pt(1 1 1) lattice. STM and LEED of this wetting layer
is shown in Fig. 43(b). Thus the growth of tin oxide on Pt(1 1 1) proceeds via a pseudoStranskiKrastanov growth mode, i.e. formation of a wetting layer followed by the growth
of three dimensional crystallites. STM images of these crystallites (Fig. 43(c)) exhibit the
structure of the SnO2(1 0 1) surface (see Fig. 24 for STM on single crystal SnO2(1 0 1)). The
growth in the (1 0 1) orientation is stabilized by the low mismatch of this orientation with
the Pt(1 1 1) substrate with a mismatch of 0.5% and 3% in the 0 1 0 and 1 0 1 SnO2 directions, respectively. This is indicated in Fig. 43(c3). These small SnO2 crystallites exhibit
bulk-like properties. Interband transitions measured by EELS indicate a bulk-like bandgap and vibrational spectroscopy show the characteristic FuchsKliewer phonon modes
for SnO2 (see also Section 3.2). These are identied in HREELS spectra shown in
Fig. 44. Multiphonon losses on these small SnO2 crystallites are strongly attenuated compared to bulk SnO2 samples, but are still observed. The 4 4 wetting layer also displays
distinctive vibrational modes that give rise to energy losses at 540 cm1 and 630 cm1.
6.4. Conclusions and outlook for SnO2 epitaxy
Epitaxial thin lms on insulating substrates may be used as model systems for gas sensors [263,264]. However, for most of these lms no atomic scale characterization of the
surface structure exists. These lms may also be ideal for studying eld eects on chemical
reactions and gas sensing properties. Applying an electric eld via a gate electrode modies
the charge carrier concentration in the lm via the eld eect [275277]. It has been
reported that this inuences the gas response, however, no studies on single crystalline
lms have been reported.
In addition to the above reported substrates other substrates would be desirable. The
SnO2(1 0 1) lms grown on Al2O3 appear to be of fairly good quality, for other lm orientations the TiO2 substrates seem to be useful. However, there are questions about interdif-

Fig. 44. HREELS spectra for SnOx-4 4 wetting layer on Pt(1 1 1) and SnO2(1 0 1) crystallites on Pt(1 1 1) (from
Ref. [188],  (2004) by the American Physical Society).

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

113

fusion of Ti and Sn at the interface. Substrates for the growth of SnO2 with other crystal
orientations with high quality are also still sought.
It also would be desirable to have a metal substrate that allows the growth of single
domain SnO2 lms and thus the characterization of highly stoichiometric and insulating
lms by electron spectroscopy and STM. None such substrate has been reported yet.
7. Additives in SnO2
As pointed out in Section 2, in most applications of SnO2 the chemical and physical
properties are modied by metal or metal-oxide additives. A distinction between additives
that are present at the surface of SnO2 in form of clusters or ad-layers and additives that
are atomically dispersed within a SnO2 host matrix can be made. Unfortunately, in many
cases the distinction between bulk and surface dopants is not that clear cut. For example
for Pd, a common additive in gas sensing applications, it is frequently assumed that it is
present at the surface where it sensitizes the gas response by either catalytic reactions or
by altering the interface electronic structure (see Section 2.3.2). There are, however, studies
that indicate that some Pd may diuse into the SnO2 matrix, in particular under oxidizing
conditions [142147]. Thus Pd may also act as a bulk dopant and for dierent gas environments the solubility of Pd in SnO2 may change which could contribute to the gas response.
On the other hand, bulk dopants may agglomerate and form inclusions within the SnO2
matrix or segregate to the surface. This may lead to a surface layer with dierent dopant
concentration than in the bulk. Nevertheless a separation in surface additives and bulk
dopants is sensible in most cases and we use this distinction for the next two sections.
7.1. Clusters and ad-layers
There exists only a few fundamental surface science studies of metal growth on SnO2.
Most studies concentrated on Pd-growth [278285], but there are also reports of the
growth of Sn [279,280,286] and Cu [287] on SnO2(1 1 0), and DFT calculations for Pt interaction with SnO2(1 1 0) [288]. Also the growth of vanadium [289] and vanadia [290] on
SnO2(1 1 0)-4 1 was studied. These studies are summarized in the following.
7.1.1. Tin
The growth of Sn on four dierently prepared SnO2(1 1 0) surfaces was investigated by
XPS, UPS, and four-point sheet conductance measurements. These four SnO2 surfaces
were 4 1 reconstructed surfaces, sputtered surfaces, as well as thermally [280] and plasma
oxidized [279] surfaces that both exhibited a 1 1 LEED structure. Studies on the same
surfaces were also performed for Pd growth and are summarized in the next section. Sn
deposition and characterization were carried out at 300 K. Fig. 45(a) summarizes the surface properties found for dierent Sn coverage on the rst three SnO2(1 1 0) surfaces mentioned above. A 50-fold increase in conductivity has been observed after only depositing
0.1 ML of Sn on the thermally oxidized surface. On the plasma oxidized surface this
increase is reported to be 500-fold [279]. For low Sn coverage no metallic Sn is detected.
It takes 0.2 ML of Sn on the 4 1, 0.4 ML of Sn on the sputtered surface, and 0.5 ML of
Sn on the thermally oxidized surface before metallic Sn is observed in XPS. This indicates
that deposited Sn reacts with surface oxygen. Both chemisorbed oxygen and lattice oxygen
is proposed to take part in the oxidation of deposited Sn. The change in conductivity for

114

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 45. The eect of Sn deposition on various parameters for three dierently prepared SnO2(1 1 0) surfaces is
shown in (a). Sheet conductivity measurements, band bending, changes in the work function, the percentage of
metallic Sn detected in XPS and the attenuation of the O peak area are shown as a function of Sn coverage. The
Sn 3d core level for Sn deposition on a plasma oxidized sample is shown in (b). An initial shift to lower binding
energy due to oxygen induced band bending is relaxed after deposition of 0.1 ML Sn. At 1.0 ML Sn coverage
metallic Sn is detected (from Ref. [279],  (1991) Elsevier and Ref. [280]).

thermally and plasma oxidized samples has been attributed to the depletion of charged
chemisorbed oxygen species and additional creation of defects by abstraction of lattice
oxygen by deposited tin atoms. In contrast to vacuum reduced surfaces, no defect states
were, however, observed by UPS within the band gap after deposition of Sn on the oxidized samples [279]. This implies either a very small density of defects or a dierent kind
of defects formed by Sn deposition compared to thermal reduction under vacuum. The
reaction of deposited Sn with charged, chemisorbed oxygen is supported by the observation of band-bending eects on the plasma oxidized surface. This is shown in Fig. 45(b).
The clean oxidized surface exhibits a binding energy shift of 0.7 eV to lower binding
energy, i.e. an upward band bending. This shift relaxes upon Sn deposition. In
Fig. 45(b) it can also be seen that metallic Sn appears at about 1 ML Sn coverage.
7.1.2. Palladium
Equivalent studies of those of Sn were also performed for Pd deposited on SnO2(1 1 0)
[279,280]. An overview of the results for dierent Pd coverage are shown in Fig. 46(a) for
sputtered, 4 1 reconstructed and thermally oxidized surfaces. Low coverage of Pd produced a small increase in conductivity. Although a shift in the Pd 3d core level is observed
to higher binding energy on all surfaces that may indicate formation of oxidized Pd, or
may be attributed to relaxation shifts for small Pd clusters. Comparison of the Pd 3d peak

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

115

Fig. 46. The eect of Pd deposition on various parameters for three dierently prepared SnO2(1 1 0) surfaces is
shown in (a). Sheet conductivity, the photoemission intensity at the Fermi edge, band bending, change in work
function, the peak binding energy of Pd 3d5/2, the metallic Sn component and the Sn 3d peak area are shown as a
function of Pd coverage. In (b) the Sn 3d5/2 core level is shown after deposition of 2.3 ML Pd on a plasma oxidized
sample and a thermally oxidized sample. More metallic Sn is detected on the thermally oxidized sample. In (c) the
Pd 3d core level for 0.1 ML Pd on the plasma and thermally oxidized surface is shown. A larger shift to higher
binding energy is observed for Pd on the plasma oxidized sample (from Ref. [279],  (1991), Elsevier and Ref.
[280]).

position for 0.1 ML Pd on thermally and plasma oxidized surfaces shown in Fig. 46(c)
indicate a stronger shift on plasma oxidized surface. Thus it may be concluded that Pd
oxide forms on the plasma oxidized surface for low coverages due to the reaction of Pd
with chemisorbed oxygen species. Time dependent conductivity measurements after the
deposition of Pd on thermally and plasma oxidized surfaces also showed dierent behavior
that could be explained by the amount of charged chemisorbed oxygen available at the
two surfaces [281].

116

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

At high coverage (>5 ML) a sharp increase in conductivity is observed that is attributed
to the formation of a contiguous Pd ad-layer. No oxidized Pd is observed at high coverage
but metallic Sn is formed. This indicates the formation of a PdSn alloy at high Pd coverage. For a diluted SnPd alloy ([Pd]/[Sn] > 6) the PdSn bonds are quite strong (over
60 kcal/Sn atom) and thus may favor the alloy formation. Fig. 46(b) shows the metallic
Sn component detected in XPS after deposition of 2.3 ML Pd on thermally oxidized
and plasma oxidized samples. It is evident that more metallic Sn is detected for thermally
oxidized samples. Formation of long range order in the Pd lm is apparent from the observation of a Pd(1 1 1) LEED pattern that is in registry with the SnO2(1 1 0) substrate for a
10 ML lm [280]. XPS and LEIS [282] indicate that Pd growth proceeds via cluster growth
with complete coverage of the substrate at around 10 ML.
STM on SnO2(1 0 1) oriented lms also suggest a cluster growth mode [278]. Below two
monolayers of Pd, Pd formed at, 0.5 nm high islands, 1020 nm in diameter, indicating
wetting of the substrate at low Pd coverage. Annealing of a 7 ML Pd lm to 670 K for
30 min resulted in the appearance of metallic Sn0 state in XPS, consistent with formation
of a PdSn alloy.
High resolution TEM studies on the epitaxy of Pd clusters on the {1 1 0} facets of SnO2
nanosticks are reported [283]. Semispherical Pd clusters were observed and an epitaxial
relationship of 1 
1 0 1 1 1Pd k0 0 1 1 1 0SnO2 was established.
The surface adsorption energies of Pd atoms on dierent (defect) sites on the SnO2(1 1 0)
surface was calculated by density functional theory [285]. In these calculations it was
found that Pd clusters on stoichiometric surfaces. On reduced surfaces Pd may substitute
surface Sn atoms. Pd adsorption was found to be strongest at in-plane oxygen vacancy
sites.
The inuence of 3 ML Pd on SnO2(1 1 0) on the gas sensitivity towards hydrogen was
studied by sheet conductivity measurements. It was found that the increase in conductivity
was 16 times greater with Pd than without [282]. This increase was attributed to spillover
of atomic hydrogen from the Pd clusters onto SnO2. Pd cluster formation was observed for
SnO2 nanosticks impregnated with Pd upon reduction by heating in an Ar atmosphere
with 5% H2 [284]. Before reduction the Pd was either within the bulk or ultradispersed
on the surface.
7.1.3. Copper
There exists one study of Cu growth on SnO2(1 1 0) [287]. Similar to the case of Pd,
island growth has been observed with the formation of Cu clusters that exhibit a
Cu(1 1 1) orientation in registry with the SnO2(1 1 0) substrate. Also metallic Sn was
observed for thick Cu lms on top of the Cu clusters. Annealing in 5 106 mbar O2
oxidized both the metallic Sn and the Cu clusters.
7.1.4. Vanadium and vanadia
Vandium deposition on vacuum prepared SnO2(1 1 0) surfaces at room temperature
results in oxidation of vanadium and formation of metallic tin [289]. Increasing vanadium
deposition results in growth of metallic vanadium islands. Annealing of the vanadium lm
to 800 K causes the oxidation of the entire vanadium lm and also re-oxidation of the
metallic Sn. XPD indicate that the formed vanadium oxide lm is rutile VO2 in registry
with the SnO2 substrate. No mixing of the two oxides has been observed. The VO2 lms
appears to be covered by a disordered tin oxide layer, though.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

117

Vanadium oxide was also deliberately grown by vapor depositing vanadium in an oxygen atmosphere on reduced SnO2(1 1 0)-4 1 surfaces. The grown vanadia lms were characterized by XPD and LEIS [290]. The lm structure appears to be quite similar to the one
obtained by vanadium deposition in vacuum and subsequent annealing. It was established
that vanadium oxide grows epitaxially with a similar rutile structure as the SnO2 substrate
and a VnO2n1 stoichiometry with 4 < n < 8. Formation of so-called Magneli phases was
proposed, where the deviation from the rutile VO2 stoichiometry is due to the formation of
non-stoichiometric crystallographic shear planes. Furthermore, it was concluded that the
vanadium oxide lm does not mix with the tin oxide substrate up to 800 K but appears to
be capped with a two-dimensional tin containing surface layer.
7.1.5. Conclusions and outlook for surface dopants
For metal deposition on SnO2, dierences have been observed depending on the oxidation procedure of the substrate. It appears that plasma oxidized samples have a higher
concentration of chemisorbed oxygen at the surface than thermally oxidized surface under
high oxygen pressure. These chemisorbed oxygen species are rather reactive to deposited
metal and cause initial oxidation of both Sn and Pd deposits. Reactive metals like vanadium oxidize at the interface to SnO2 even if the SnO2 substrate is reduced by vacuum
preparation procedures. It is interesting to note that Sn deposition on an oxidized surface
resulted in a much larger conductivity increase in the surface layer than was observed for
Pd deposition in the submonolayer regime. This was attributed to lattice oxygen abstraction and formation of defects in addition to oxidation of chemisorbed oxygen. The lack of
defect states in UPS spectra however indicates that either a dierent kind of defects form,
distinct form the defects observed on vacuum reduced samples or only a small amount of
defects are formed. In either case it is not obvious of how such defects would contribute to
the drastic increase in conductivity. It is also possible that samples modied with Pd show
a lesser increase in conductivity than Sn dosed samples, because the conductivity increase
due to consumption of oxygen species is opposed by the formation of a Schottky barrier at
a SnO2/PdO interface. In future studies it may be interesting to see how reactive elements
like vanadium, which have demonstrated to easily abstract lattice oxygen from the SnO2
surface, impact the sample conductivity.
Another interesting observation is the apparently easy formation of Sn alloys upon Pd or
Cu deposition. This is dierent to studies on the related Cu/ZnO system where there have
been no reports that unambiguously identify the formation of a CuZn alloy upon Cu deposition on ZnO under UHV conditions [291]. Alloy formation suggests a gradient in the
chemical potential of Sn that drives Sn diusion from the SnO2 substrate into the metal
clusters. Many metals form strong intermetallic compounds with Sn, in addition to Cu
and Pd other frequent additives to SnO2 gas sensors such as Ag, Au and Pt show strong
exothermic mixing with Sn. It has been shown for SnPt and SnPd alloys that Sn is extracted
out of these alloys in an oxidizing atmosphere and form SnOx overlayers on Pt (see also
Section 6.3) and Pd [193]. This falls into the eld of strong-metal-support interaction
(SMSI) discussed extensively in heterogeneous catalysis. These ultrathin tin oxide lms
on metal supports are known to reduce easier than bulk tin oxide [274]. Thus depending
on the oxidation potential of the atmosphere the metal clusters may be encapsulated in a
tin oxide lm or the Sn is alloyed with the metal cluster. These variations will inuence
the catalytic activity and the electronic response of a SnO2 gas sensor. No detailed studies
of the SMSI eect for supported metal clusters on SnO2 have been reported yet.

118

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Another question that needs to be answered is the extend and inuence of mixing of
metal additives with SnO2. It has been shown on powder samples that some Pd mixes with
SnO2 under oxidizing conditions. If this is part of the gas sensing mechanism then this has
to result in a decrease of charge carriers in the sample due to p-type doping eect of Pd.
First the stability of Pd in a SnO2 matrix has to be assessed. Electrical transport measurements of Pd doped epitaxial lms, for instance, could give information of the inuence of
Pd dopants on the charge carrier concentration.
7.2. Lattice dopants
A great variety of dopants have been studied in order to improve the properties of SnO2
for certain applications. Most of these studies were performed on complex samples, polycrystalline at best. These studies include Mo [292294], V [295297], Al [298,299], Ta [300
302], Nb [299,303306], In [306,307], Ge [308], Ru [309], Cr [310312], Bi [305], Ga [313],
Li [314], F [294,315], and SnO2 doped with rare earth elements Ce, La, and Y [316]. The
most thoroughly studied system is Sb doped SnO2 and these studies are summarized in the
following subsection. Fe [317] and Co doped [318] SnO2 has attracted some recent interest
because of its ferromagnetic properties that may make it a useful material for spintronic
applications. Section 7.2.2 briey discusses Co doped lms.
7.2.1. Antimony-doped SnO2
As discussed in Section 2.1 Sb is a common n-type dopant in SnO2. Stjerna et al. [294]
reported a strong increase in the free electron concentration in SnO2 when doped with Sb.
For 0%, 2%, 4%, and 8% Sb doped SnO2 they measured charge carrier concentrations of
0.015 1021 cm3, 0.35 1021 cm3, 0.54 1021 cm3, and 0.80 1021 cm3, respectively.
They also noted that the optical band gap increased by 0.35 eV for heavily Sb doped samples compared to pure SnO2. This is in agreement with the BursteinMoss eect described
in Section 2.1. The transmission of highly doped SnO2 lms was observed to decrease[319]
and a slight blue discoloration is often reported.
Sb dopants occupy Sn sites in the rutile structure. This has been established by EXAFS
studies [313,320]. XANES measurements [320] showed that the largest fraction of Sb has a
valency of 5+, although 3+ was also observed in nanocrystalline SnO2. The Sb3+ species
may be due to a strong tendency of Sb to segregate to the surface [200,321]. These surfacesegregated Sb atoms may form oxygen vacancy complexes at the surface and may be
important active sites for oxygen and water adsorption [321]. Surface segregation of antimony has also been concluded from photoemission measurements [306,322325].
Although a surface enrichment of antimony is rmly established no increase in free charge
carriers near to the surface is observed. In contrary, some studies observe depletion of
charge carriers near the surface [324]. The lower charge carrier concentration in the surface
layer is explained by the trapping of electrons in Sb(III) like lone pairs [326]. It is also
interesting to point out that UPS measurements showed that the surface states due to surface oxygen deciency and Sn2+ species observed on pure SnO2 samples (see Section 5.2)
are far less pronounced at Sb doped samples. This is in agreement with Sb3+ replacing
Sn2+ surface species and a lower binding energy of the Sb3+ 5s5p lone pair electrons compared to the Sn2+ 5s5p electrons [325]. Thus the lone pair electron levels are pushed down
into the bulk valence band region and are not observed in the band-gap region. No studies
of Sb doped single crystals have been reported.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

119

UPS measurements identied electronic states at the Fermi level conrming the lling
of the conduction band by Sb doping. Furthermore, a downward shift of the valence band

Fig. 47. The left panel shows ts of the Sn 3d5/2 peaks in Sb doped SnO2 to two Voigt components associated
with the screened and unscreened nal states for the dierent doping levels. The right panel shows (a) the
variation in the full width half maximum height of the Sn 3d5/2 peak as a function of the charge carrier
concentration. (b) The screened to unscreened ratio as a function of charge carrier concentration. (c) Separation
between unscreened and screened components as a function of charge carrier concentration. (d) Variation in the
baricenter of the Sn 3d5/2 binding energy (solid line) and of the binding energy of the rst peak of the valence band
spectrum (solid circles) as a function of charge carrier concentration (from Ref. [327],  (2003), Elsevier).

120

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

was observed [239] as is expected due to a shift of the Fermi level into the conduction
band. However, after evaluating the charge carrier concentration from measuring the plasmon losses in EELS [322] and taking the conduction band dispersion into account the
measured downward shift of the valence band was concluded to be smaller than the occupied part of the conduction band. Consequently, shrinkage of the bulk band gap due to
attractive dopant electron interactions and screening of the Coulomb repulsion between
conduction and valence electrons was proposed.
Sn core level photoemission peaks exhibit a satellite component for Sb doped SnO2
[322,324,327]. Fig. 47 shows the Sn 3d5/2 peak for photoemission from SnO2 with dierent
Sb concentration. The observed shift of the baricenter in the binding energy of the core
level with Sb doping is due to a shift of the Fermi level due to occupation of the conduction band and the above described band-gap shrinkage. The asymmetric shape of the Sn 3d
peak may be attributed due to two peaks associated with screened and unscreened nal
states. The separation between the two components and the relative intensity of the
screened component increase with increasing Sb-induced charge carrier concentration.
Thus, as expected, the screening response of a dilute electron gas due to the creation of
a corehole decreases as the carrier concentration decreases. It has been shown that the
separation between the screened and unscreened nal states is of the same order of magnitude as the plasmon energy [327]. The strong plasmon satellite in the Sn core level peaks
observed for Sb doped lms can be expected for other samples with strong n-type doping
such as F or Cl doped SnO2 as well. The splitting of the satellite peak may thus be used as
a measure for the charge carrier concentration in SnO2.
7.2.2. Cobalt-doped SnO2
Co doped SnO2 was grown by pulsed laser deposition [318] as well as by oxygen plasma
assisted MBE [262] on r-plane alumina (see also Section 6.2). It has been demonstrated
that Co doping may result in a ferromagnetic lm with a giant magnetic moment of
7.5 0.5lB/Co and a Curie temperature of 650 K for oxygen decient lms with 5% Co
doping. A possible explanation of the high magnetic moments observed in dilute ferromagnetic oxides is given in Ref. [328]. All the lms were rutile structure with a (1 0 1) orientation. X-ray absorption spectroscopy at the Co L2,3 edge suggest that Co atoms occupy
oxygen octahedral coordinated cation sites [329]. In X-ray photoelectron diraction studies (XPD) on MBE grown lms a similar variation of the Co 2p photoemission intensity
and the Sn 3d intensities with polar emission angle has been observed [262]. This strongly
suggests that a large fraction of the Co dopants substitutionally occupy Sn sites in the
SnO2 lattice. XPS indicates that these Co dopants have a 2+ charge state. The exact origin
of the ferromagnetism in dilute oxides is not completely understood yet, it seems likely
though that it is charge carrier mediated and thus the oxygen vacancy concentration as
intrinsic dopants may play an important role for the ferromagnetic properties of Co doped
SnO2.
8. Surface chemistry of SnO2
Despite the importance of molecular interactions with SnO2 in its applications as gas
sensor and heterogenous catalyst, relatively few surface science studies of molecular
adsorption and reactions on SnO2 single crystals have been performed. Furthermore, most
of these studies have been concentrated on the SnO2(1 1 0) surface. Much of this work has

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

121

been performed by Cox and co-workers. In their studies they assumed simple bulk terminations with dierent oxygen concentration. It has been discussed in Section 4, however,
that the surface structure of SnO2(1 1 0) can be quite complex. In particular recent DFT
calculations question if a surface with only the bridging oxygen atoms removed can be stable. Nevertheless, many of the conclusions drawn by Cox and co-workers regarding surface-composition dependent reactivity are still valid regardless of the precise surface
structure. In the next subsection we give some general remarks about molecular reactions
on SnO2 surfaces before we summarize adsorption studies of specic molecules.
8.1. General remarks on molecular reactions at SnO2 surfaces
Many reactions of organic molecules at oxide surfaces are successfully described in the
framework of surface acidbase reactions [330] where the oxygen anions can act as
Brnsted or Lewis base sites and the metal cations are described as Lewis acid sites. Many
Brnsted acids dissociate on oxide surfaces, causing protonation of surface oxygen anions,
and adsorption of the conjugated base of the acid to metal cations at the surface. The key
requirement for the dissociative adsorption thus is the availability of pairs of undercoordinated metal cations and surface base sites (oxygen). Thus the coordination environment
of the surface cations and anions are the critical surface characteristics for describing surface reactions of Brnsted acids.
From the previous discussion of the surface composition and structure it is obvious that
the coordinative environment and the valency of Sn are critically dependent on the surface
preparation of SnO2 surfaces. This opens opportunities to study the structurereactivity
relationship of one oxide surface by varying the preparation condition. This was recognized by Cox et al. who proposed that one could prepare three dierent kind of (1 1 0) surfaces by high pressure oxygen annealing or plasma treatment. These three surfaces are: (i)
stoichiometric (1 1) surfaces, (ii) (1 1) surfaces with bridging oxygen rows removed,
and (iii) surfaces that had all bridging oxygen atoms removed and had additional in-plane
oxygen vacancies. Similarly, for the (1 0 1) and (1 0 0) surfaces, stoichiometric and reduced
SnO2 surfaces with Sn4+ and Sn2+ surface cations can be obtained (see Section 4.2). Thus
by preparing these surfaces adsorption sites with dierent electronic and steric properties
could be designed and site specic reactions may be probed. Consequently one has the
unique possibility on SnO2 surfaces to reduce the number of oxygen atoms, i.e. Brnsted
or Lewis base sites, by surface reduction and this ought to have signicant impact on the
surface reactivity. Furthermore, a transformation of surface Sn species from a valency of
IV to a valency of II and the formation of fairly inactive Sn 5s lone pair electrons is
expected to aect molecular adsorption. Thus diverse surface reactivities may be probed
for the SnO2 surfaces depending on the coordination environment and charge state of
the Sn cation. For instance, dissociation of Brnsted acids like methanol, formic acid,
and water, is very dierent for oxygen rich and reduced surfaces. This is described in some
more detail below. Briey, it was observed that reduction of the (1 1 0) surface initially
increases the activity for dissociative adsorption of methanol but inhibits dissociative
adsorption for a more reduced surface. Cox and coworkers argue that the reason for
the decrease in activity with reduction of the SnO2 surface (in contrast to the generally
observed increase on reduced oxides) may be twofold. One reason may be the lack of
Brnsted base sites in form of oxygen atoms on the reduced surface, the other reason
may have to do with the dual valency of Sn that inuences the ideal coordination of

122

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

the Sn cations. From the experimental observations on the (1 1 0) surface it was concluded
that fourfold coordinated Sn2+ atoms have the highest activity for methanol dissociation
but the formation of in-plane oxygen vacancies, i.e. a further reduction of the surface, inhibits methanol dissociation due to the formation of less than four coordinated Sn2+ atoms.
One fact that may be very important, but is often neglected, for understanding SnO2 in
its applications in gas sensors and catalysts is the role of atomic scale surface defects. It is
important to point out that a single missing oxygen atom, for instance in the bridging oxygen row of the (1 1 0) surface, creates a site that is distinct from a surfaces with all the
bridging oxygen atoms removed. This is because in the latter case the Sn atoms may adopt
a Sn2+ valency and as such there are no broken bonds while for a single oxygen vacancy
there is one broken bond for each of the two neighboring Sn surface atoms. Thus one may
expect that such Sn cation sites may act as particularly strong Lewis acid sites and consequently to be particular reactive. Furthermore, molecules adsorbing at an oxygen vacancy
can interact with neighboring bridging oxygen sites. From STM studies on clean surfaces
it is apparent that defect sites are common features. There are, however, no experimental
surface science studies on single crystals that describe the impact of specic defect constellations on the surface reactivity. This is because of the diculty to prepare surfaces with a
single defect geometry. For other oxide surfaces it is, however, well established that oxygen vacancies are particularly active sites. For instance the strong reactivity of oxygen
vacancy sites on TiO2(1 1 0) surfaces is rmly established [34]. There are a few theoretical
studies that also suggest that vacancies are critical to understand adsorption of oxygen and
the eect oxygen has on the gas response of SnO2 (see below). Also, the observation by
Cox et al. that dissociation of Brnsted acids initially increases with reduction of the
SnO2(1 1 0) surface may be explained by the role oxygen vacancies play.
8.2. Adsorption and reactions of specic molecules
8.2.1. Oxygen
In Section 2.3.1 we discussed the importance of charged, chemisorbed oxygen species
for explaining the gas sensing mechanism of SnO2. From ESR measurements [331334]
it has been proposed that adsorbed oxygen changes to various oxygen anion species transferring an electron form SnO2 to the chemisorbed oxygen according the following processes [335]:

2
2
O2 gas () O2 ad () O
2 ad () O ad () O ad () O lattice.

Chang [336] compared EPR signals measured on powder samples with conductivity measurements on SnO2 thin lms in an oxygen atmosphere. A transition temperature of
150 C was found. Below this temperature oxygen adsorption at the surface was mainly

2
in form of O
was found.
2 while above 150 C chemisorbed oxygen in form of O or O
This transition temperature correlates to an anomalous temperature dependence found for
the thin lm conductance. Below 160 C the sample conductance increased with temperature, which is expected for a thermal excitation of charge carriers in a semiconductor.
Above 160 C the inverse was observed, i.e. the sample conductance decreased with
increasing temperature. This was interpreted in terms of the conversion of chemisorbed

2
O
2 into O (O ) and thus additional acceptance of electrons from the tin oxide by the
oxygen adsorbates.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

123

Thermal desorption studies of oxygen from SnO2 powder in a ow of helium or nitrogen showed oxygen desorption peaks at 100 C and 500 C and a constant increase in
oxygen desorption above 600 C [337]. The peaks were sensitive to the preparation conditions. The low temperature peak was only observed for oxygen adsorption at or below
150 C and cooling to RT in the oxygen atmosphere. The high temperature peak was present if oxygen was adsorbed at 400 C. From ESR spectra it was concluded that the low
temperature peak was due to O
2 ions and the high temperature peaks were assigned to
O and/or O2 species. The increase of oxygen above 600 C was believed to be due to
decomposition of SnO2 and evolution of lattice oxygen. The only UHV single crystal
TPD study was performed on 4 1 reconstructed and the pseudo-1 1 high temperature phases of SnO2(1 1 0) [338]. In these studies oxygen was adsorbed at 100 K and
desorption peaks were observed at 200 K with a shoulder at 250 K.
XPS studies of nanocrystalline SnO2 lms showed an upward band bending of 0.2 eV
after exposure to 104 L O2 at 120 C and 250 C [339]. Although no chemisorbed oxygen
could be detected by XPS this was believed to be due to adsorption of small amounts of
oxygen. Subsequent exposure to CH4 inverted the band bending which indicated a reaction of the chemisorbed oxygen with CH4. Band bending induced by chemisorbed oxygen
was also observed by Mizsei and Lantto [340], and Semancik and Cox [341]. A stronger
band bending of up to 0.7 eV was observed on single crystal SnO2 samples that were
exposed to an oxygen plasma rather than molecular oxygen (see Fig. 45(b)).
XPS studies of oxygen adsorption on Ar-ion etched SnO2 lms showed additional components for the O 1s core level after oxygen adsorption [342,343]. This was interpreted as

O, O2
2 , and O2 species adsorbed at oxygen vacancies. Three surface oxygen species were
also observed in the valence band by UPS measurements [342]. However, the Ar etching
procedure is likely to produce more complicated defect structures at the surface and thus
these assignments should be taken with caution.
At temperatures above 1000 K, Schierbaum et al. [344] observed reversible changes in
the conductivity of a thin SnO2 lm with oxygen partial pressure. The conductivity
1=6
changes followed a power law r  pO2
which the authors ascribed to an equilibrium
condition of doubly positively charged oxygen vacancies with the gas phase oxygen.
Although there is evidence for complex oxygen/SnO2 interactions, which are instrumental for triggering gas responses, there is little known about specic adsorption geometries.
To better understand the interaction of O2 with SnO2, calculations were conducted. The
adsorption of O2 on stoichiometric (1 1 0) [345348] and (1 0 1) [349] surfaces as well as
at oxygen vacancies has been investigated theoretically. Neutral O2 adsorbs extremely
weakly on stoichiometric SnO2 surfaces, with calculated adsorption energies less than
0.02 eV for the (1 1 0) surface [346]. At bridging-oxygen vacancies adsorption is however
strong. Dissociation of O2 adsorbed at vacancy sites with one O atom lling the vacancy
and the other adsorbing at neighboring vefold coordinated Sn sites has been shown to be
exothermic for the (1 1 0) surface [345,350]. A thorough investigation of oxygen adsorption
on stoichiometric and various reduced surfaces by DFT calculations was performed by
Oviedo and Gillan [348]. They conrm that O2 cannot adsorb exothermally on the stoichiometric (1 1 0) surface. Weak adsorption is possible at the vefold coordinated Sn sites
with an adsorption energy less than 0.4 eV. Much stronger adsorption is possible at bridging-oxygen vacancy sites with an adsorption energy of up to 1.8 eV. There was also evidence that a bridging oxygen vacancy can bind three O2 molecules simultaneously with
one at the vacancy site and two at the neighboring vefold Sn sites. Using a point-charge

124

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 48. Geometries of adsorbed oxygen on SnO2(1 1 0) surfaces. Favorable adsorption geometry for atomic
oxygen are shown for (a) a surface with excess charges and (b) a neutral surface. Adsorption geometries of O
2
and O
3 if there is an additional electron in the unit cell are shown in (c) and (d) respectively (from Ref. [350], 
(2005), Elsevier).

model Yamaguchi et al. considered the stability of dierent oxygen anion species stabilized
at bridging-oxygen vacancy sites at the (1 1 0) surface [347]. Considering the energy diagram for dierent oxygen species they concluded that a low concentration of O
2 and
O can be expected. Maki-Jaskari et al. studied the charge accumulation of oxygen species
adsorbed on an ideal and defective SnO2(1 1 0) surface with excess charge by DFT calculations. An excess charge of 10211022 cm3 inuenced the oxygen surface chemistry of
SnO2 and allowed to model the ionosorption of oxygen. Fig. 48 shows the preferred
adsorption geometry of atomic oxygen on the ideal (1 1 0) surface with (Fig. 48(a)) and
without (Fig. 48(b)) excess charge. With additional charges an O ion adsorbs at vefold
Sn sites, while without additional charges a O2
2 complex is favorable. This latter structure
is identical to the adsorption of O2 at a bridging-oxygen vacancy site and can be viewed as
the initial step of oxygen adsorption. An energy of 2 eV=O
ads is being released upon transformation of the initial adsorption geometry (Fig. 48(a) to the nal geometry (Fig. 48(b))

in the presence of excess charge. Two O
ads can combine to form O2 (see Fig. 48(c). Also,


Oads can react with O2 and form a O3 complex (see Fig. 48(d)). The energy release for the
latter reaction was evaluated to be 0:5 eV=O
ads , while the energy release by combining two


O
ads to O2 has been calculated to be 0:9 eV=Oads .
For the SnO2(1 0 1) surface, DFT calculations showed that O2 adsorbes at oxygen
vacancy sites with a signicant charge transfer forming an O
2 species [349].
8.2.2. Carbon monoxide CO
No experimental studies of CO adsorption on single crystal surfaces have been
reported. There are, however, reports on the interaction of CO on SnO2 powder and
porous thin lms. Transmission infrared spectroscopy was used to characterize CO
adsorption on a pressed SnO2 powder disk at 400500 K [351]. Two broad absorption
bands were identied at 1440 cm1 and 1370 cm1. These were associated with an

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

125

unidentate carbonate species. It was proposed that CO reacts with a surface oxide forming a CO2 molecule that is adsorbed onto a Sn2+ species. Mossbauer spectroscopy of
samples that had been exposed to CO also showed a greater Sn(II) resonance indicating
a consumption of lattice oxygen and thus reduction of the sample by CO. No evidence
of adsorption of carbon monoxide was found in this study. More recent FTIR studies
on powder samples at 120 K at various CO pressures drew a dierent picture of CO
adsorption [352]. CO adsorption on samples prepared at 773 K in vacuum, i.e. on possibly reduced SnO2 surfaces, and samples that were oxidized and cooled to RT in O2
were examined. Both samples showed very similar CO adsorption. At low CO pressures
(up to 0.04 Torr) an absorption band at 22102196 cm1 develops, which was assigned
to CO bound to a coordinational undersaturated (cus) Sn4+ surface species. At higher
CO pressures (0.040.52 Torr) a dierent kind of Sn4+ cus sites is being occupied giving
rising to second absorption band at 21802177 cm1. Physisorbed CO developed at CO
pressures higher than 0.52 Torr. In both infrared spectroscopy studies mentioned above
hydroxyl groups were present at the surface that may have strongly inuenced the CO
adsorption.
TPD studies from oxidized SnO2 powder under He ow showed two desorption peaks
after exposing the sample to 1 Torr CO [353]. CO2 was observed to desorb at 220 C and
oxygen desorbed at 550 C. The amount of CO2 desorption decreased with increasing
reduction of the sample. It was concluded that CO was captured by surface oxygen to
form a surface carbonate species that subsequently desorbs as CO2 at elevated temperatures in agreement with the infrared studies of Ref. [351]. Coadsorption experiments with
NO, N2O and H2O showed that these strongly adsorbed molecules block adsorption sites
for CO.
The interaction of CO with the vefold Sn site at the stoichiometric SnO2(1 1 0) surface
was investigated by Melle-Franco and Pacchioni by quantum-mechanical calculations
[354]. Only CO adsorption with the C-end oriented to the surface was considered as the
favored adsorption geometry. It was found that electrostatic eects play a dominant role
in the interaction. Chemical bonding with charge donation from the CO 5r level to the
empty Sn states and CO polarization was also found. Unlike for TiO2 [355] or Cu2O
[356,357] no back donation from the oxide to the CO molecule was observed. Overall,
the CO adsorption was rather weak with an adsorption energy of 0.25 eV, but a large
vibrational frequency shift of 80 cm1 towards higher values was predicted. Ciriaco
et al. [358] calculated the inuence of CO adsorption on the electronic structure of
SnO2. CO adsorption at Sn sites resulted in a shift of the Fermi level. They also discussed
the possibility of abstraction of lattice oxygen to from CO2 and found that this is a thermodynamically disfavored process.
8.2.3. Carbon dioxide CO2
Cluster and periodic ab initio calculations showed that the stoichiometric SnO2(1 1 0)
surface is rather unreactive towards CO2 [359]. CO2 interacts electrostatically with vefold
Sn cations with the molecular axis perpendicular to the surface. On the bridging oxygen
atoms CO2 is chemisorbed with the formation of a surface carbonate. This process is however almost thermoneutral and the chemisorbed state is metastable.
Infrared spectroscopy of CO2 adsorption on SnO2 powder identied mono- and bidentate carbonate species [351]. This indicates interaction of CO2 with lattice oxygen.
No single crystal experiments of CO2 adsorption have been reported.

126

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

8.2.4. Nitrogen dioxide NO2


DFT calculations of NO2 adsorption on SnO2(1 0 1) in view of an observed gas response
of SnO2 nanoribbons (or nanobelts) (see Section 9) were performed [349]. It was found
that NO2 adsorbs either with a single bond to a single vefold coordinated Sn surface
atom or in a bidentate structure to two Sn atoms. The dierence in the binding energy
between these two congurations is small implying that a bidentate bonded NO2 molecule
can easily break one bond, rotate around the remaining bond, and re-bond to another Sn
atom. Thus NO2 molecules are expected to be mobile at room temperature. If two NO2
molecules meet at the surface an oxygen atom is transferred from an NO2 molecule to
the other forming a strongly bidentate bonded NO3 and a weakly bonded NO that may
desorb from the surface. The NO3 molecule is not expected to be mobile at room temperature. There is signicant charge transfer from the SnO2 substrates of 0.4 electrons to the
NO3 molecule. This should result in a drop of the conductivity of the nanoribbons. Nitrogen K-edge XANES spectra conrmed the presence of NO2 and NO3 at the surface of
these nanoribbons with NO3 being the dominant species [349].
8.2.5. Ammonia NH3
Adsorption of Lewis base molecules like NH3 were used to probe the acidity of surface
adsorption sites on the SnO2(1 1 0) surface [360]. Only NH3 desorption was detected in
TPD studies from SnO2(1 1 0) surfaces at dierent reduction states, implying that NH3
does not dissociate at tin oxide surfaces. A downward band bending and a surface dipole
moment were observed after adsorption of NH3. The strongest band bending and also
strongest dipole moment was present at the reduced surface. From these studies a stronger
interaction of NH3 with Sn2+ than Sn4+ surface atoms was concluded. This is surprising
since Sn4+ is supposed to have a higher acidity and is therefore expected to interact stronger with a Lewis base. Formation of bonds between NH3 and the Sn surface atom of more
covalent character, rather than ionic, involving Sn 5s and 5p related bands may explain
these results.
8.2.6. Methanol (CH3OH)
The adsorption site requirements for methanol dissociation was investigated on
SnO2(1 1 0) surfaces at dierent reduction states [212]. In TPD studies the only desorption
products beside methanol were formaldehyde (H2CO) and water. The conversion rate of
methanol to formaldehyde was found to be strongly dependent on the reduction state of
the sample. Formaldehyde production initially increased with a reduction of the surface
but then decreased for highly reduced surfaces. This was interpreted in terms of the coordination and valence state of the available Sn sites. Fourfold coordinated Sn2+ sites, created by desorption of the bridging oxygen atoms from the stoichiometric (1 1 0) surface,
were deemed to be most active for methanol dissociation. Further reduction of the surface,
i.e. formation of in-plane oxygen vacancies, was assumed to reduce the reactivity of Sn2+
sites. These two sites were also believed to cause the two desorption peaks for formaldehyde at 450 K and at 540 K.
Infrared spectroscopy of methanol adsorption on SnO2 powder showed that methanol
chemisorbed to give methoxide species which were readily oxidized to a surface formate at
320 K [361]. More recent FTIR studies of methanol adsorption were reported by Ouyang
et al. [362]. They identied methanol, two kinds of methoxide species, monodentate
(1165 cm1) and bi-(or poly-)dentate (1056 cm1), and OH after methanol exposure at

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

127

310 K. The methanol and the monodentate methoxide species disappeared after heat treatment at 373 K. A new band assigned to dioxymethylene (or adsorbed formaldehyde)
appeared. These FTIR studies on SnO2 powder appear to be in good agreement with
the SnO2(1 1 0) single crystal TPD studies described above.
Methanol adsorption on SnO2 thin lms was probed by XPS [363,364]. Reduced surfaces were prepared by Ar+ sputtering and oxidized surfaces were obtained by exposing
the presputtered surface to oxygen at RT. This procedure produced oxygen species at
the surface with dierent O 1s core level components. Such an oxidized surface was in
favor of abstraction of hydrogen from adsorbed methanol, while a sputtered and thus
heavily reduced surface favored dissociation of the CO bond in methanol.
A theoretical study of the adsorption and dissociation of methanol on stoichiometric
SnO2(1 1 0) surfaces was carried out by Calatayud et al. [365]. Dissociative adsorption

via CH3OH ) CH3O + H+ and via CO bond cleavage, i.e. CH3 OH ) CH
3 OH
were considered. For one monolayer coverage the CO bond cleavage process was
favored. This appears to be in contradiction to the experimental results discussed above
where methoxide and formaldehyde production was observed. However, oxygen vacancies
were important in these experiments and defects were not taken into account in the theoretical studies.
8.2.7. BF3
The basicity of lattice oxygen sites of SnO2(1 1 0) was probed by adsorption of BF3
[366]. On reduced surfaces BF3 interacts molecularly with three-coordinated surface oxygen sites to form a Lewis acid/base adduct. BF3 desorbs intact from these oxygen sites at
300 K and thus the desorption temperature provides a reasonable measure of the basicity
of these sites. However, BF3 reacts irreversible with the more labile twofold coordinated
bridging oxygen atoms of the stoichiometric (1 1 0) surface. No distinctive BF3 desorption
feature is observed. Instead the gas phase reaction products HF and F2 are detected and
build up of boron oxide at the surface is observed.
8.2.8. Formic acid (HCOOH)
Formic acid adsorption and reaction with SnO2(1 1 0) surfaces was studied with XPS
and TPD by Gercher and Cox [367]. It was found that formic acid reacts with the surface
to form CO, CO2, H2CO and H2O with the conversion being dependent on the initial surface composition. From these studies it was concluded that at 170 K formic acid adsorbs
molecularly and dissociates to form formate. The acidic proton from the formic acid dissociation reacts with lattice oxygen to form water at 400 K. Formate decomposition is the
limiting step in the production of all the other reaction products. The dissociation probability showed an apparent linear dependence on the density of in-plane lattice oxygen
atoms, while the coordination and valency of the Sn cations appeared to have little eect
on the extent of formic dissociation. CO and H2CO production near 530 K was determined to be due to formate decomposition at cations associated with in-plane vacancies
while CO2 production at 600 K was believed to be due to formate decomposition at cations away from in-plane oxygen vacancies. Consequently the oxygen concentration at the
surface inuenced the selectivity of the process.
Adsorption of formic acid on the reduced surfaces of SnO2(1 1 0) formed by high temperature vacuum annealing was studied by Irwin et al. [368]. Both the high temperature
pseudo-1 1 structure as well as the 1 2 reconstruction was studied. It was found that

128

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

for both surface structures formic adsorbed molecularly at 105 K and is desorbed by
annealing to 375 K. A conversion of the 1 2 reconstruction to a structure with a 1 1
LEED pattern was observed after desorption of formic acid. This indicates some interaction of the formic acid with the oxygen atoms in the 1 2 reconstruction.
8.2.9. Benzene (C6H6)
Benzene adsorption on stoichiometric and reduced SnO2(1 0 1) surfaces was studied by
ARUPS [369]. No adsorption at RT was observed. At 120 K benzene adsorbed molecularly on both surfaces but no preferential orientation of the molecule could be detected
by polarization dependent studies. The dierence in the work function of 1 eV between the
reduced and the stoichiometric surface caused a shift of the binding energies of the molecular orbitals of the adsorbed benzene by the same amount. This is demonstrated in the
UPS spectra shown in Fig. 49. Thus this study demonstrates the possibility of tuning
the band alignment between a SnO2 electrode and an organic lm by controlling the composition of the oxide surface.
8.2.10. Water
The omnipresence of water and the consequently unavoidable contamination of any
surface with water exposed to ambient conditions ensures continuous interest of surface
scientists in the interaction of water with inorganic materials [370,371]. The adsorption
of water on tin-dioxide surfaces is also of importance for its applications as a catalyst, electrode material, and as gas sensing material. For these applications the presence of water at
the surface impacts device operations considerably. It has been reported for instance that
water adsorption facilitates the catalytic conversion of CO to CO2 over SnO2 catalysts.
Furthermore, many gas sensors are intended to detect harmful gases in the environment
under ambient conditions and water is constantly present in these applications. In the
use of SnO2 gas sensors it was early on recognized that SnO2 was not just sensitive to
inammable gases but also exhibits sensitivity to humidity [372]. This gas sensitivity
sparked early interest in the adsorption of water on SnO2 gas sensor materials. Water
adsorption on pressed SnO2 powder or thin lms prepared by spray pyrolysis was studied
by transmission infrared spectroscopy [351], temperature programmed desorption in a Hecarrier gas [373,337,353], and conductivity measurements [337,374]. Thornton and Harrison [361] found that molecular water was removed at 150 C under atmospheric pressure,
while hydroxyl groups started to desorb at 250 C but some remained up to 500 C. These
infrared spectroscopy results were taken to interpret later TPD studies, i.e. desorption
peaks below 200 C were assigned to desorption of molecular water and various high temperature desorption features were believed to originate from recombination of hydroxyls
[337,373]. The TPD results diered for samples preannealed in a He atmosphere from
those annealed in oxygen [373]. More hydroxyl groups per unit surface area were found
on the oxygen pretreated samples. In addition oxygen desorbed concurrent with water
at high temperatures (620 C) only for oxygen pretreated samples. This dependence on
sample pretreatments suggested that oxygen vacancies and other defect sites at the surface
may play an important role for water dissociation. The authors also pointed out that much
more dissociated water has been observed on SnO2 samples than for TiO2 under identical
experimental conditions [375]. It is well established that water adsorption on SnO2 gas sensors increases its conductivity [337,376379,344]. Yamazoe et al. [337] correlated
the decrease in conductivity with the high temperature desorption peak in TPD and

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

129

Fig. 49. (a) Photoemission spectrum of benzene in the gas phase. This spectrum is displayed on the same scale as
the photoemission spectra below but is arbitrarily shifted. (b) Valence band spectra of the clean reduced and (c)
stoichiometric SnO2 surfaces. Dierence spectra for various benzene exposures on these surfaces are also shown.
The entire benzene spectrum is shifted by 1 eV to lower binding energies on the stoichiometric surface compared
to the reduced surface. On the stoichiometric surface an additional, small shift of 0.2 eV towards higher binding
energies is observed for submonolayer coverage compared to monolayer coverage for the HOMO band. (d)
Schematic diagram of the band alignment of the HOMO of benzene with the valence band of SnO2(1 0 1) for
reduced and stoichiometric surfaces. Note that all the energy levels in this diagram are referenced to the vacuum
level, while in the photoemission measurements the Fermi edge EF of the SnO2 substrates is the reference (adapted
from Ref. [369]).

consequently concluded that hydroxyl groups cause the change in conductivity upon water
adsorption on SnO2. The complex nature of the polycrystalline and powder samples considered makes it however impossible to make any denitive statement about water adsorption geometries and reactions. To obtain a better fundamental understanding of water

130

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

adsorption on the stoichiometric SnO2(1 1 0) surface as well as on the reduced and stoichiometric SnO2(1 0 1) surfaces DFT calculations and experiments on single crystals have
been recently performed. These studies are summarized in the following for two low index
SnO2 surfaces.
8.2.10.1. Water on the SnO2(1 1 0) surface. First principle DFT calculations favor (partial)
dissociative adsorption on the stoichiometric (1 1 0) surface. Goniakowski and Gillan
found that for SnO2(1 1 0) dissociated water is the most stable conguration [380]. However, as pointed out by Lindan [381] because inter-molecular bonding was not explicitly
addressed in these calculations these conclusions are of limited use [382]. Lindan studied
half and full monolayer coverage of water and found that in both cases complete dissociation is the most energetically favored adsorption. However, while for half monolayer coverage no molecular adsorbed water was stable, at full coverage a half dissociated/half
molecular adsorbed water geometry was found to exhibit an only slightly lower adsorption
energy compared to the fully dissociated structure. The considered adsorption geometries
and the adsorption energies are shown in Fig. 50. Lindan also compared the results for
SnO2(1 1 0) with the much more frequently studied TiO2(1 1 0) surface. He suggests that
geometrical dierences, i.e. a 5% larger lattice constant of SnO2 compared to TiO2, is signicant in explaining dierences in water adsorption on these two surfaces. In his calculations for full monolayer water coverage at the TiO2(1 1 0) surface, the mixed as well as
the molecular adsorption states have a 0.1 eV higher adsorption energy compared to the
fully dissociated adsorption states. He explains this contrast between TiO2 and SnO2 by
the larger separation between adsorbed water on SnO2 compared to water on TiO2. On
TiO2 water can form H bonds between neighboring water molecules and thus stabilizes
molecular adsorption, while on SnO2 such hydrogen bonding is less advantageous because

Fig. 50. Schematic representation of the adsorption geometries of water on SnO2(1 1 0) by DFT calculations.
Oxygen, tin and hydrogen are represented by large, small lled and small open circles respectively. Half coverage
(top) and monolayer coverage (bottom) are considered. Adsorption energies are also given (from Ref. [381], 
(2000), Elsevier).

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

131

of the greater separation. Bates [383] conrmed many of Lindans conclusions for the full
monolayer coverage on SnO2(1 1 0) and extended the studies using a larger surface unit
cell.
Although all the theoretical studies point towards preferential dissociation of water on
SnO2(1 1 0) experimental studies have not unambiguously conrmed this conclusion. Gercher and Cox [384] performed UPS and TPD studies of water adsorption on SnO2(1 1 0)
single crystals. Molecular desorption peaks at 200 and 300 K were observed and desorption at 435 K was attributed to OH disproportionation. In their study they found only 10
15% dissociated water on the stoichiometric, reduced (all bridging oxygen atoms removed)
and highly defective surfaces (with in-plane oxygen vacancies). An increase in dissociation
to about 35% was observed on a less defective surface. This suggested that a limited number of in-plane vacancies promotes dissociation, but further increase in vacancies decreases
the dissociation probabilities. Although the samples were prepared with care it is very difcult to make quantitative statement about the defect concentration and structure on the
(1 1 0) surface [209]. This also makes it questionable if the observed dissociated water is due
to defects or occurs on perfect surfaces.
A downward band bending of 0.1 eV was observed for water adsorption on stoichiometric SnO2(1 1 0) surfaces [341]. This implies an electron transfer from water to the substrate and is consistent with the increase in conductivity of SnO2 gas sensors in a humid
atmosphere.
8.2.10.2. Water on the SnO2(1 0 1) surface. Water adsorption on the (1 0 1) surface was
studied on both the stoichiometric and reduced bulk terminations (see Fig. 16(d)) by
UPS measurements and DFT calculations [385]. A strong dierence in the water adsorption was observed depending on the oxidation state of the surface. Adsorption studies at
110 K are shown in Fig. 51(a)(c). The shift in the binding energy of the water molecular
orbitals is a consequence of the dierent work function of the reduced versus stoichiometric surface (see Section 8.2.9). On the stoichiometric surface weak additional features
can be observed in the dierence spectra (spectra that have the contribution of the clean
surface subtracted from the water exposed samples) that may be assigned to orbitals of
OH from dissociated water. A stronger dierence is observed between the two surfaces
if water is adsorbed at higher temperatures (160 K) (Fig. 51(d) and (e)). At this temperature signicant water adsorption is observed on the stoichiometric surface only. Furthermore, there is clear indication of dissociated water in the dierence spectrum for the
stoichiometric surface. In addition to valence band photoemission, band bending eects
upon water adsorption was also investigated. Fig. 51(c) shows that water adsorption at
110 K causes a much stronger downward band bending on the stoichiometric surface compared to the reduced surface. This is consistent with the notion that dissociated water
causes most of the band bending and thus is responsible for the humidity gas response
of SnO2.
The dierences in the water adsorption on the reduced and stoichiometric SnO2(1 0 1)
surfaces has been veried by DFT calculations. Fig. 51(f)(h) shows the results of these
calculations. On the reduced surface only molecular water is stable and adsorbs weakly,
this is most likely because of the lack of surface oxygen that could accept hydrogen.
For the stoichiometric surface both molecular and dissociative adsorption is possible,
however, dissociation is thermodynamically favored and exhibits the highest adsorption
energy of the congurations considered.

132

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Fig. 51. Water adsorption on the reduced and stoichiometric SnO2(1 0 1) surface studied by UPS measurements
and DFT calculations. Adsorption at 110 K is shown in (a)(c) for the reduced surface (a) and stoichiometric
surface (b). Band bending eects upon water exposure is plotted in (c) for the two surfaces. UPS measurements
for water exposure at 160 K is shown in (d) for the reduced surface and in (e) for the stoichiometric surface. Only
on the stoichiometric surface (e) adsorption and dissociation of water is observed at 160 K. Models of water
adsorption derived from DFT calculations are shown in (f) for the reduced surface and (g) and (h) for the
stoichiometric surface for molecular and dissociative water adsorption, respectively. Calculated adsorption
energies are also shown (adapted from Ref. [385]).

9. Nanomaterials
A variety of self-organized tin oxide materials with two or all three dimensions at the
nano- to micro-meter scale have been recently discovered and characterized. The most
often utilized approach for synthesizing nanomaterials is a simple vapor phase transport
method [386]. A schematic diagram of the equipment used for such a synthesis is shown in
Fig. 52. Tin oxide powder is evaporated at one end of a furnace and subsequently transported in a gas ow along the furnace tube. The tin oxide molecules then condense in a
colder region of the furnace. This experimental set-up is similar to the vapor phase transport technique that has been employed for the growth of macroscopic single crystals [387].
In fact, the presence of a wool-like byproduct has been reported for the growth of single
crystals, but it was not the sought single crystal material and thus discarded as waste. It
has not been up to recently that the nature of this wool-like material was investigated.
It was found that it contains ribbon- (or belt) -like structures several hundred micrometer
long, and with a rectangular cross-section of 1030 nm times 80300 nm. Furthermore,
these nanobelts have a single-crystalline rutile structure with the side faces being {1 0 1}

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

133

Fig. 52. Schematic diagram of the experimental set-up for the growth of tin oxide nanomaterials by a vapor
phase transport method (from Ref. [5],  (2002) American Chemical Society).

and {1 0 0} for the wide and narrow sides of the belt, respectively. Fig. 53 shows TEM
images of SnO2 nanobelts. Tin oxide is not unique in the formation of nanobelts, many
other semiconductors and oxide materials also self-organize in similar nanostructures.
Most prominently, in the context of this review, ZnO and In2O3 form nanobelts
[18,388]. The growth mechanism of SnO2 nanobelts is still not entirely resolved. It is, however, important to point out that the SnO2 nanostructures do not require a hetero-metal
catalyst to grow, contrary to many other materials. At this point it can, however, not be
excluded that the growth is self-catalyzed, i.e. that a thin metallic Sn layer exists at the end
of the nanobelt that catalyzes the growth reaction. It is more likely, though, that the
growth of SnO2 nanobelts proceeds via a vapor solid process only, where SnO or SnO2
molecules arrive at a nucleated nanobelt and attach themselves to the rough growth front.
In such a scenario it is assumed that newly arriving molecules at the side faces of the nanobelt are mobile enough, at the relatively high growth temperatures, to diuse to the growth
front where they are attached to the belt (or re-evaporate from the side faces). Although

Fig. 53. TEM results of SnO2 nanoribbons grown by a vapor phase transport method displayed in Fig. 52. (a)
TEM bright-eld image of SnO2 nanoribbon, showing strain contrast introduced by bending of the nanoribbon.
(b) Cross-sectional TEM image of a nanoribbon. (c) HRTEM recorded near the edge of a SnO2 nanoribbon
along the direction perpendicular to the wide surface of the nanoribbon, where columns of Sn cations are clearly
resolved. The inset shows the corresponding electron diraction pattern (from Ref. [19],  (2001), Elsevier).

134

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

kinetics obviously plays an important role in the formation of these nanobelts this growth
scenario implies that the side faces have a low energy that disallows growth of these faces.
It is interesting, however, that the lowest energy surface of stoichiometric SnO2, the (1 1 0)
surface, is not present in these nanobelts. As pointed out in Section 4.2.2 the surface
energy depends on the oxygen chemical potential and the surface termination may not
be stoichiometric under the growth conditions. Thus under the thermodynamic conditions
during growth the {1 0 1} surfaces may be energetically favored over the {1 1 0} termination. Also, the thermodynamic stability of the nanobelts has been demonstrated by using
a Wul-construction equation [197].
Apart from nanobelts a variety of other nanostructured materials can be grown by the
vapor transport technique. The kind of materials obtained can be controlled by choosing
the parameters such as evaporation material (SnO2 or SnO), temperature, gas ow, and
gas composition. Rutile structured nanotubes, and orthorhombic structured nanowires
have been grown. Also, sandwiched nanobelts with a rutile core but an orthorhombic
outer layer have been found [21]. The orthorhombic phase of SnO2 is the thermodynamic
stable form at high pressures [159] (see Section 3) but it is surprising to nd it in nanostructured materials. Tin monoxide nanodiskettes have been found to form in the low temperature region between 200 and 400 C within the furnace [5,6]. SEM images of these

Fig. 54. Electron microscopy micrographs for a variety of tin-oxide nanomaterials. (a) SnO nanodisks (from Ref.
[5],  (2002) American Chemical Society), (b) shbone-like SnO2 nanoribbons (from Ref. [389],  (2003),
Elsevier), (c) SnO2 cubes (from Ref. [390],  (2004) American Chemical Society), (d) SnO2 nano-needles (from
Ref. [395], reproduced by permission of The Royal Society of Chemistry), (e) SnO2 nano-box-beams (from Ref.
[396],  (2004), WILEY-VCH), and (f) SnO2 nanorods (from Ref. [398],  (2004) American Chemical Society).

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

135

structures are shown in Fig. 54(a). These nanodiskettes can be oxidized to SnO2 in an oxygen atmosphere. It was suggested that the oxidation proceeds in two steps (see also Section
3) from SnO to Sn3O4 and then from Sn3O4 to SnO2 [5].
By using a mixture of Sn powder and Fe3(NO3)3 Hu et al. obtained shbone-like nanoribbons (Fig. 54(b)) [389]. Photoluminescence measurements showed a strong green emission at 500 nm of these strongly branched shbone-like nanostructures. Ramgir et al.
added 0.1 to 1 wt.% RuO2 to the reaction mixture and found the formation of SnO2 nanobipyramids and cubes (Fig. 54(c)) in regions of the furnace that were between 200 C and
500 C, and 950 C respectively [390]. It has been proposed that RuO2 acts as a promoter/
nucleating agent for these nanomaterials.
Other techniques have also been employed to synthesize nanostructures. A slight modication of the vapor phase transport is the use of a carbothermal reduction process, which
lowers the synthesization temperature [391,392]. Laser ablation can also be tuned to result
in the formation of SnO2 nanoribbons [393]. A very dierent method to produce SnO2
nanoribbons was described by Ye et al. By adding cetyltrimethylammonium bromide
and NaOH solution to a 0.015 M SnCl2 solution resulted in the formation of SnO2 nanoribbons [394]. SnO2 nanoneedles were produced by carrying SnH4 in an O2 ow into a
glow discharge. The nanoneedles were deposited on a gold coated porous silicon substrate
placed below the discharge plasma (Fig. 54(d)) [395]. Nano-box-beams, i.e. hollow SnO2
boxes of 0.52 lm in width and up to 7 lm in length were produced in a uniform distribution on a quartz substrate via combustion chemical vapor deposition in the open atmosphere [396]. The side walls of the boxes are SnO2 {1 1 0} planes and the growth direction is
the [0 0 1] crystallographic orientation. The hollow boxes are capped by which could be
{1 0 1} faces (Fig. 54(e)). SnO2 nanorods with a diameter of 4813 nm were produced by
heating 1,10 phenanthroline-capped Sn nanoparticles (25 nm) in a NaCl ux [397]. These
SnO2 nanorods often exhibit V-shaped junctions that were identied as twin boundaries
with a twin plane of (1 0 1) and a twin direction of 1 0 1 . SnO2 nanorods with diameters
of the order of molecules (3.4 nm) were synthesized in a solution-phase growth [398]. A
Sn4+ precursor (SnCl4 5H2O, 0.001 mol) was dissolved in a basic mixture of alcohol and
water (pH  12). After heating at 150 C for 12 h a white precipitate could be collected.
Analysis showed that the precipitate consisted of uniform rutile structured SnO2 nanorods
with a mean aspect ratio of 4:1 (Fig. 54(f)). Other solution based techniques resulted in
nanorods on a slightly large length scale [399].
There exist a number of potential applications of these nanomaterials, in particular the
very long nanobelts have attracted attention for device applications. One problem of polycrystalline gas sensors is grain coarsening during operation that causes an alteration of the
gas response. Gas sensors fabricated from single crystalline nanobelts could overcome this
problem but still exhibit a high surface area to volume ratio essential for a strong gas
response. Furthermore, with further advancement of nanotechnology gas sensors based
on single nanobelts can be envisaged. Already, some studies of the gas sensitivity of nanobelts have been undertaken. Comini et al. demonstrated the gas sensitivity of the electrical
conductance of SnO2 nanobelts to CO, NO2, and ethanol [400]. The adsorption of NO2 on
these nanoribbons with {1 0 1} and {1 0 0} surfaces was further investigated by DFT calculations [349,401]. See Section 8.2.4 for details. The electrical properties of nanobelts
was investigated in terms of depletion layers induced by charged surface adsorbates. It
was concluded that the quasi-two-dimensional nanobelts should be superior to polycrystalline lms [402].

136

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

Field eect transistors have been fabricated by placing a single nanobelt on prefabricated electrodes [403]. Furthermore, the gas sensitivity of these FETs has been demonstrated. In another approach, thin lm transistors were fabricated by growing SnO2
nanobelts on Au/Pt electrodes in situ [404].
The mechanical strength of nanobelts was investigated by nanoindentations [405]. In
these studies it was demonstrated that nanomachining of these nanobelts with an AFM
tip was possible. The thermal conductivity of a single nanobelt was found to be signicantly lower than the bulk values of SnO2 [406]. This may be explained by increased phonon boundary scattering and a modied phonon dispersion. Other applications and
properties of nanobelts are discussed in a recent review article by Wang [388].
10. Synopsis and outlook
A review of the materials science of tin oxide with an emphasis on the surface properties
of SnO2 is being presented. The other main form of tin oxide, stannic oxide (SnO), has not
been extensively studied. This is because stannic oxide is not as stable as SnO2, which complicates the preparation of stannic oxide samples. Furthermore, it is not of any technological relevance. Nevertheless, Section 3 gives a comparison of the properties of these two
forms of tin oxide. The existence of Sn(II)O in addition to Sn(IV)O2 illustrates the dual
valency of Sn that is crucial in understanding the surface properties of the SnO2 low index
surfaces. In Section 4 it is demonstrated that depending on the oxygen chemical potential of
the system (i.e. oxygen pressure and temperature) surfaces with either Sn(IV)O2 or Sn(II)O
composition are thermodynamically stable. Consequently, preparation under UHV conditions (low oxygen chemical potential) usually results in a reduced surface. However, surfaces prepared at high oxygen pressures or with atomic oxygen remain stoichiometric in
UHV at temperatures below 300 C and thus allowing their study by surface science techniques. The easy variation of the oxygen composition of SnO2 surfaces is fundamental for
the understanding of many of its surface properties discussed in this review.
The three main applications of SnO2 are discussed in Section 2. These applications are
the use of SnO2 as a material that combines electrical conductivity with optical transparency, as a heterogeneous oxidation catalyst, and as a solid state gas sensor. Although these
applications relate to dierent elds of research, there are some interesting connections
where the same fundamental physical and chemical properties of SnO2 are responsible
for its suitability in these dierent applications.
The suitability of SnO2 as a transparent conductor arises from its band structure. Dissallowed interband electron excitations in the visible due to a large fundamental band gap
and an internal band gap in the conduction band ensure transparency even for large
charge carrier concentrations. On the other hand, the strongly dispersing s-type conduction band with its minimum at the C-point causes a low eective mass and highly mobile
conduction electrons. The high charge carrier mobility and the strongly dispersing conduction band is also an important ingredient to make it a good gas sensing material.
The basic concept of gas sensing is the acceptor or donor like adsorption of molecules
that consequently induces a band bending in the Debye layer of the SnO2 surface region.
This band bending causes the conductivity change and thus the gas response. Although
this is a simple mechanism, the details of the adsorption and charge transfer of specic
molecules are complex and more research is needed to gain a fundamental understanding
of the interaction of specic molecules with SnO2.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

137

Adsorbed oxygen plays an outstanding role in the sensing of combustible gases. Oxygen
can be present in various charge states at the surface (Section 8.2.1) and reaction of combustible molecules with these ionosorbed oxygen species causes an increasing conductivity.
The reaction of molecules with chemisorbed oxygen is thus also a route for oxidation
catalysis. But is this the dominant way a SnO2 oxidation catalyst works? In Section 4 it
is shown that all low index SnO2 surfaces easily reduce, i.e. surface lattice oxygen desorbs
from the stoichiometric surface. Thus it is also plausible to propose a catalytic reaction
that proceeds via the consumption of lattice oxygen by adsorbed molecules, i.e. the so
called Marsvan Krevelen mechanism. For instance, SnO2 is known to be a good gas sensing material for CO. This implies that CO reacts with chemisorbed oxygen species to form
CO2 and this causes the gas response. Does CO, however, also react with lattice oxygen to
form CO2? The literature (see Section 8.2.2) is ambiguous in this respect.
Surface science studies of larger molecules (e.g. methanol and formic acid, see Sections
8.2.6 and 8.2.8 respectively) show that these molecules react with SnO2 surfaces without
the presence of (co-)adsorbed oxygen. However, the reaction products seem to depend
on the defect structure of the surface stressing the importance for the preparation of
well-dened surfaces with known structure and oxygen concentration. In order to obtain
surfaces with known properties the authors of this article are advocating the use of
SnO2(1 0 1) or (1 0 0) surfaces instead of the predominantly studied (1 1 0) surface. In Section 4 the structure and composition of low index SnO2 surfaces are described in detail.
The (1 1 0) surface exhibits complex structures that are not well understood. The (1 0 1)
and (1 0 0) surfaces on the other hand can be prepared with two bulk terminations with
and without terminating, bridging oxygen atoms. In the context of surface chemistry it
is important that the preparation of stoichiometric and reduced (1 0 1) and (1 0 0) surfaces
allows probing the surface chemical properties with and without terminating surface lattice oxygen. Such a study is described in Section 8.2.10 for the adsorption of water. It
is shown that water dissociates on the stoichiometric surface (with bridging oxygen atoms)
but not on the reduced SnO2(1 0 1) surface. This may be easily understood by the availability of Brnsted base sites for accepting a proton from the water on the stoichiometric surface and the lack of such sites on the reduced surface.
The dierent chemical surface properties of the stoichiometric and reduced surface open
the possibility of tuning the moleculesurface interaction by removal of surface oxygen
under vacuum. Thus the surfaces can be functionalized [407] for selective adsorption by
controlling their surface composition. A general projection would be that weak acids
may dissociatively adsorb on the stoichiometric surfaces but adsorb only weakly on the
Sn-terminated reduced surface. In this context, it may also be noteworthy that the reduced
SnO2 surface forms Sn 5s surface states. These Sn 5s surface electrons can be described as
lone pairs, which are not very chemically active. Therefore no dangling bonds are created by removing the surface oxygen and this explains the weak adsorption of, for example, water on the reduced SnO2(1 0 1) surface (see Section 8.2.10).
The electronic band structure of the stoichiometric and reduced low index SnO2 surfaces is described in detail in Section 5. The main conclusion from these electronic structure studies is that the Sn 5s derived surface states on the reduced SnO2 surfaces are lying
just above the bulk valence band maximum. Thus, even for the reduced surface a significant band gap is maintained. The electronic structure of the (1 1 0) surface is complex
and depends critically on the preparation conditions of the sample, in agreement with
the complex surface structures discussed in Section 4. The (1 0 1) and (1 0 0) surfaces, on

138

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

the other hand, can be prepared reliably with and without Sn 5s surface states corresponding to the surface oxygen composition described in Section 4.
The easily adjustable surface oxygen composition also allows for tuning the work function of SnO2 by 1 eV (see Section 8.2.9). This has implication for the use of SnO2 as a
transparent electrode material for contacting for instance optoelectronic materials. In
optoelectronic devices the band alignment between the valence band of the hole-injecting
oxide electrode and the HOMO in, for example, organic LED materials is crucial for ecient charge transfer and device operation. This band alignment depends on the work
function of the electrode and thus by controlling the surface composition of SnO2 the band
alignment can be tuned. For the technologically more important ITO it is well documented
that oxygen plasma treatment improves the charge carrier injection properties from the
electrode into optically active molecular lms [408]. However, there exists no atomic scale
understanding on how this comes about. The similarities with SnO2 may lead to the speculation that In2O3 also easily reduces its surface and causing a change in work function.
Thus the fundamental understanding gained on SnO2 surfaces may be applied, or at least
provide ideas, for the behavior of related materials.
Although there remain still many open questions regarding the surface science of pure
SnO2, additives need also to be considered if an attempt is being made to understand practical applications of SnO2. In the use of SnO2 as a transparent conductor, additives, such
as Sb, F, Cl, play the role of donor atoms to increase the charge carrier concentration and
are thus fairly well understood in the bulk. However, eects such like surface segregation
of Sb would be worthwhile to investigate by surface science techniques on single crystal
samples. In gas sensors and heterogenous catalysts the role of additives is often obscure.
Basic concepts of how additives contribute to the operation of catalysts and gas sensors
are described in Sections 2.2 and 2.3.2, respectively. In practice, however, a fundamental
description of the function of many specic additives has not been gained. This is mainly
due to the immense diculties in obtaining structural and electronic characterizations of
doped SnO2 materials under conditions relevant for their applications. In Section 7, the
few surface science studies of well-dened systems are reviewed. This is a start, but more
studies are needed. In particular a study of the changes additives undergo at high pressure
conditions would be important in order to make statements about their role under more
realistic operation conditions. Further, studies of lattice doping by for instance ion
implantation in single crystals or by growing doped epitaxial lms would be helpful for
gaining a better understanding of additives in SnO2. A new development in dopant-modied SnO2 is the discovery of dilute ferromagnetism for Co and Fe doped SnO2 (Section
7.2.2). This opens new avenues for the use of SnO2 in spintronics applications.
The main obstacle that restricts surface science investigations of SnO2 is the lack of easily
available single crystals. Novel properties found in SnO2 related materials such as the above
mentioned dilute ferromagnetism of doped SnO2 increases the interest in this material and
may again sporn the growth of single crystals. For instance, crystals grown in Boatners lab
have been recently reported [37]. However, there are also other routes for the preparation
and study of SnO2 surfaces. In Section 6 the growth of epitaxial SnO2 lms has been
described. The best substrate to date for the growth of SnO2 is r-cut a-Al2O3. On this substrate SnO2 grows in a (1 0 1) orientation. The insulating character of this substrate may not
be a limiting factor for many surface science techniques, though, because the SnO2 lms can
be made suciently conductive to allow the use of ionizing radiation without signicant
surface charging and even STM imaging on these lms is possible. Furthermore, the good

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

139

thermal conductivity of alumina may also allow thermal desorption measurements without
signicant thermal gradients in the lm. To date no metal substrate has been reported for
the growth of single domain SnO2 epitaxial lms. Small SnO2(1 0 1) crystallites can, however, been grown on Pt(1 1 1). This is discussed in Section 6.3.
The SnO2 nanomaterials, described in Section 9, are likely to grab most attention in the
near future. One aspect of these newely synthesized materials is to exploit their properties
in applications. For instance, the use of SnO2 nanobelts as gas sensors is being pursuit.
Since these materials have a high surface to volume ratio they may be ideal for gas sensing
applications. Furthermore, nanobelts may have the additional technological advantage
that unlike for porous, polycrystalline lms, aging, i.e. sintering of grains, is not expected
to be an issue. From a surface science perspective these materials are interesting because
they are of single crystalline nature with well-dened surfaces and their dimensions are not
small enough to exhibit any properties signicantly dierent from bulk SnO2. Thus conventional surface science studies on single crystal surfaces are excellently suited to understand processes at the surfaces of these nanomaterials. In a sense the advent of these
nanomaterials is a remarkable turn of events. Before these materials were discovered it
was widely accepted that surface science studies had to go to ever more complicated morphologies in order to capture the complex nature of polycrystalline and powder materials
used in many applications. Now that it has been shown that SnO2 nanobelts make excellent gas sensing materials the materials gap between surface science studies and applications may be closed from the application side.
Acknowledgment
The authors like to thank the Donors of the American Chemical Society Petroleum Research Fund, NSF, and NASA for their nancial support of our research on the surface
properties of tin oxide.
References
[1] W.-X. Li, C. Stamp, M. Scheer, Insights into the function of silver as an oxidation catalyst by ab initio
atomistic thermodynamics, Phys. Rev. B 68 (2003) 165412.
[2] B.L.M. Hendriksen, S.C. Bobaru, J.W.M. Frenken, Oscillatory CO oxidation on Pd(1 0 0) studied with
in situ scanning tunneling microscopy, Surf. Sci. 552 (2004) 229.
[3] H. Over, M. Muhler, Catalytic CO oxidation over rutheniumbridging the pressure gap, Prog. Surf. Sci. 72
(2003) 3.
[4] G. Centi, in: B. Cornils, W.A. Hermann, R. Schlogel, C.-H. Wong (Eds.), Catalysis from A to Z, WileyVCH, Weinheim, 2000.
[5] Z.R. Dai, Z. Wei Pan, Z.L. Wang, Growth and structure evolution of novel tin oxide diskettes, J. Am.
Chem. Soc. 124 (2002) 8673.
[6] Z.L. Wang, Nanobelts, nanowires, and nanodiskettes of semiconducting oxidesfrom materials to
nanodevices, Adv. Mater. 15 (2003) 432.
[7] M.H. Huang, Y.Y. Wu, H. Feick, N. Tran, E. Weber, R. Russo, P.D. Yang, Room-temperature ultraviolet
nanowire nanolasers, Science 292 (2001) 1897.
[8] M.H. Huang, Y.Y. Wu, H. Feick, N. Tran, E. Weber, P.D. Yang, Catalytic growth of zinc oxide nanowires
by vapor transport, Adv. Mater. 13 (2001) 113.
[9] Y.C. Kong, D.P. Yu, B. Zhang, W. Fang, S.Q. Feng, Ultraviolet-emitting ZnO nanowires synthesized by a
physical vapor deposition approach, Appl. Phys. Lett. 78 (2001) 407.
[10] C.H. Liang, G.W. Meng, Y. Lei, F. Phillipp, L.D. Zhang, Catalytic growth of semiconducting In2O3
nanobers, Adv. Mater. 13 (2001) 1330.

140

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

[11] Y.Q. Zhu, W.K. Hsu, M. Terrones, N. Grobert, H. Terrones, J.P. Hare, H.W. Kroto, D.R.M. Walton, 3D
silicon oxide nanostructures: from nanoowers to radiolarian, J. Mater. Chem. 8 (1998) 1859.
[12] X.C. Wu, W.H. Song, K.Y. Wang, T. Hu, B. Zhao, Y.P. Sun, J.J. Du, Preparation and photoluminescence
properties of amorphous silica nanowires, Chem. Phys. Lett. 336 (2001) 53.
[13] Z.L. Wang, R.P. Gao, J.L. Gole, J.D. Stout, Silica nanotubes and nanober arrays, Adv. Mater. 12 (2000)
1938.
[14] X.C. Wu, W.H. Song, W.D. Huang, M.H. Pu, B. Zhao, Y.P. Sun, J.J. Du, Crystalline gallium oxide
nanowires: intensive blue light emitters, Chem. Phys. Lett. 328 (2000) 5.
[15] H.Z. Zhang, Y.C. Kong, Y.Z. Wang, X. Du, Z.G. Bai, J.J. Wang, D.P. Yu, Y. Ding, Q.L. Hang, S.Q. Feng,
Ga2O3 nanowires prepared by physical evaporation, Solid State Commun. 109 (1999) 677.
[16] C.H. Liang, G.W. Meng, G.Z. Wang, Y.W. Wang, L.D. Zhang, S.Y. Zhang, Catalytic synthesis and
photoluminescence of b-Ga2O3 nanowires, Appl. Phys. Lett. 78 (2001) 3202.
[17] Z.G. Bai, D.P. Yu, H.Z. Zhang, Y. Ding, Y.P. Wang, X.Z. Gai, Q.L. Hang, G.C. Xiong, S.Q. Feng, Nanoscale GeO2 wires synthesized by physical evaporation, Chem. Phys. Lett. 303 (1999) 311.
[18] Z.W. Pan, Z.R. Dai, Z.L. Wang, Nanobelts of semiconducting oxides, Science 291 (2001) 1947.
[19] Z.R. Dai, Z.W. Pan, Z.L. Wang, Ultra-long single crystalline nanoribbons of tin oxide, Solid State
Commun. 118 (2001) 351.
[20] Z.W. Pan, Z.R. Dai, Z.L. Wang, Lead oxide nanobelts and phase transformation induced by electron beam
irradiation, Appl. Phys. Lett. 80 (2002) 309.
[21] Z.R. Dai, J.L. Gole, J.D. Stout, Z.L. Wang, Tin oxide nanowires, nanoribbons, and nanotubes, J. Phys.
Chem. B 106 (2002) 1274.
[22] P.D. Yang, C.M. Lieber, Nanorod-superconductor composites: a pathway to materials with high critical
current densities, Science 273 (1996) 1836.
[23] H.-J. Freund, E. Umbach, Adsorption on Ordered Surfaces of Ionic Solids and Thin FilmsSpringer Series
in Surface Science, vol. 33, Springer, Berlin, 1993.
[24] V.E. Heinrich, P.A. Cox, The Surface Science of Metal-oxides, Cambridge University Press, Cambridge,
1994.
[25] C. Noguera, Physics and Chemistry of Oxide Surfaces, Cambridge University Press, Cambridge, 1996.
[26] H.-J. Freund, Adsorption of gases on complex solid surfaces, Angew. Chem. Int. Ed. Engl. 36 (1997) 452.
[27] C.T. Campbell, Ultrathin metal lms and particles on oxide surfaces: structural, electronic and
chemisorptive properties, Surf. Sci. Rep. 27 (1997) 1.
[28] J.P. LaFemina, Total energy computations of oxide surface reconstructions, Crit. Rev. Surf. Chem. 3 (1994)
297.
[29] M.A. Barteau, Site requirements of reactions on oxide surfaces, J. Vac. Sci. Tech. A 11 (1993) 2162.
[30] S.A. Chambers, Epitaxial growth and properties of thin lm oxides, Surf. Sci. Rep. 39 (2000) 105.
[31] D.A. Bonnell, Scanning tunneling microscopy and spectroscopy of oxide surfaces, Prog. Surf. Sci. 57 (1998)
187.
[32] R.J. Lad, Interactions at metal/oxide and oxide/oxide interfaces studied by ultrathin lm growth on singlecrystal oxide substrates, Surf. Rev. Lett. 2 (1995) 109.
[33] H.A. Al-Abadleh, V.H. Grassian, Oxide surfaces as environmental interfaces, Surf. Sci. Rep. 52 (2003) 63.
[34] U. Diebold, The surface science of titanium dioxide, Surf. Sci. Rep. 48 (2003) 53.
[35] W. Weiss, W. Ranke, Surface chemistry and catalysis on well-dened epitaxial iron-oxide layers, Prog. Surf.
Sci. 70 (2002) 1.
[36] S. Surnev, M.G. Ramsey, F.P. Netzer, Vanadium oxide surface studies, Prog. Surf. Sci. 73 (2003) 117.
[37] Y.W. Heo, J. Kelly, D.P. Norton, A.F. Hebard, S.J. Pearton, J.M. Zavada, L.A. Boatner, Eects of high
dose Ni, Fe, Co, and Mn implantation into SnO2, Electrochem. Solid State Lett. 7 (2004) G309.
[38] J.F. Wagner, Transparent electronics, Science 300 (2003) 1245.
[39] R.L. Homan, B.J. Norris, J.F. Wagner, ZnO-based transparent thin-lm transistors, Appl. Phys. Lett. 82
(2003) 733.
[40] S. Masuda, K. Kitamura, Y. Okumura, S. Miyatake, H. Tabata, T. Kawai, Transparent thin lm
transistors using ZnO as an active channel layer and their electric properties, J. Appl. Phys. 93 (2003) 1624.
[41] R.E. Presley, C.L. Munsee, C.-H. Park, D. Hong, J.F. Wager, D.A. Keszler, Tin oxide transparent thin-lm
transistors, J. Phys. D 37 (2004) 2810.
[42] C.G. Granqvist, A. Hultaker, Transparent and conducting ITO lms: new developments and applications,
Thin Solid Films 411 (2002) 1.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

141

[43] B.G. Lewis, D.C. Paine, Applications and processing of transparent conducting oxides, MRS Bull. 25
(2000) 22.
[44] I. Hamberg, C.G. Granqvist, Evaporated Sn-doped In2O3 lms: basic optical properties and applications to
energy-ecient windows, J. Appl. Phys. 60 (1986) R123.
[45] K.L. Chopra, S. Major, D.K. Pandya, Transparent conductorsa status review, Thin Solid Films 102
(1983) 1.
[46] T.J. Coutts, D.L. Young, X. Li, Characterization of transparent conducting oxides, MRS Bull. 25 (2000)
58.
[47] T. Minami, New n-type transparent conducting oxides, MRS Bull. 25 (2000) 38.
[48] A.J. Freeman, K.R. Poeppelmeier, T.O. Mason, R.P.H. Chang, T.J. Marks, Chemical and thin-lm
strategies for new transparent conducting oxides, MRS Bull. 25 (2000) 45.
[49] H. Kawazoe, H. Yanagi, K. Ueda, H. Hosono, Transparent p-type conducting oxides: design and
fabrication of pn heterojunctions, MRS Bull. 25 (2000) 28.
[50] R.G. Gordon, Criteria for choosing transparent conductors, MRS Bull. 25 (2000) 52.
[51] H.L. Hartnagel, A.L. Dawar, A.K. Jain, C. Jagadish, Semiconducting Transparent Thin Films, IOP
Publishing, Bristol, 1995.
[52] Y. Mi, H. Odaka, S. Iwata, Electronic structures and optical properties of ZnO, SnO2 and In2O3, Jpn. J.
Appl. Phys. 38 (1999) 3453.
[53] Z.M. Jarzebski, Preparation and physical properties of transparent conducting oxide lms, Phys. Status
Solidi A 71 (1982) 13.
[54] K.J. Button, D.G. Fonstad, W. Dreybradt, Determination of the electron masses in stannic oxide by
submillimeter cyclotron resonance, Phys. Rev. B 4 (1971) 4539.
[55] G. Zwicker, K. Jacobi, Experimental band structure of ZnO, Solid State Commun. 54 (1985) 701.
[56] K. Ellmer, Resistivity of polycrystalline zinc oxide lms: current status and physical limit, J. Phys. D: Appl.
Phys. 34 (2001) 3097.
[57] C. Kilic, A. Zunger, Origins of coexistence of conductivity and transparency in SnO2, Phys. Rev. Lett. 88
(2002) 095501.
[58] D.S. Ginley, C. Bright, Transparent conducting oxides, MRS Bull. 25 (2000) 15.
[59] J.A. Marley, R.C. Dockerty, Electrical properties of stannic oxide single crystals, Phys. Rev. 140 (1965)
A304.
[60] K. Ishiguro, T. Sasaki, T. Arai, I. Imai, Optical and electrical properties of tin oxide lms, J. Phys. Soc. Jpn.
13 (1958) 296.
[61] M. Nagasawa, S. Shionoya, S. Makishima, Electron eective mass of SnO2, J. Phys. Soc. Jpn. 20 (1965)
1093.
[62] S.-C. Chang, Oxygen chemisorption on tin oxide: correlation between electrical conductivity and EPR
measurements, J. Vac. Sci. Technol. 17 (1980) 366.
[63] E.E. Kohnke, Electrical and optical properties of natural stannic oxide crystals, J. Phys. Chem. Solids 23
(1962) 1557.
[64] H.E. Matthews, E.E. Kohnke, Eect of chemisorbed oxygen on the electrical conductivity of Zn-doped
polycrystalline SnO2, J. Phys. Chem. Solids 29 (1968) 653.
[65] Y. Mizokawa, S. Nakamura, ESR and electric conductance studies of the ne-powdered SnO2, Jpn. J. Appl.
Phys. 14 (1975) 779.
[66] K.C. Mishra, K.H. Johnson, P.C. Schmidt, Electronic-structure of antimony doped tin oxide, Phys. Rev. B
51 (1995) 13972.
[67] O.N. Mryasov, A.J. Freeman, Electronic band structure of indium tin oxide and criteria for transparent
conducting behavior, Phys. Rev. B 64 (2001) 233111.
[68] H. Odaka, Y. Shigesato, T. Murakami, S. Iwata, Electronic structure analyses of Sn-doped In2O3, Jpn. J.
Appl. Phys. 40 (2001) 3231.
[69] O. Lang, C. Pettenkofer, J.F. Sanchez-Royo, A. Segura, A. Klein, W. Jaegermann, Thin lm growth and
band lineup of In2O3 on the layered semiconductor InSe, J. Appl. Phys. 86 (1999) 5687.
[70] I. Hamberg, C.G. Granquist, K.F. Berggren, B.E. Senelius, L. Engstrom, Band-gap widening in heavily Sndoped In2O3, Phys. Rev. B 30 (1984) 3240.
[71] R.L. Weiher, R.P. Ley, Optical properties of indium oxide, J. Appl. Phys. A 37 (1966) 299.
[72] W.G. Haines, R.H. Bube, Eects of heat treatment on the optical and electrical properties of indiumtin
oxide lms, J. Appl. Phys. 49 (1978) 304.

142

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

[73] A.J. Steckel, G. Mohammed, The eect of ambient atmosphere in the annealing of indium tin oxide lms,
J. Appl. Phys. 51 (1980) 3890.
[74] M. Mizuhashi, Electrical properties of vacuum-deposited indium oxide and indium tin oxide lms, Thin
Solid Films 70 (1980) 91.
[75] A.P. Roth, J.B. Webbs, D.F. Williams, Band-gap narrowing in heavily defect-doped ZnO, Phys. Rev. B 25
(1982) 7836.
[76] O. Carporaletti, Electrical and optical properties of bias sputtered ZnO thin lms, Solar Energy Mater. 7
(1982) 65.
[77] Y. Dou, R.G. Egdell, T. Walker, D.S.L. Law, G. Beamson, N-type doping of CdO: a study by EELS and
photoemission spectroscopy, Surf. Sci. 398 (1998) 241.
[78] Y. Dou, T. Fishlock, R.G. Egdell, D.S.L. Law, G. Beamson, Band-gap shrinkage in n-type-doped CdO
probed by photoemission spectroscopy, Phys. Rev. B 55 (1997) R13381.
[79] Y. Dou, R.G. Egdell, Surface properties of indium-doped Cd2SnO4 ceramics studied by EELS and
photoemission spectroscopy, Surf. Sci. 372 (1997) 289.
[80] A.J. Nozik, Optical and electrical properties of Cd2SnO4: a defect semiconductor, Phys. Rev. B 6 (1972)
453.
[81] E. Leja, K. Budzyska, T. Pisarkiewicz, T. Stapiski, Cd2SnO4 thin lms obtained by d.c. reactive sputtering
of CdSn alloys, Thin Solid Films 100 (1983) 203.
[82] T. Arai, The study of the optical properties of conducting tin oxide lms and their interpretation in terms of
a tentative band scheme, J. Phys. Soc. Jpn. 15 (1960) 916.
[83] E. Shanti, A. Banerjee, V. Dutta, K.L. Chopra, Electrical and optical properties of undoped and antimonydoped tin oxide lms, J. Appl. Phys. 51 (1980) 6243.
[84] E. Shanti, A. Banerjee, K.L. Chopra, Dopant eects in sprayed tin oxide lms, Thin Solid Films 88 (1982) 93.
[85] K. Suzuki, M. Mizuhashi, Structural, electrical and optical properties of r.f.-magnetron-sputtered SnO2:Sb
lm, Thin Solid Films 97 (1982) 119.
[86] G. Sanon, R. Rup, A. Mansingh, Band-gap narrowing and band structure in degenerate tin oxide
(SnO2)lms, Phys. Rev. B 44 (1991) 5672.
[87] K.B. Sundaram, G.K. Bhagavat, Optical absorption studies on tin oxide lms, J. Phys. D 14 (1981) 921.
[88] E. Burstein, Anomalous optical absorption limit in InSb, Phys. Rev. 93 (1954) 632.
[89] T.S. Moss, Proc. Phys. Soc. Lond. Ser. B 67 (1954) 775.
[90] M.J. Fuller, M.E. Warwick, The catalytic oxidation of carbon monoxide on tin(IV) oxide, J. Catal. 29
(1973) 441.
[91] M.J. Fuller, M.E. Warwick, The catalytic oxidation of carbon monoxide on SnO2CuO gels, J. Catal. 34
(1974) 445.
[92] G.C. Bond, L.R. Molloy, M.J. Fuller, Oxidation of carbon monoxide over palladiumtin(IV) oxide
catalysts: an example of spillover catalysis, J. Chem. Soc. Chem. Commun. (1975) 796.
[93] G. Croft, M.J. Fuller, Water-promoted oxidation of carbon monoxide over tin(IV) oxide-supported
palladium, Nature 269 (1977) 585.
[94] M.J. Fuller, M.E. Warwick, The catalytic reduction of nitric oxide by carbon monoxide over SnO2CuO
gels, J. Catal. 42 (1976) 418.
[95] F. Solymosi, J. Kiss, Adsorption and reduction of NO on tin(IV) oxide doped with chromium(III) oxide,
J. Catal. 54 (1978) 42.
[96] P.G. Harrison, C. Bailey, W. Azelee, Modied tin(IV) oxide (M/SnO2 M = Cr, LA, Pr, Nd, Sm, Gd)
catalysts for oxidation of carbon monoxide and propane, J. Catal. 186 (1999) 147.
[97] P.W. Park, H.H. Kung, D.-W. Kim, M.C. Kung, Characterization of SnO2/Al2O3 lean NOx catalysts,
J. Catal. 184 (1999) 440.
[98] P.G. Harrison, in: P.G. Harrison (Ed.), Chemistry of Tin, Blackie, Glasgow, 1989 (Chapter 12).
[99] K. Sekizawa, H. Widjaja, S. Maeda, Y. Ozawa, K. Eguchi, Low temperature oxidation of methane over
Pd/SnO2 catalyst, Appl. Catal. A 200 (2000) 211.
[100] H. Widjaja, K. Sekizawa, K. Eguchi, Low-temperature oxidation of methane over Pd supported on SnO2based oxides, Bull. Chem. Soc. Jpn. 72 (1999) 313.
[101] T. Takeguchi, O. Takeoh, S. Aoyama, J. Ueda, R. Kikuchi, K. Eguchi, Strong chemical interaction between
PdO and SnO2 and the inuence on catalytic combustion of methane, Appl. Catal. A 252 (2003) 205.
[102] P.G. Harrison, N.C. Lloyd, W. Daniell, The nature of the chromium species formed during the thermal
activation of chromium-promoted tin(IV) oxide catalysts: an EPR and XPS study, J. Phys. Chem. B 102
(1998) 10672.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

143

[103] P.G. Harrison, N.C. Lloyd, W. Daniell, C. Bailey, W. Azelee, Evolution of microstructure during thermal
activation of chromium-promoted tin(IV) oxide catalysts: an FT-IR, FT-Raman, XRD, TEM and
XANES/EXAFS study, Chem. Mater. 11 (1999) 896.
[104] P.G. Harrison, W. Azelee, A.T. Mubarak, C. Bailey, W. Daniell, N.C. Lloyd, in: N. Kruse, A. Frennet,
J.-M. Bastin (Eds.), Catalysis and Automotive Pollution Control. IV, Studies in Surface Science and
Catalysis, vol. 116, Elsevier, Amsterdam, 1988.
[105] G. Centi, F. Triro`, Oxidation catalysts based on antimony mixed oxides with rutile-type structures, Catal.
Rev. 28 (1986) 165.
[106] F.J. Berry, Tin-antimony oxide catalysts, Adv. Catal. 30 (1981) 97.
[107] H.J. Herniman, D.R. Pyke, R. Reid, An investigation of the relationship between the bulk and surface
composition of tin and antimony mixed oxide catalysts and the oxidative dehydrogenation of 1-butene to
butadiene, J. Catal. 58 (1979) 68.
[108] Y.M. Cross, D.R. Pyke, An X-ray photoelectron spectroscopy study of the surface composition of tin and
antimony mixed metal oxide catalysts, J. Catal. 58 (1979) 61.
[109] Y. Boudeville, F. Figueras, M. Forissier, J.L. Portefaix, J.C. Vedrine, Correlations between X-ray
photoelectron spectroscopy data and catalytic properties in selective oxidation on SbSnO catalysts,
J. Catal. 58 (1979) 52.
[110] F. Sala, F. Triro`, Oxidation catalysts based on tin-antimony oxides, J. Catal. 34 (1974) 68.
[111] N.G. Valente, L.E. Cadus, O.F. Gorriz, L.A. Arrua, J.B. Rivarola, Synergy in SnMoO catalysts: the
selective oxidation of methanol, Appl. Catal. A 153 (1997) 119.
[112] L.J. Lakshmi, E.C. Alyea, ESR, FT-Raman spectroscopic and ethanol partial oxidation studies on
MoO3/SnO2 catalysts made by metal oxide vapor synthesis, Catal. Lett. 59 (1999) 73.
[113] M. Niwa, H. Yamada, Y. Murakami, Activity for the oxidation of methanol of a molybdena monolayer
supported on tin oxide, J. Catal. 134 (1992) 331.
[114] T. Ono, H. Kamisuki, H. Hisashi, H. Miyata, A comparison of oxidation activities and structures of Mo
oxides highly dispersed on group IV oxide supports, J. Catal. 116 (1989) 303.
[115] M. Daturi, L.G. Appel, Infrared spectroscopic studies of surface properties of Mo/SnO2 catalyst, J. Catal.
209 (2002) 427.
[116] M. Niwa, M. Mizutani, M. Takahashi, Y. Murakami, Mechanism of methanol oxidation over oxide
catalysts containing MoO3, J. Catal. 70 (1981) 14.
[117] R. Skolmeistere, L. Leitis, M. Shymanska, J. Stoch, Inuence of preparation conditions on activity and
physical properties of V2O5SnO2 catalyst for oxidative destruction of residual pyridine bases, Catal. Today
17 (1993) 79.
[118] S.L.T. Andersson, S. Jaras, Activity measurements and ESCA investigations of a V2O5/SnO2 catalyst for
the vapor-phase oxidation of alkylpyridines, J. Catal. 64 (1980) 51.
[119] S. Bordoni, F. Catellani, F. Cavani, F. Triro`, M. Gazzano, Nature of vanadium species in SnO2V2O5based catalysts. Chemistry of preparation, characterization, thermal stability and reactivity in ethane
oxidative dehydrogenation over VSn mixed oxides, J. Chem. Soc. Faraday Trans. 90 (1994) 2981.
[120] P.G. Harrison, Tin(IV)oxide: surface chemistry, catalysis and gas sensing, in: P.G. Harrison (Ed.),
Chemistry of Tin, Blackie, Glasgow, 1989.
[121] P. Aigrain, C. Dugas, Adsorption sur les semi-conducteurs, Z. Elektrochem. 56 (1952) 363.
[122] P.B. Weisz, Electronic barrier layer phenomena in chemisorption and catalysis, J. Chem. Phys. 20 (1952)
1483.
[123] K. Haue, H.J. Engel, Zum Mechanismus der Chemisorption vom Standpunkt der Fehlordnungstheorie,
Z. Elektrochem. 56 (1952) 366.
[124] K. Haue, The application of the theory of semiconductors to problems of heterogenous catalysis,
Adv. Catal. 7 (1955) 213.
[125] Th. Wolkenstein, Sur les dierents types de liaisons lors de Ladsorption chimique sur des semi-conducteurs,
Adv. Catal. 9 (1957) 807.
[126] Th. Wolkenstein, Sur le mecanisme de laction catalytique des semi-conducteurs, Adv. Catal. 9 (1957) 818.
[127] J. Haber, M. Witko, Oxidation catalysis-electronic theory revisited, J. Catal. 216 (2003) 416.
[128] J. Haber, Molecular mechanism of heterogeneous oxidationorganic and solid state chemists views, Stud.
Surf. Sci. Catal. 110 (1997) 1.
[129] A. Bielanski, J. Deren, in: Electronic Phenomena in Chemisorption and Catalysis on Semiconductors,
DeGruyter, Berlin, 1969, p. 149.

144

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

[130] Chemical sensors, in: T. Seiyama, K. Fueki, J. Shiokawa, S. Suzuki (Jpn Eds.), Proceedings of the
International Meeting on Chemical Sensors, held in Fukuoka, Published by Kodansha, Tokyo and Elsevier,
Amsterdam, 1983.
[131] W. Gopel, Chemisorption and charge transfer at ionic semiconductor surfaces: implications in designing gas
sensors, Prog. Surf. Sci. 20 (1985) 9.
[132] M.J. Madou, S.R. Morrison, Chemical Sensing with Solid State Devices, Academic Press, Boston, 1989.
[133] F. Cosandey, G. Skandan, A. Singhal, Materials and processing issues in nanostructured semiconductor gas
sensors, JOM-e 52 (10) (2000).
[134] D.E. Williams, Modelling the response of semiconducting oxides as gas sensors, In: C.A. English,
J.R. Matthews, H. Rauh, A.M. Stonehead FRS, R. Thetford (Eds.), Materials Modelling: From Theory to
Technology, Inst. Phys. Publ., Bristol, 1991.
[135] K.D. Schierbaum, U. Weimar, W. Gopel, R. Kowalkowski, Conductance, work function and catalytic
activity of SnO2-based gas sensors, Sensors Actuat. B 3 (1991) 205.
[136] C. Malagu`, V. Guidi, M. Stefancich, M.C. Carotta, G. Mertinelli, Model for Schottky barrier and surface
states in nanostructured n-type semiconductors, J. Appl. Phys. 91 (2002) 808.
[137] A.M. Gaskov, M.N. Rumyantseva, Materials for solid-sate gas sensors, Inorg. Mater. 36 (2000) 369.
[138] V. Lantto, T.T. Rantala, T.S. Rantala, Atomistic understanding of semiconductor gas sensors, J. Eur.
Ceram. Soc. 21 (2001) 1961.
[139] N. Barsan, M. Schweizer-Berberich, W. Gopel, Fundamental and practical aspects in the design of
nanoscale SnO2 gas sensors: a status report, Fres. J. Anal. Chem. 365 (1999) 287.
[140] N. Barsan, U. Weimar, Conduction model of metal oxide gas sensors, J. Electroceram. 7 (2001) 143.
[141] J.W. Erickson, S. Semancik, An economical ultrahigh vacuum four-point resistivity probe, J. Vac. Sci.
Technol. A 5 (1987) 115.
[142] K.D. Schierbaum, U.K. Kirner, J.F. Geiger, W. Gopel, Schottky-barrier and conductivity gas sensors based
upon Pd/SnO2 and Pt/SnO2, Sensors Actuat. B 4 (1991) 87.
[143] S.R. Davis, A.V. Chadwick, J.D. Wright, A combined EXAFS and diraction study of pure and doped
nanocrystalline tin oxide, J. Phys. Chem. B 101 (1997) 9901.
[144] K.D. Schierbaum, X. Wei-Xing, W. Gopel, Solid/gas-interactions of surface-doped oxides: CV, IV, XPS,
UPS, and ELS studies on Pt/TiO2 and Pd/SnO2(1 1 0), Ber. Bunsenges. Phys. Chem. 97 (1993) 363.
[145] W. Gopel, New materials and transducers for chemical sensors, Sensors Actuat. B 18/19 (1994) 1.
[146] G. Zhang, M. Liu, Eect of particle size and dopant on properties of SnO2-based gas sensors, Sensors
Actuat. B 69 (2000) 144.
[147] J.F. Geiger, P. Beckmann, K.D. Schierbaum, W. Gopel, Interaction of Pd-overlayers with SnO2:
comperative comperative XPS, SIMS, and SNMS studies, Fres. J. Anal. Chem. 341 (1991) 25.
[148] N. Yamazoe, Y. Kurokawa, T. Seiyama, Eects of additives on semiconductor gas sensors, Sensors Actuat.
4 (1983) 283.
[149] K.M. Sancier, Hydrogen migration on alumina/palladium catalysts for benzene hydrogenation, J. Catal. 20
(1971) 106.
[150] G.E. Batley, A. Ekstrom, D.A. Johnson, Studies of topochemical heterogeneous catalysis. IV. Catalysis of
the reaction of zinc sulphide with oxygen, J. Catal. 36 (1975) 285.
[151] P.A Sermon, G.C. Bond, Hydrogen spillover, Catal. Rev. 8 (1973) 211.
[152] S.C. Tsang, C.D.A. Bulpitt, P.C.H. Mitchell, A.J. Ramirez-Cuesta, Some new insights into the sensing
mechanism of palladium promoted tin(IV) oxide sensor, J. Phys. Chem. B 105 (2001) 5737.
[153] C.-N. Xu, J. Tamaki, N. Miura, N. Yamazoe, Nature of sensitivity promotion in Pd-loaded SnO2 gas
sensor, J. Electrochem. Soc. 143 (1996) L148.
[154] N. Yamamoto, S. Tonomura, T. Matsuoka, H. Tsubomura, Eect of various substrates on the hydrogen
sensitivity of palladium-semiconductor diodes, J. Appl. Phys. 52 (1981) 6227.
[155] J. Tamaki, T. Maekawa, N. Miura, N. Yamazoe, CuOSnO2 element of highly sensitive and selective
detection of H2S, Sensors Actuat. B 9 (1992) 197.
[156] M.N. Rumyantseva, R.B. Vasiliev, I.B. Kutsenok, Thermodynamic approach to the development of H2S
selective SnO2 based gas sensors, Proceedings of the 11th European Conference on Solid State Transducers,
Eurosensors XI, Warsaw 1 (1997) 503.
[157] J. Liu, X. Huang, G. Ye, W. Lei, Z. Jiao, W. Chao, Z. Zhou, Z. Yu, H2S detection sensing characteristic of
CuO/SnO2 sensor, Sensors 3 (2003) 110.
[158] M.N. Boulova, T. Pagnier, M.N. Rumyantseva, In situ study of the reactivity of SnO2 nanocrystallites with
H2S by coupled electrical and Raman measurements, NATO Sci. Ser., Ser. 3: High Technol. 62 (1998) 285.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

145

[159] K. Suito, N. Kawai, Y. Masuda, High pressure synthesis of orthorhombic SnO2, Mater. Res. Bull. 10 (1975)
677.
[160] L. Luxmann, R. Dobner, Metall (Berlin) 34 (1980) 821.
[161] F. Lawson, Tin oxideSn3O4, Nature 215 (1967) 955.
[162] G. Murken, M. Tromel, Ueber das bei der Disproportionierung von SnO entstehende Zinnoxid, Sn2O3,
Z. Anorg. Allg. Chem. 897 (1973) 117.
[163] K. Hasselbach, G. Murken, M. Tromel, Mossbauer-Messungen an Sn2O3, Z. Anorg. Allg. Chem. 897
(1973) 127.
[164] Y. Li-Zi, S. Zhi-Tong, W. Chan-Zheng, A thermodynamic study of tin oxides by coulometric titration,
J. Solid State Chem. 113 (1994) 221.
[165] F.D. Rossini, D.D. Wagman, W.H. Evans, S. Levin, I. Jae, Selected Values of Chemical Thermodynamic
Properties, Circular of the National Bureau of Standards 500, U.S. Government Printing Oce,
Washington, DC, 1952.
[166] D.N. Klushin, O.V. Nadinskaya, K.G. Bogatina, Zh. Prikl. Khim. (Leningrad) 32 (1959) 273.
[167] J. Geurts, S. Rau, W. Richter, F.J. Schmitte, SnO lms and their oxidation to SnO2: Raman scattering, IR
reectivity and X-ray diraction studies, Thin Solid Films 121 (1984) 217.
[168] T. Yamazaki, U. Mizutani, Y. Iwama, Formation of vapour-deposited SnO2 thin lms studied by
Rutherford backscattering, Jpn. J. Appl. Phys. 21 (1982) 440.
[169] J. Pannetier, G. Denes, Tin(II) oxide: structure renement and thermal expansion, Acta Crystallogr., Sect.
B: Struct. Crystallogr. Cryst. Chem. 36 (1980) 2763.
[170] G.W. Watson, The origin of the electron distribution in SnO, J. Chem. Phys. 114 (2001) 758.
[171] J. Terra, D. Guenzburger, Electronic structure and electric-eld gradients of crystalline Sn(II) and Sn(IV)
compounds, Phys. Rev. B 44 (1991) 8584.
[172] M. Meyer, G. Onida, M. Palummo, L. Reining, Ab initio pseudopotential calculation of the equilibrium
structure of tin monoxide, Phys. Rev. B 64 (2001) 045119.
strom, Stereochimie des elements comportant des paires non liees:
[173] J. Galy, G. Meunier, S. Andersson, A. A
Ge(II), As(III), Se(IV), Br(V), Sn(II), Sb(III), Te(IV), I(V), Xe(VI), Tl(I), Pb(II), et Bi(III) (oxydes, uorures
et oxyuorures), J. Solid State Chem. 13 (1975) 142.
[174] R.J. Gillespie, Molecular Geometry, Van Nostrand Reinhold, London, 1972.
[175] I. Lefebvre, M. Szymanski, J. Olivier-Fourcade, J.C. Jumas, Electronic structure of tin monochalcogenides
from SnO to SnTe, Phys. Rev. B 58 (1998) 1896.
[176] A.A. Bolzan, C. Fong, B.J. Kennedy, C.J. Howard, Structural studies of rutile-type metal dioxides, Acta
Crystallogr. Sect. BStruct. Sci. 53 (1997) 373.
[177] L. Sangaletti, L.E. Depero, B. Allieri, F. Pioselli, E. Comini, G. Sberveglieri, M. Zocchi, Oxidation of Sn
thin lms to SnO2: micro-Raman mapping and X-ray diraction studies, J. Mater. Res. 13 (1998) 2457.
[178] K. Mcguire, Z.W. Pan, Z.L. Wang, D. Milkie, J. Menendez, A.M. Rao, Raman studies of semiconducting
oxide nanobelts, J. Nanosci. Nanotech. 2 (2002) 499.
[179] R.S. Katiyar, P. Dawson, M.M. Hargreave, G.R. Wilkinson, Dynamics of the rutile structure. III. Lattice
dynamics, infrared and Raman spectra of SnO2, J. Phys. C: Solid St. Phys. 4 (1971) 2421.
[180] A. Dieguez, A. Romano-Rodriguez, A. Vila, J.R. Morante, The complete Raman spectrum of nanometric
SnO2 particles, J. Appl. Phys. 90 (2001) 1550.
[181] P. Merle, J. Pascual, J. Camassel, H. Mathieu, Uniaxial-stress dependence of the rst-order Raman
spectrum of rutile. I. Experiments, Phys. Rev. B 21 (1980) 1617.
[182] T. Sato, T. Asari, Temperature dependence of the linewidth of the rst-order Raman spectrum for SnO2
crystal, J. Phys. Soc. Jpn. 64 (1995) 1193.
[183] P.S. Peercy, B. Morosin, Pressure and temperature dependence of Raman-active phonons in SnO2, Phys.
Rev. B 7 (1972) 2779.
[184] R. Summitt, Infrared absorption in single-crystal stannic oxide: optical lattice-vibration modes, J. Appl.
Phys. 39 (1968) 3762.
[185] P.S. Peercy, B. Morosin, Pressure and temperature dependences of the Raman-active phonons in SnO2,
Phys. Rev. B 7 (1973) 2779.
[186] R. Fuchs, K.L. Kliewer, Optical modes of vibration in an ionic crystal slab, Phys. Rev. 140 (1965) A2076.
[187] P.A. Cox, R.G. Egdell, W.R. Flavell, R. Helbig, Observation of surface optical phonons on SnO2(1 1 0),
Vacuum 33 (1983) 835.
[188] M. Batzill, J. Kim, D.E. Beck, B.E. Koel, Epitaxial growth of tin-oxide on Pt(1 1 1): structure and properties
of wetting layers and SnO2 crystallites, Phys. Rev. B 69 (2004) 165403.

146

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

[189] P.A. Cox, Rg. Egdell, C. Harding, W.R. Patterson, P.J. Taverner, Surface properties of antimony doped
tin(IV) oxide: a study by electron spectroscopy, Surf. Sci. 123 (1982) 179.
[190] V.M. Jimenez, J.A. Mejias, J.P. Espinos, A.R. Gonzalez-Elipe, Interface eects for metal oxide thin lms
deposited on another metal oxide. II. SnO2 deposited on SiO2, Surf. Sci. 366 (1996) 545.
[191] L. Kovar, Z. Kovacs, R. Sanjines, G. Moretti, G. Margaritondo, J. Palinkas, H. Adachi, Surf. Interface
Anal. 23 (1995) 467.
[192] V.M. Jimenez, J.P. Espinos, A.R. Gonzalez-Elipe, Interface eects for metal oxide thin lms deposited on
another metal oxide. III. SnO and SnO2 deposited on MgO(1 0 0) and the use of chemical state plots, Surf.
Sci. 366 (1996) 556.
[193] A.F. Lee, R.M. Lambert, Oxidation of Sn overlayers and the structure and stability of Sn oxide lms on
Pd(1 1 1), Phys. Rev. B 58 (1998) 4156.
[194] J.-M. Themlin, M. Chtaib, L. Henrard, P. Lambin, J. Darville, J.-M. Gilles, Characterization of tin oxides
by X-ray-photoemission spectroscopy, Phys. Rev. B 46 (1992) 2460.
[195] D.A. Ashbury, G. Hound, A surface study of the oxidation of polycrystalline tin, J. Vac. Sci. Technol. A 5
(1987) 1132.
[196] C.D. Wagner, in: D. Briggs, M.P. Seah (Eds.), Practical Surface Analysis, vol. 1, Wiley, New York, 1983, p.
648.
[197] A. Beltran, J. Andres, E. Longo, E.R. Leite, Thermodynamic argument about SnO2 nanoribbon growth,
Appl. Phys. Lett. 83 (2003) 635.
[198] J. Oviedo, M.J. Gillan, Energetics and structure of stoichiometric SnO2 surfaces studied by rst-principles
calculations, Surf. Sci. 463 (2000) 93.
[199] P.A. Mulheran, J.H. Harding, The stability of SnO2 surfaces, Modell. Simul. Mater. Sci. Eng. 1 (1992) 39.
[200] B. Slater, C.R. Catlow, D.H. Gay, D.E. Williams, V. Dusastre, Study of surface segregation of antimony on
SnO2 surfaces by computer simulation techniques, J. Phys. Chem. B 103 (1999) 10644.
[201] W. Bergermayer, I. Tanaka, Reduced SnO2 surfaces by rst-principles calculations, Appl. Phys. Lett. 84
(2004) 909.
[202] M. Batzill, A.M. Chaka, U. Diebold, Oxygen chemistry of a gas sensing material: SnO2 (1 0 1), Europhys.
Lett. 65 (2004) 61.
[203] E. de Fresart, J. Darville, J.M. Gilles, Inuence of the surface reconstruction on the work function and
surface conductance of (1 1 0)SnO2, Appl. Surf. Sci. 11/12 (1982) 637.
[204] C.L. Pang, S.A. Haycock, H. Raza, P.J. Mller, G. Thornton, Structures of the 4 1 and 1 2
reconstructions of SnO2(1 1 0), Phys. Rev. B 62 (2000) R7775.
[205] D.F. Cox, T.B. Fryberger, S. Semancik, Surface reconstructions of oxygen decient SnO2(1 1 0), Surf. Sci.
224 (1989) 121.
[206] F.H. Jones, R. Dixon, J.S. Foord, R.G. Egdell, J.B. Pethica, The surface structure of SnO2(1 1 0) (4 1)
revealed by scanning tunneling microscopy, Surf. Sci. 376 (1997) 367.
[207] M. Sinner-Hettenbach, M. Gothelid, J. Weissenrieder, H. von Schenck, T. Wei, N. Barsan, U. Weimar,
Oxygen-decient SnO2(1 1 0): a STM, LEED and XPS study, Surf. Sci. 477 (2001) 50.
[208] A. Atrei, E. Zanazzi, U. Bardi, G. Rovida, The SnO2(1 1 0)(4 1) structure determined by LEED intensity
analysis, Surf. Sci. 475 (2001) L223.
[209] M. Batzill, K. Katsiev, U. Diebold, Surface morphologies of SnO2(1 1 0), Surf. Sci. 529 (2003) 295.
[210] D.F. Cox, T.B. Fryberger, S. Semancik, Oxygen vacancies and defect electronic states on the SnO2(1 1 0)1 1 surface, Phys. Rev. B 38 (1988) 2072.
[211] Z.S. Li, Q.L. Guo, P.J. Mller, Electronic properties of Cu clusters and islands and their reaction with O2
on SnO2(1 1 0) surfaces, Z. Physik D 40 (1997) 550.
[212] V.A. Gercher, D.F. Cox, J.-M. Themlin, Oxygen-vacancy-controlled chemistry on a metal-oxide surfacemethanol dissociation and oxidation on SnO2(1 1 0), Surf. Sci. 306 (1994) 279.
[213] V.A. Gercher, D.F. Cox, Water-adsorption on stoichiometric and oxygen decient SnO2(1 1 0) surfaces,
Surf. Sci. 332 (1995) 177.
[214] R. Cavicchi, V. Sukkarev, S. Semancik, Preparation of well-oredered, oxygen-rich SnO2(1 1 0) surfaces via
oxygen plasma treatment, J. Vac. Sci. Technol. A 8 (1990) 2347.
[215] J. Erickson, S. Semancik, Surface conductivity changes in SnO2(1 1 0)-eects of oxygen, Surf. Sci. 187 (1987)
L658.
[216] J.-M. Themlin, J.-M. Gilles, R.L. Johnson, Oxygen 2s spectroscopy of tin oxides with synchrotron
radiation-induced photoemission, J. Physique IV 4 (1994) C9183.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

147

[217] G. Margaritondo, A. Franciosi, Synchrotron radiation photoemission spectroscopy of semiconductor


surfaces and interfaces, Ann. Rev. Mater. Sci. 14 (1984) 67.
[218] J.J. Yeh, I. Lindau, Atomic data Nucl. Data Tables 32 (1985) 1.
[219] C. Stamp, M.V. Ganduglia-Pirovano, K. Reuter, M. Scheer, Catalysis and corrosion: the theoretical
surface-science context, Surf. Sci. 500 (2002) 368.
[220] K. Reuter, M. Scheer, Composition, structure, and stability of RuO2(1 1 0) as a function of oxygen
pressure, Phys. Rev. B 65 (2002) 035406.
[221] X.G. Wang, A. Chaka, M. Scheer, Eect of the Environment on a-Al2O3(0 0 0 1) Surface Structures, Phys.
Rev. Lett. 84 (2000) 3650.
[222] Q. Sun, K. Reuter, M. Scheer, Eect of a humid environment on the surface structure of RuO2(1 1 0),
Phys. Rev. B 67 (2003) 205424.
[223] K. Reuter, M. Scheer, First-principles atomistic thermodynamics for oxidation catalysis: surface phase
diagrams and catalytically interesting regions, Phys. Rev. Lett. 90 (2003) 046103.
[224] K. Reuter, M. Scheer, Composition and structure of the RuO2(1 1 0) surface in an O2 and CO
environment: implications for the catalytic formation of CO2, Phys. Rev. B 68 (2003) 045407.
[225] X.-G. Wang, W. Weiss, Sh.K. Shaikhutdinov, M. Ritter, M. Petersen, F. Wagner, R. Schlogel, M. Scheer,
The hematite (a-Fe2O3)(0 0 0 1) surface: evidence for domains of distinct chemistry, Phys. Rev. Lett. 81
(1998) 1038.
[226] O. Dulub, U. Diebold, G. Kresse, Novel stabilization mechanism on polar surfaces: ZnO(0 0 0 1)Zn, Phys.
Rev. Lett. 90 (2003) 016102.
[227] G. Kresse, O. Dulub, U. Diebold, Competing stabilization mechanism for the polar ZnO(0 0 0 1)Zn
surface, Phys. Rev. B 68 (2003) 245409.
[228] M. Batzill, K. Katsiev, J.M. Burst, U. Diebold, A.M. Chaka, B. Delley, Gas phase-dependent properties of
SnO2(1 1 0), (1 0 0), and (1 0 1) single crystal surfaces: structure, composition, and electronic properties, Phys.
Rev. B, in press.
[229] J. Oviedo, M.J. Gillan, The energetics and structure of oxygen vacancies on the SnO2(1 1 0) surface, Surf.
Sci. 467 (2000) 35.
[230] P.W. Murray, N.G. Condon, G. Thornton, Eect of stoichiometry on the structure of TiO2(1 1 0), Phys.
Rev. B 51 (1995) 10989.
[231] H. Onishi, Y. Iwasawa, Dynamic visualization of a metal-oxide-surface/gas-phase reaction: time-resolved
observation by scanning tunneling microscopy at 800 K, Phys. Rev. Lett. 76 (1996) 791.
[232] I. Kocemba, S. Szafran, J. Rynkowski, T. Paryjczak, Relationship between the catalytic and detection
properties of SnO2 and Pt/SnO2 systems, Ads. Sci. Technol. 20 (2002) 897.
[233] C.L. Lau, G.K. Wertheim, Oxidation of tin: an ESCA study, J. Vac. Sci. Technol. 15 (1978) 622.
[234] V.M. Jimenez, G. Lassaletta, A. Fernandez, J.P. Espinos, F. Yubero, A.R. Gonzalez-Elipe, L. Soriano,
J.M. Sanz, D.A. Papaconstantopoulos, Resonant photoemission characterization of SnO, Phys. Rev. B 60
(1999) 11171.
[235] S. Munnix, M. Schmeits, Electronic structure of tin dioxide surfaces, Phys. Rev. B 27 (1983) 7624.
[236] J.M. Themlin, R. Sporken, J. Darville, R. Caudano, J.M. Gilles, Resonant-photoemission study of SnO2:
cationic origin of the defect band-gap states, Phys. Rev. B 42 (1990) 11914.
[237] M. Sinner-Hettenbach, M. Gothlid, T. Weiss, N. Barsan, U. Weimar, H. von Schemck, L. Giovanelli, G. Le
Lay, Electronic structure of SnO2(1 1 0)-4 1 and sputtered SnO2(1 1 0) revealed by resonant photoemission,
Surf. Sci. 499 (2002) 85.
[238] R.G. Egdell, S. Erickson, W.R. Flavell, Oxygen decient SnO2(1 1 0) and TiO2(1 1 0)-a comperative-study by
photoemission, Solid State Commun. 60 (1986) 835.
[239] R.G. Egdell, J. Rebane, T.J. Walker, D.S.L. Law, Competition between initial- and nal-state eects in
valence- and core-level X-ray photoemission of Sb-doped SnO2, Phys. Rev. B 59 (1999) 1792.
[240] D.F. Cox, S. Semancik, P.D. Szuromi, Structural and electronic-properties of clean and water dosed
SnO2(1 1 0), J. Vac. Sci. Technol. A 4 (1986) 627.
[241] D.F. Cox, T.B. Fryberger, J.W. Erickson, S. Semancik, Surface-properties of clean and gas-dosed
SnO2(1 1 0), J. Vac. Sci. Technol. A 5 (1988) 1170.
[242] M.A. Maki-Jaskari, T.T. Rantala, Band structure and optical parameters of the SnO2(1 1 0) surface, Phys.
Rev. B 64 (2001) 075407.
[243] M.A. Maki-Jaskari, T.T. Rantala, Theoretical study of oxygen-decient SnO2(1 1 0) surfaces, Phys. Rev. B
65 (2002) 245428.

148

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

[244] I. Manassidis, J. Goniakowski, L.N. Kantorovich, M.J. Gillan, The structure of the stoichiometric and
reduced SnO2(1 1 0) surface, Surf. Sci. 339 (1995) 258.
[245] Ph. Ebert, Nano-scale properties of defects in compound semiconductor surfaces, Surf. Sci. Rep. 33 (1999) 121.
[246] M. Batzill, B. Katsiev, D.J. Gaspar, U. Diebold, Variations of the local electronic surface properties of
TiO2(1 1 0) induced by intrinsic and extrinsic defects, Phys. Rev. B 66 (2002) 235401.
[247] M. Batzill, E.L.D. Hebenstreit, W. Hebenstreit, U. Diebold, The inuence of subsurface, charged impurities
on the adsorption of chlorine at TiO2(1 1 0), Chem. Phys. Lett. 367 (2003) 319.
[248] S. Munnix, M. Schmeits, Electronic structure and point defects on oxide surfaces, Phys. Rev. B 33 (1986)
4136.
[249] T. Suzuki, H. Wakabayashi, Y. Nishi, M. Fujimoto, Transmission electron microscopy study of defect
structure in epitaxial SnO2 rutile thin lm, J. Ceram. Soc. Jpn. 110 (2002) 86.
[250] H. Wakabayashi, T. Suzuki, Y. Iwazaki, M. Fujimoto, Defect structure of heteroepitaxial SnO2thin lms
grown on TiO2substrates, Jpn. J. Appl. Phys. 40 (Pt. 1) (2001) 6081.
[251] J.E. Dominguez, L. Fu, X.Q. Pan, Eect of crystal defects on the electrical properties in epitaxial tin dioxide
thin lms, Appl. Phys. Lett. 81 (2002) 5168.
[252] J.E. Dominguez, X.Q. Pan, L. Fu, P.A. Van Rompay, Z. Zhang, J.A. Nees, P.P. Pronko, Epitaxial SnO2
thin lms grown on (1 0 1 2) sapphire by femtosecond pulsed laser deposition, J. Appl. Phys. 91 (2002)
1060.
[253] R.E. Cavicchi, S. Semancik, M.D. Antonik, R.J. Lad, Layer-by-layer growth of epitaxial SnO2 on sapphire
by reactive sputter deposition, Appl. Phys. Lett. 61 (1992) 1921.
[254] R. Cavicchi, S. Semancik, The growth of thin, epitaxial SnO2 lms for gas sensing applications, Thin Solid
Films 206 (1991) 81.
[255] A. Tarre, A. Rosental, J. Sundqvist, A. Harsta, T. Uustare, V. Sammelselg, Nanoepitaxy of SnO2 on
a-Al2O3(0 1 2), Surf. Sci. 532535 (2003) 514.
[256] J. Lu, J. Sundqvist, M. Ottosson, A. Tarre, A. Rosental, J. Aarik, A. Harsta, Microstructure
characterisation of ALD-grown epitaxial SnO2 thin lms, J. Cryst. Growth 260 (2004) 191.
[257] M. Nagano, Chemical vapor deposition of SnO2 thin lms on (1 0 0) surfaces of rutile single crystals,
J. Cryst. Growth 69 (1984) 465.
[258] M. Nagano, Chemical vapor deposition of SnO2 thin lms on rutile single crystals, J. Cryst. Growth 67
(1984) 639.
[259] V. Caslavska, R. Roy, Epitaxial growth of SnO2on rutile single crystals, J. Appl. Phys. 40 (1969) 3414.
[260] A. Tarre, A. Rosental, A. Aidla, J. Aarik, J. Sundqvist, A. Harsta, New routes to SnO2 heteroepitaxy,
Vacuum 67 (2002) 571.
[261] J. Sundqvist, A. Tarre, A. Rosental, A. Harsta, Atomic layer deposition of epitaxial and polycrystalline
SnO2 lms from the SnI4/O2 precursor combination, Chem. Vapor Deposition 9 (2003) 21.
[262] M. Batzill, J.M. Burst, U. Diebold, Pure and Co-doped SnO2(1 0 1) lms grown by MBE on Al2O3, Thin
Solid Films 484 (2005) 132.
[263] A. Rosental, A. Tarre, A. Gerst, J. Sundqvist, A. Harsta, A. Aidla, J. Aarik, V. Sammelselg, T. Uustare,
Gas sensing properties of epitaxial SnO2 thin lms prepared by atomic layer deposition, Sensors Actuat. B
93 (2003) 552.
[264] J. Vetrone, Y.-W. Chung, R. Cavicchi, S. Semancik, Role of initial conductance and gas pressure on the
conductance response of single-crystal SnO2 thin lms to H2, O2, and CO, J. Appl. Phys. 73 (1993) 8371.
[265] J.R. Sambrano, L.A. Vasconcellos, J.B.L. Martins, M.R.C. Santos, E. Longo, A. Beltran, A theoretical
analysis on electronic structure of the (1 1 0) surface of TiO2SnO2 mixed oxide, J. Mol. Struct. (Theochem.)
629 (2003) 307.
[266] F.R. Sensato, R. Custodio, E. Longo, A. Beltran, J. Andres, Electronic and structural properties of
SnxTi1xO2 solid solutions: a periodic DFT study, Catal. Today 85 (2003) 145.
[267] K. Zakrzewska, Mixed oxides as gas sensors, Thin Solid Films 391 (2001) 229.
[268] R.J. Choudhary, S.B. Ogale, S.R. Shinde, V.N. Kulkarni, T. Venkatesan, K.S. Harshavardhan,
M. Strikovski, B. Hannoyer, Pulsed-electron-beam deposition of transparent conducting SnO2 lms and
study of their properties, Appl. Phys. Lett. 84 (2004) 1483.
[269] G.E. Poirier, R.E. Cavicchi, S. Semancik, Paricle growth of palladium on epitaxial tin oxide lms, J. Vac.
Sci. Technol. A 12 (1994) 2149.
[270] M. Hoheisel, S. Speller, W. Heiland, A. Atrei, U. Bardi, G. Rovida, Adsorption of oxygen on Pt3Sn(1 1 1)
studied by scanning tunneling microscopy and x-ray photoelectron diraction, Phys. Rev. B 66 (2002)
165416.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

149

[271] A. Atrei, U. Bardi, G. Rovida, M. Torrini, M. Hoheisel, S. Speller, Test of structural models for the (4 4)
phase formed by oxygen adsorption on the Pt3Sn(1 1 1) surface, Surf. Sci. 526 (2003) 193.
[272] M. Batzill, D. Beck, B.E. Koel, Structure of monolayer tin oxide lms on Pt(1 1 1) formed using NO2 as an
ecient oxidant, Phys. Rev. B 64 (2001) 245402.
[273] M. Batzill, D.E. Beck, B.E. Koel, Self-organized molecular-sized, hexagonally ordered SnOx nanodot
superlattices on Pt(1 1 1), Appl. Phys. Lett. 78 (2001) 2766.
[274] M. Batzill, D.E. Beck, D. Jerdev, B.E. Koel, Tin-oxide overlayer formation by oxidation of PtSn(1 1 1)
surface alloys, J. Vac. Sci. Technol. A 19 (2001) 1953.
[275] W. Hellmich, G. Muller, Ch. Bosch-von Braunmuhl, T. Doll, I. Eisele, Field-eect-induced gas sensitivity
changes in metal oxides, Sensors Actuat. B 43 (1997) 132.
[276] J. Dalin, Fabrication and characterization of a novel MOSFET gas sensor, PhD thesis, Linkopings Institute
of Technology.
[277] Y. Zhang, A. Kolmakov, S. Chretien, H. Metiu, M. Moskovits, Control of catalytic reactions at the surface
of a metal oxide nanowire by manipulating electron density inside it, Nano Letters 4 (2004) 403.
[278] G.E. Poirier, R.E. Cavicchi, S. Semancik, Particle growth of palladium on epitaxial tin oxide thin lms,
J. Vac. Sci. Technol. A 12 (1994) 2149.
[279] R. Cavicchi, S. Semancik, Reactivity of Pd and Sn adsorbates on plasma and thermally oxidized
SnO2(1 1 0), Surf. Sci. 257 (1991) 70.
[280] J.W. Erickson, T.B. Fryberger, S. Semancik, Metal/semiconductor interfaces on SnO2(1 1 0), J. Vac. Sci.
Technol. A 6 (1988) 1593.
[281] R. Cavicchi, V. Sukharev, S. Semancik, Time-dependent conductance of Pd-dosed SnO2(1 1 0), Surf. Sci.
418 (1998) L81.
[282] T.B. Freyberger, J.W. Erickson, S. Semancik, Chemical and electronic properties of Pd/SnO2(1 1 0) model
gas sensors, Surf. Interf. Anal. 14 (1989) 83.
[283] J. Arbiol, A. Cirera, F. Peiro, A. Cornet, J.R. Morante, J.J. Delgado, J.J. Calvino, Optimization of tin
dioxide nanosticks faceting for the improvement of palladium nanocluster epitaxy, Appl. Phys. Lett. 80
(2002) 329.
[284] J. Arbiol, R. Diaz, A. Ciera, F. Peiro, A. Cornet, J.R. Morante, F. Sanz, C. Mira, J.J. Delgado, G. Blanco,
J.J. Calvino, Eects of in-situ and ex-situ reduction of Pd/SnO2 studied by HRTEM, Inst. Phys. Conf. Ser.
169 (2001) 73.
[285] M.A. Maki-Jaskari, T.T. Rantala, Density functional study of Pd adsorbates at SnO2(1 1 0) surfaces, Surf.
Sci. 537 (2003) 168.
[286] P. De Padova, R. Larciprete, M. mangiantini, M. Fanfoni, Cr, Sn, and Ag/SnO2 interface formation
studied by synchrotron radiation induced UPS, J. Electr. Spec. Rel. Phenom. 76 (1995) 499.
[287] Z. Li, Q. Guo, P.J. Mller, Electronic properties of Cu clusters and islands and their reaction with O2 on
SnO2(1 1 0) surfaces, Z. Phys. D 40 (1997) 550.
[288] Y. Yamaguchi, K. Tabata, E. Suzuki, Density functional theory calculations for the interaction of oxygen
with reduced M/SnO2(1 1 0) (M = Pd, Pt) surfaces, Surf. Sci. 526 (2003) 149.
[289] A. Atrei, U. Bardi, C. Tarducci, G. Rovida, Growth, composition, and structure of ultrathin vanadium
lms deposited on the SnO2(1 1 0) surface, J. Phys. Chem. B 104 (2000) 3121.
[290] A. Atrei, T. Cecconi, B. Cortigiani, U. Bardi, M. Torrini, G. Rovida, Composition and structure of
ultrathin vanadium oxide layers deposited on SnO2(1 1 0), Surf. Sci. 513 (2002) 149.
[291] P. Lazcano, M. Batzill, U. Diebold, in preparation.
[292] E. Zampiceni, E. Bontempi, G. Sberveglieri, L.E. Depero, Mo inuence on SnO2 thin lm properties, Thin
Solid Films 418 (2002) 16.
[293] Ph. de Montgoler, P. Meriaudeau, Y. Boudeville, M. Che, Hyperne and superhyperne interactions of
Mo5+ in SnO2, Phys. Rev. B 14 (1976) 1788.
[294] B. Stjerna, E. Olsson, C.G. Granqvist, Optical and electrical properties of radio frequency sputtered tin
oxide lms doped with oxygen vacancies, F, Sb, or Mo, J. Appl. Phys. 76 (1994) 3797.
[295] A.E. Taverner, C. Rayden, S. Warren, A. Gulino, P.A. Cox, R.G. Egdell, Comparison of energies of
vanadium donor levels in doped SnO2 and TiO2, Phys. Rev. B 51 (1995) 6833.
[296] R.G. Egdell, A. Gulino, C. Rayden, G. Peacock, P.A. Cox, Nature of donor states in V-doped SnO2,
J. Mater. Chem. 5 (1995) 499.
[297] I.K. Ball, P.G. Harrison, C. Bailey, F.J. Allison, D. Eyre, The role of vanadium in vanadium promoted
tin(IV) oxide catalysis: an FTIR, XRD and EXAFS study, Main Group Metal Chem. 22 (1999) 617.

150

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

[298] M.-M. Bagheri-Mohagheghi, M. Shokooh-Saremi, The inuence of Al doping on the electrical, optical and
structural properties of SnO2 transparent conducting lms deposited by the spray pyrolysis technique,
J. Phys. D: Appl. Phys. 37 (2004) 1248.
[299] L. Poupon, P. Iacconi, C. Pijolat, Physical and quasi-chemical study of point defects in aluminum or
niobium-doped polycrystalline tin oxide, J. Eur. Ceram. Soc. 19 (1999) 747.
[300] S.W. Lee, Y.-W. Kim, H. Chen, Electrical properties of Ta-doped SnO2 thin lms prepared by the metalorganic chemical-vapor deposition methode, Appl. Phys. Lett. 78 (2001) 350.
[301] S.W. Lee, A. Daga, Z.K. Xu, H. Chen, Characterization of MOCVD grown optical coatings of Sc2O3 and
Ta-doped SnO2, Mater. Sci. Eng. B 99 (2003) 134.
[302] J.-F. Wang, H.-C. Chen, W.-X. Wang, W.-B. Su, G.-Z. Zang, Electrical nonlinearity of (Ni, Ta) doped
SnO2 varistors, Mater. Sci. Eng. B 99 (2003) 465.
[303] Y.J. Wang, J.F. Wang, C.P. Li, H.C. Chen, W.B. Su, W.L. Zhong, P.L. Zhang, L.Y. Zhao, Eects of
niobium dopant on the electrical properties of SnO2-based varistor system, J. Mater. Sci. Lett. 20 (2001) 19.
[304] W.-X. Wang, J.-F. Wang, H.-C. Chen, W.-B. Su, G.Z. Zang, Electrical nonlinearity of (Cu, Ni, Nb)-doped
SnO2 varistors system, Mater. Sci. Eng. B 99 (2003) 457.
[305] C.S. Rastomjee, R.S. Schaer, F.H. Jones, R.G. Egdell, G.C. Georgiadis, M.J. Lee, T.J. Tate, L.L. Cao, An
investigation of doping of SnO2 by ion implantation and application of ion-implanted lms as gas sensors,
Thin Solid Films 279 (1996) 98.
[306] D. Szczuko, J. Werner, S. Oswald, G. Behr, K. Wetzig, XPS investigations of surface segregation of doping
elements in SnO2, Appl. Surf. Sci. 179 (2001) 301.
[307] Z. Ji, Z. He, Y. Song, K. Liu, Z.Z. Ye, Fabrication and characterization of indium-doped p-type SnO2 thin
lms, J. Cryst. Growth 259 (2003) 282.
[308] C. Nozaki, K. Tabata, E. Suzuki, Synthesis and characterization of homogeneous germanium-substituted
tin oxide by solgel method, J. Solid State Chem. 154 (2000) 579.
[309] B.S. Niranjan, S.R. Sainkar, K. Vijayamohanan, I.S. Mulla, Ruthenium: tin oxide thin lm as a highly
selective hydrocarbon sensor, Sensors Actuat. B 82 (2002) 82.
[310] H. Hikita, K. Takeda, Y. Kimura, Analytical determination of the local oxygen structure around Cr3+ in
SnO2 rutile-type crystals by use of electron paramagnetic resonance, Phys. Rev. B 46 (1992) 14381.
[311] E. Lopez-Navarette, A. Caballero, V.M. Orera, F.J. Lazaro, M. Ocana, Oxidation state and localization of
chromium ions in Cr-doped cassiterite and Cr-doped malayaite, Acta Mater. 51 (2003) 2371.
[312] E. Lopez-Navarette, A.R. Gonzalez-Elipe, M. Ocana, Non-conventional synthesis of Cr-doped SnO2
pigments, Ceram. Int. 29 (2003) 385.
[313] A.V. Chadwick, EXAFS studies of dopant sites in metal oxides, Solid State Ion. 6365 (1993) 721.
[314] M.-M. Bagheri-Mohagheghi, M. Shokooh-Saremi Electrical, optical and structural properties of Li-doped
SnO2 trasnparent conducting lms deposited by spray pyrolysis technique: a carrier-type conversion study,
Semicond. Sci. Technol. 19 (2004) 764.
[315] M.C. Esteves, D. Gouvea, P.T.A. Sumodjo, Eect of uorine doping on the properties of tin oxide based
powders prepared via Pechinis method, Appl. Surf. Sci. 229 (2004) 24.
[316] I.T. Weber, A.P. Maciel, P.N. Lisboa-Filho, E. Longo, E.R. Leite, C.O. Paiva-Santos, Y. Maniette,
W.H. Schreiner, Eects of synthesis and processing on supersaturated rare earth-doped nanometric SnO2
powders, Nano Lett. 2 (2002) 970.
[317] J.M.D. Coey, A.P. Douvalis, C.B. Fitzgerald, M. Venkatesan, Ferromagnetism in Fe-doped SnO2 thin
lms, Appl. Phys. Lett. 84 (2004) 1332.
[318] S.B. Oagle, R.J. Choudhary, J.P. Buban, S.E. Loand, S.R. shinde, S.N. Kale, V.N. Kulkarni, J. Higgins,
C. Lanci, J.R. Simpson, N.D. Browning, S. Das Sarma, H.D. Drew, R.L. Greene, T. Venkatesan, High
temperature ferromagnetism with giant magnetic moment in transparent Co-doped SnO2d, Phys. Rev.Lett.
91 (2003) 077205.
[319] G. Jain, R. Kumar, Electrical and optical properties of tin oxide and antimony doped tin oxide lms,
Opt. Mater. 26 (2004) 27.
[320] J. Rockenberger, U. zum Felde, M. Tischer, L. Troger, M. Haase, H. Weller, Near edge X-ray absorption
ne structure measurements (XANES) and extended X-ray absorption ne structure measurements
(EXAFS) of the valence state and coordination of antimony in doped nanocrystalline SnO2, J. Chem. Phys.
112 (2000) 4296.
[321] V. Dusastre, D.E. Williams, Sb(III) as active sites for water adsorption on Sn(Sb)O2, and its eect on
catalytic activity and sensor behavior, J. Phys. Chem. B 102 (1998) 6732.

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

151

[322] R.G. Egdell, W.R. Flavell, P. Tavener, Antimony-doped tin(IV) oxide: surface composition and electronic
structure, J. Solid State Chem. 51 (1984) 345.
[323] C. McGinley, S. Al Moussalami, M. Riedler, M. Pughoet, H. Borchert, M. Haase, A.R.B. de Castro,
H. Weller, T. Moller, Pure and Sb-doped SnO2 nanoparticles studied by photoelectron spectroscopy,
Eur. Phys. J. D 16 (2001) 225.
[324] C. McGinley, H. Borchert, M. Pughoet, S. Al Moussalami, A.R.B. de Castro, M. Haase, H. Weller,
T. Moller, Dopant atom distribution and spatial connement of conduction electrons in Sb-doped SnO2
nanoparticles, Phys. Rev. B 64 (2001) 245312.
[325] P.A. Cox, R.G. Egdell, C. Harding, W.R. Patterson, P.J. Tavener, Surface properties of antimony doped
tin(IV) oxide a study by electron spectroscopy, Surf. Sci. 123 (1982) 179.
[326] C.S. Rastomjee, R.G. Egdell, M.J. Lee, T.J. Tate, Observation of conduction electrons in Sb-implanted
SnO2 by ultraviolet photoemission spectroscopy, Surf. Sci. Lett. 259 (1991) L769.
[327] R.G. Egdell, T.J. Walker, G. Beamson, The screening response of a dilute electron gas in core level
photoemission form Sb-doped SnO2, J. Electr. Spectr. Rel. Phenom. 128 (2003) 59.
[328] J.M. Coey, M. Venkatesan, C.B. Fitzgerald, Donor impurity band exchange in dilute ferromagnetic oxides,
Nature Mater. 4 (2005) 173.
[329] A. Lussier, J. Dvorak, Y.U. Idzerda, S.B. Oagle, S.R. Shinde, R.J. Choudary, T. Venkatesan, Comparative
X-ray absorption spectroscopy study of Co-doped SnO2 and TiO2, J. Appl. Phys. 95 (2004) 7190.
[330] M.A. Barteau, Organic reactions at well-dened oxide surfaces, Chem. Rev. 96 (1996) 1413.
[331] J.H.C. van Hoo, J.F. van Helden, Formation of peroxo radicals on tin dioxide, J. Catal. 8 (1967) 199.
[332] J.H.C. van Hoo, Formation of paramagnetic surface species during the oxidation of nonstoichiometric
TiO2(A), SnO2, and ZnO, J. Catal. 11 (1968) 277.
[333] P. Meriaudeau, C. Naccache, A.J. Tench, Paramagentic oxygen species adsorbed on reduced SnO2, J. Catal.
21 (1971) 208.
[334] C. Hauser, Considerations on oxygen paramagnetic centers adsorbed on TiO2 and SnO2, Chem. Phys. Lett.
18 (1973) 205.
[335] Y. Mizokawa, S. Nakamura, ESR study of adsorbed oxygen on tin dioxide, Oyo Buturi 46 (1977) 580 (in
Japanese).
[336] S.C. Chang, Oxygen chemisorption on tin oxide: correlation between electrical conductivity and EPR
measurements, J. Vac. Sci. Technol. 17 (1980) 366.
[337] N. Yamazoe, J. Fuchigami, M. Kishikawa, T. Seiyama, Interactions of tin oxide with O2, H2O and H2,
Surf. Sci. 86 (1979) 335.
[338] G.L. Shen, R. Casanova, G. Thornton, Interaction of O2 with SnO2(1 1 0)1 1 and 4 1, Vacuum 43 (1992)
1129.
[339] T.G.G. Maes, G.T. Owen, M.W. Penny, T.K.H. Starke, S.A. Clark, H. Ferkel, S.P. Wilks, Nanocrystalline SnO2 gas sensor response to O2 and CH4 at elevated temperature investigated by XPS, Surf. Sci.
520 (2002) 29.
[340] J. Mizsei, V. Lantto, Simultaneous response of work function and resistivity of some SnO2-based samples to
H2 and H2S, Sensors Actuat. B 4 (1991) 163.
[341] S. Semancik, D.F. Cox, Fundamental characterization of clean and gas-dosed tin oxide, Sensors Actuat. 12
(1987) 101.
[342] Y. Nagasawa, T. Choso, T. Karasuda, S. Shimomura, F. Ouyang, K. Tabata, Y. Yamaguchi,
Photoemission study of the interaction of a reduced thin lm SnO2 with oxygen, Surf. Sci. 433435
(1999) 226.
[343] T. Kawabe, S. Shimomura, T. Karasuda, K. Tabata, E. Suzuki, Y. Yamaguchi, Photoemission study of
dissociatively adsorbed methane on a pre-oxidized SnO2 thin lm, Surf. Sci. 448 (2000) 101.
[344] K.D. Schierbaum, H.D. Wiemhofer, W. Gopel, Defect structures and sensing mechansism of SnO2 gas
sensors: comparative electrical and spectroscopic studies, Solid State Ion. 2830 (1988) 1631.
[345] B. Slater, C.R.A. Catlow, D.E. Williams, A.M. Stoneham, Dissociation of O2 on the reduced SnO2(1 1 0)
surface, Chem. Commun. 0 (2000) 1235.
[346] C.R.A. Catlow, C.M. Baker, R.G. Bell, S.T. Bromley, D.S. Coombers, F. Cora, S. French, B. Slater,
A.A. Sokol, NATO ASI Series, Kluwer Academic Press.
[347] Y. Yamaguchi, Y. Nagasawa, A. Murakami, K. Tabata, Stability of oxygen anions and hydrogen
abstraction from methane on reduced SnO2(1 1 0) surfaces, Int. J. Quant. Chem. 69 (1998) 669.
[348] J. Oviedo, M.J. Gillan, First-principles study of the interaction of oxygen with the SnO2(1 1 0) surface, Surf.
Sci. 490 (2001) 221.

152

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

[349] A. Maiti, J.A. Rodriguez, M. Law, P. Kung, J.R. McKinney, P. Yang, SnO2 nanoribbons as NO2 sensors:
insights from rst principles calculations, Nano Lett. 3 (2003) 1025.
[350] M.A. Maki-Jaskari, T.T. Rantala, V.V. Golovanov, Computational study of charge accumulation at
SnO2(1 1 0) surface, Surf. Sci. 577 (2005) 127.
[351] E.W. Thornton, P.G. Harrison, Tin oxide surfaces, J. Chem. Soc. Faraday Trans. 1 71 (1975) 461.
[352] N. Sergent, P. Gelin, L. Perier-Camby, H. Praliaud, G. Thomas, FTIR study of low-temperature CO
adsorption on high surface area tin(IV) oxide: probing Lewis and Brnsted acidity, Phys. Chem. Chem.
Phys. 4 (2002) 4802.
[353] J. Tamaki, M. Nagaishi, Y. Teraoka, N. Miura, N. Yamazoe, K. Moriya, Y. Nakamura, Adsorption
behavior of CO and interfering gases on SnO2, Surf. Sci. 221 (1989) 183.
[354] M. Melle-Franco, G. Pacchioni, CO adsorption on SnO2(1 1 0): cluster and periodic ab initio calculations,
Surf. Sci. 461 (2000) 54.
[355] G. Pacchioni, A.M. Ferrari, P.S. Bagus, Cluster and band structure ab initio calculations on the adsorption
of CO on acid sites of the TiO2(1 1 0) surface, Surf. Sci. 350 (1996) 159.
[356] T. Bredow, G. Pacchioni, Comparative periodic and cluster ab initio study on Cu2O(1 1 1)/CO, Surf. Sci.
373 (1997) 21.
[357] T. Bredow, A.M. Marquez, G. Pacchioni, Analysis of electronic contributions to the vibrational frequency
of CO/Cu2O(1 1 1), Surf. Sci. 430 (1999) 137.
[358] F. Ciriaco, L. Cassidei, M. Cacciatore, G. Petrella, First principle study of processes modifying the
conductivity of substoichiometric SnO2 materials upon adsorption of CO from atmosphere, Chem. Phys.
303 (2004) 55.
[359] M. Melle-Franco, G. Pacchioni, A.V. Chadwick, Cluster and periodic ab initio calculations on the
adsorption of CO2 on the SnO2(1 1 0) surface, Surf. Sci. 478 (2001) 25.
[360] M.W. Abee, D.F. Cox, NH3 chemisorption on stoichiometric and oxygen-decient SnO2(1 1 0) surfaces,
Surf. Sci. 520 (2002) 65.
[361] E.W. Thornton, P.G. Harrison, Tin oxide surfaces. Part 3.Infrared study of the adsorption of some small
organic molecules on tin(IV) oxide, J. Chem Soc. Faraday Trans. I 71 (1975) 2468.
[362] F. Ouyang, S. Yao, K. Tabata, E. Suzuki, Infrared identication of reaction route from adsorbed species
derived from adsorption of methanol on SnO2, Appl. Surf. Sci. 158 (2000) 28.
[363] T. Kawabe, K. Tabata, E. Suzuki, Y. Nagasawa, Methanol adsorption on an oxidized and a reduced SnO2
thin lm, Surf. Sci. 454456 (2000) 374.
[364] T. Kawabe, K. Tabata, E. Suzuki, Y. Nagasawa, Methanol adsorption on SnO2 thin lms with dierent
morphologies, Surf. Sci. 482485 (2001) 183.
[365] M. Calatayud, J. Andres, A. Beltran, A theoretical analysis of adsorption and dissociation of CH3OH on
the stoichiometric SnO2(1 1 0) surface, Surf. Sci. 430 (1999) 213.
[366] M.W. Abee, D.F. Cox, BF3 adsorption on stoichiometric and oxygen-decient SnO2(1 1 0) surfaces, J. Phys.
Chem. B 107 (2003) 1814.
[367] V.A. Gercher, D.F. Cox, Formic acid decomposition on SnO2(1 1 0), Surf. Sci. 312 (1994) 106.
[368] J.S.G. Irwin, A.F. Prime, P.J. Hardman, P.L. Wincott, G. Thornton, The interconversion of SnO2(1 1 0)
reduced phases by reaction with formic acid, Surf. Sci. 480 (1996) 352.
[369] M. Batzill, K. Katsiev, U. Diebold, Tuning the oxide/organic interface: benzene on SnO2(1 0 1), Appl. Phys.
Lett. 85 (2004) 5766.
[370] M.A. Henderson, The interaction of water with solid surfaces: fundamental aspects revisited, Surf. Sci. Rep.
46 (2002) 1.
[371] P.A. Thiel, T.E. Madey, The interaction of water with solid surfaces: fundamental aspects, Surf. Sci. Rep. 7
(1987) 211.
[372] J.F. Boyle, K.A. Jones, The eects of CO, water vapor and surface temperature on the conductivity of a
SnO2 gas sensor, J. Electron. Mater. 6 (1977) 717.
[373] M. Egashira, M. Nakashima, S. Kawasumi, T. Seiyama, Temperature programmed desorption study of
water adsorbed on metal oxides: 2. Tin oxide surfaces, J. Phys. Chem. 85 (1981) 4125.
[374] G. Korotchenkov, V. Brynzari, S. Dmitriev, Electrical behavior of SnO2 thin lms in humid atmosphere,
Sensors Actuat. B 54 (1999) 197.
[375] M. Egashira, S. Kawasumi, S. Kagawa, T. Seiyama, Temperature programmed desorption study of water
adsorbed on metal oxides. I. Anatase and rutile, Bull. Chem. Soc. Jpn. 51 (1978) 3144.
[376] D.E. Williams, in: P.T. Moseley, B.C. Toeld (Eds.), Solid State Gas Sensors, 71, Adam Hilger IOP,
Publishing, 1987 (Chapter 5.1).

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

153

[377] N. Barsan, R. Ionescu, The mechanism of the interaction between CO and an SnO2 surface: the role of
water vapour, Sensors Actuat. B 12 (1993) 71.
[378] R. Ionescu, A. Vancu, C. Moise, A. Tomescu, Role of water vapour in the interaction of SnO2 gas sensors
with CO and CH4, Sensors Actuat. B 61 (1999) 39.
[379] S.H. Hahn, N. Barsan, U. Weimar, S.G. Ejakov, J.H. Visser, R.E. Soltis, CO sensing with SnO2 thick lm
sensors: role of oxygen and water vapour, Thin Solid Films 436 (2003) 17.
[380] J. Goniakowski, M.J. Gillan, The adsorption of H2O on TiO2 and SnO2(1 1 0) studied by rst-principles
calculations, Surf. Sci. 350 (1996) 145.
[381] P.J.D. Lindan, Water chemistry at the SnO2(1 1 0) surface: the role of inter-molecular interactions and
surface geometry, Chem. Phys. Lett. 328 (2000) 325.
[382] P.J.D. Lindan, N.M. Harrison, M.J. Gillan, Mixed dissociative and molecular adsorption of water on the
rutile (1 1 0) surface, Phys. Rev. Lett. 80 (1998) 762.
[383] S.P. Bates, Full-coverage adsorption of water on SnO2(1 1 0): the stabilization of the molecular species, Surf.
Sci. 512 (2002) 29.
[384] V.A. Gercher, D.F. Cox, Water adsorption on stoichiometric and defective SnO2(1 1 0) surfaces, Surf. Sci.
322 (1995) 177.
[385] M. Batzill, W. Bergermayer, I. Tanaka, U. Diebold. Tuning the chemical response of a gas sensitive
material: water adsorption on SnO2(1 0 1), unpublished.
[386] Z.R. Dai, Z.W. Pan, Z.L. Wang, Novel nanostructures of functional oxides synthesized by thermal
evaporation, Adv. Funct. Mater. 13 (2003) 9.
[387] B. Thiel, R. Helbig, Growth of SnO2 single crystals by a vapour phase reaction method, J. Cryst. Growth 32
(1976) 259.
[388] Z.L. Wang, Functional oxide nanobelts: materials, properties and potential applications in nanosystems
and biotechnology, Annu. Rev. Phys. Chem. 55 (2004) 159.
[389] J.Q. Hu, Y. Bando, D. Goldberg, Self-catalyst growth and optical properties of novel SnO2 shbone-like
nanoribbons, Chem. Phys. Lett. 372 (2003) 758.
[390] N.S. Ramgir, I.S. Mulla, K.P. Vijayamohanan, Shape selective synthesis of unusal nanobipyramids, cubes,
and nanowires of RuO2: SnO2, J. Phys. Chem. B 108 (2004) 14815.
[391] E.R. Leite, J.W. Gomes, M.M. Oliviera, E.J.H. Lee, E. Longo, J.A. Varela, C.A. Paskocimas, T.M. Boschi,
F. Lanciotti Jr., P.S. Pizani, P.C. Soares Jr., Synthesis of SnO2 nanoribbons by a carbothermal reduction
process, J. Nanosci. Nanotech. 2 (2002) 125.
[392] J.X. Wang, D.F. Liu, X.Q. Yan, H.J. Yuan, L.J. Ci, Z.P. Zhou, Y. Gao, L. Song, L.F. Liu, W.Y. Zhou, G.
Wang, S.S. Xie, Growth of SnO2 nanowires with uniform branched structures, Solid State Commun. 130
(2004) 89.
[393] J. Hu, Y. Bando, Q. Liu, D. Goldberg, Laser-ablation growth and optical properties of wide and long
single-crystal SnO2 ribbons, Adv. Funct. Mater. 13 (2003) 493.
[394] C. Ye, X. Fang, Y. Wang, T. Xie, A. Zhao, L. Zhang, Novel synthesis of tin dioxide nanoribbons via a mild
solution approach, Chem. Lett. 33 (2004) 54.
[395] C.-F. Wang, S.-Y. Xie, S.-C. Lin, X. Cheng, X.H. Zhang, R.-B. Huang, L.S. Zheng, Glow discharge growth
of SnO2 nano-needles from SnH4, Chem. Commun. (2004) 1766.
[396] Y. Liu, J. Dong, M. Liu, Well-aligned nano-box-beams of SnO2, Adv. Mater. 16 (2004) 353.
[397] Y. Wang, J.Y. Lee, T.C. Deivaraj, Controlled synthesis of V-shaped SnO2 nanorods, J. Phys. Chem. B 108
(2004) 13589.
[398] B. Cheng, J.M. Russell, W. Shi, L. Zhang, E.T. Samulski, Large-scale, solution-phase growth of singlecrystalline SnO2 nanorods, J. Am. Chem. Soc. 126 (2004) 5972.
[399] D.-F. Zhang, L.-D. Sun, J.-L. Yin, C.-H. Yan, Low-temperature fabrication of highly crystalline SnO2
nanorods, Adv. Mater. 15 (2003) 1022.
[400] E. Comini, G. Faglia, G. Sberveglieri, Z. Pan, Z.L. Wang, Stable and highly sensitive gas sensors based on
semiconducting oxide nanobelts, Appl. Phys. Lett. 81 (2002) 1869.
[401] A. Maiti, Electromechanical and chemical sensing at he nanoscale: Molecular modeling applications, Mol.
Simul. 30 (2004) 191.
[402] E. Comini, V. Guidi, C. Malagu`, G. Martinelli, Z. Pan, G.-S. Sberveglieri, Z.L. Wang, Electrical properties
of tin dioxide two-dimensional nanostructures, J. Phys. Chem. B 108 (2004) 1882.
[403] M.S. Arnold, P. Avouris, Z.W. Pan, Z.L. Wang, Field-eect transistors based on single semiconducting
oxide nanobelts, J. Phys. Chem. B 107 (2003) 659.

154

M. Batzill, U. Diebold / Progress in Surface Science 79 (2005) 47154

[404] Q.H. Li, Y.J. Chen, Q. Wan, T.H. Wang, Thin lm transistors fabricated by in situ growth of SnO2
nanobelts on Au/Pt electrodes, Appl. Phys. Lett. 85 (2004) 1805.
[405] S.X. Mao, M. Zhao, Z.L. Wang, Nanoscale mechanical behavior of individual semiconducting nanobelts,
Appl. Phys. Lett. 83 (2003) 993.
[406] L. Shi, Q. Hao, C. Yu, N. Mingo, X. Kong, Z.L. Wang, Thermal conductivities of individual tin dioxide
nanobelts, Appl. Phys. Lett. 84 (2004) 2638.
[407] G. Cicero, A. Catellani, G. Galli, Atomic control of water interaction with biocompatible surfaces: the case
of SiC(0 0 1), Phys. Rev. Lett. 93 (2004) 016102.
[408] J.-S. Kim, F. Cacialli, R. Friend, Surface conditioning of indium-tin oxide anodes for organic light-emitting
diodes, Thin Solid Films 445 (2003) 358.

Вам также может понравиться