The morphological evolution of the Bi2Mo3O 12(010) surface in air-H2O atmospheres

Article abstract of DOI:10.1016/S0021-9517(02)00024-6

Atomic force microscopy (AFM) has been used to examine the morphological evolution of the Bi₂Mo₃O₁₂(010) surface at 400-600°C in dry air and air-2.3% H₂O. The (010) cleavage surface is characterized by atomically flat terraces separated by straight steps that are integer multiples of b/2 (5.75 A) in height. During treatments at or above 500°C, the surface is etched due to the volatilization of Mo. In dry air, etching affects both steps and flat terraces and results in step recession, the development of half-unit-cell (b/2) step loops (pits and islands), and the accumulation of Bi-rich surface deposits. In air-2.3% H₂O, steps are etched with preference to terraces, and this leads to step recession as well as the formation of Bi-rich deposits. Mo volatilization proceeds at an enhanced rate in air-2.3% H₂O and culminates in the nucleation and growth of Bi₂MoO₆ and Bi₂Mo₂O₉ precipitates at 500 and 600°C, respectively.

Full text of DOI:10.1016/S0021-9517(02)00024-6

Journal of Catalysis 213 (2003) 151–162
www.elsevier.com/locate/jcat
The morphological evolution of the Bi2Mo3O12(010) surface
in air–H2O atmospheres
Svetlana V. Yanina and Richard L. Smith ∗
Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
Received 16 April 2002; revised 20 September 2002; accepted 26 September 2002
Abstract
Atomic force microscopy (AFM) has been used to examine the morphological evolution of the Bi Mo O (010) surface at 400–600 C
◦
2
3 12
in dry air and air–2.3% H O. The (010) cleavage surface is characterized by atomically flat terraces separated by straight steps that are
2
◦
integer multiples of b/2 (5.75 Å) in height. During treatments at or above 500 C, the surface is etched due to the volatilization of Mo. In
dry air, etching affects both steps and flat terraces and results in step recession, the development of half-unit-cell (b/2) step loops (pits and
islands), and the accumulation of Bi-rich surface deposits. In air–2.3% H O, steps are etched with preference to terraces, and this leads to
2
step recession as well as the formation of Bi-rich deposits. Mo volatilization proceeds at an enhanced rate in air–2.3% H O and culminates
2
◦
in the nucleation and growth of Bi MoO and Bi Mo O precipitates at 500 and 600 C, respectively.
2
6
2
2 9
2002 Elsevier Science (USA). All rights reserved.
Keywords: Molybdate; Scanning probe microscopy; Partial oxidation; Volatilization; Molybdenum oxide
1. Introduction
The results of bulk structural studies have generally
indicated that Bi2Mo3O12 is stable at elevated tempera-
tures (T ꢀ 500 ◦C) in air and under conditions typical
for the selective oxidation of alkenes [8–11]. At the same
time, investigations with local bulk and surface-sensitive
probes have shown that the structure and composition near
Bi2Mo3O12 surfaces can depart from that of the bulk. For
example, using in situ transmission electron microscopy
(TEM), Gai [13,14] observed the formation of a phase simi-
lar to B◦i2Mo2O9 (as well as Bi and MoO3) within minutes at
∼ 440 C in propene and propene–O2 (1 : 1) atmospheres.
The Bi2Mo2O9-like precipitates grew with reaction time and
extended into the Bi2Mo3O12 bulk, but they were unsta-
ble outside of the reducing environment of the in situ TEM
cell [13]. X-ray photoelectron spectroscopy (XPS) analy-
ses of calcined Bi2Mo3O12 powders have generally indi-
cated near-surface Bi/Mo ratios similar to that expected
based on the bulk stoichiometry (i.e., Bi/Mo ∼ 2/3), but ra-
tios near unity have been observed in Bi2Mo3O12–Bi2MoO6
mixtures and in nonstoichiometric Bi2Mo3O12 (excess Bi)
and Bi2MoO6 (excess Mo) [7,8,15,16]. Matsuura and co-
workers [15] attributed this shared surface chemistry to a
common surface layer similar to Bi2Mo2O9, but alternate
explanations based on surface oxide mixtures with an aver-
age Bi/Mo ∼ 1 have also been advanced [17,18]. In a more
Bismuth molybdates have been the subject of numer-
ous investigations related to their ability to catalyze the
selective oxidation and ammoxidation of alkenes [1–12].
While industrial catalysts are usually complex multicom-
ponent molybdates, most research has focused on the sim-
pler binary oxides of the Bi2O3–MoO3 system, particu-
larly Bi2Mo3O12 (also known as α-bismuth molybdate),
Bi2Mo2O9 (β-bismuth molybdate), Bi2MoO6 (γ -bismuth
molybdate), and their mixtures [1–6]. As a result of these
studies, a considerable amount is known about the bulk
structures of the individual compounds as well as the overall
mechanisms of partial oxidation over them [6]. Compara-
tively little is known, however, about the surface structures
of the bismuth molybdates and, as a result, the role surface
structure plays in their reactivity [7]. These uncertainties
motivated this atomic force microscopy (AFM) study of the
Bi2Mo3O12(010) surface and its evolution during treatments
in air and air–2.3% H2O.
*
Corresponding author.
E-mail address: smithrl@mit.edu (R.L. Smith).
URL address: http://web.mit.edu/smithrl/www/.
0021-9517/02/$ – see front matter
PII: S0021-9517(02)00024-6
2002 Elsevier Science (USA). All rights reserved.

152
S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
recent XPS study, Briand and co-workers [19] reported sur-
face enrichment of Mo (Bi/Mo ∼ 0.55) in co-precipitated
Bi2Mo3O12 powders that had been calcined at 500 ◦C. Based
on reactions with a probe molecule (methanol), the authors
deduced that Bi2Mo3O12 surfaces are largely composed of
MoOx species, in the form of a two-dimensional overlayer,
due to spontaneous spreading similar to that observed when
oxide supports are mixed with MoO3 [19–21].
While XPS and TEM have provided some insight into
the chemistry and phase composition near the surfaces of
Bi2Mo3O12 particles, little is known about the structure and
morphology of the individual surfaces that actually bound
the particles [7]. Furthermore, the structures and morpholo-
gies of these surfaces are expected to change during reac-
tions at elevated temperatures, in response to a variety of
thermochemical stimuli. Understanding these details of sur-
face structure and dynamics is important because they bear
directly on the population and configurations of surface sites
and, therefore, reactivity. The objectives of this study were
to characterize the structure of the Bi2Mo3O12(010) sur-
face and its evolution during treatments in dry air and air–
2.3% H2O between 400 and 600 ◦C. This has been accom-
plished through a combination of atomic force microscopy
(AFM), X-ray diffraction (XRD), and X-ray photoelectron
spectroscopy (XPS) experiments on single crystal cleavage
surfaces.
a 78% MoO3 (99+%, Aldrich)–22% Bi2O3 (99%, Aldrich)
◦
molar mixture was first melted at 710 C in an Al2O3 cru-
cible. After 5 h at◦710 ◦C, the melt was cooled slowly
◦
(0.5 ◦C/h) to 600 C and then more rapidly (∼ 20 C/h)
to room temperature. The solidified product consisted of a
crumbly mixture of Bi2Mo3O12 and MoO3 crystals from
which Bi2Mo3O12 plates up to 1.5 cm on edge could be
separated. In the directional solidification approach, crystals
were grown by slowly cooling a MoO3–25% Bi2O3 melt
across a temperature gradient in a horizontal tube furnace.
The growths were conducted in quartz tubes (i.d. = 13 mm)
and yielded irregular boules (3–4 cm in length) composed of
several Bi2Mo3O12 crystals/grains.
2.2. Surface treatments
Bi2Mo3O12 single crystals were treated in a continuous
flow of the desired gas phase in a quartz reaction tube
(length = 1 m, i.d. = 2.5 cm). Prior to treatment, fresh
(010) and (010) surfaces were formed by cleavage with
a razor blade in the ambient. Typically, 3–4 crystals were
cleaved in half to yield 3–4 pairs of opposing (010) and
(010) surfaces. The terminations of the Bi2Mo3O12(010)
and (010) cleavage surfaces were not, at the outset, expected
to be equivalent. Based on our observations, however, they
appear to be related in a simple manner, which will be
elaborated in Section 2.1. Following cleavage, the crystals
were immediately placed in a high-purity Al2O3 boat and
loaded into the reaction tube. The tube was then doubly
purged by alternately evacuating with an oil-less pump and
backfilling with the gas mixture of interest. A continuous
gas flow of 100 cm3 min−1 at 1 atm pressure was then
established. The samples were subsequ◦ently heated to the
desired reaction temperature (400–600 C) in a horizontal
split-tube furnace. It took less than 10 min to reach this
temperature, and the actual temperature did not exceed
the desired temperature by more than 1% throughout the
thermal cycle. Temperature was monitored externally with
a thermocouple that was positioned in contact with the
reaction tube, directly adjacent to the samples.
The (010) surface of Bi2Mo3O12 was the focus of
study for a number of reasons. Bi2Mo3O12 crystallites are
characteristically platy with large (010) and (010) faces
and, as a result, these faces compose a significant fraction
of the total surface areas of powdered samples. From a
more practical experimental standpoint, Bi2Mo3O12 crystals
are readily cleaved normal to ꢀ010ꢁ, making it possible
to easily and reproducibly prepare fresh (010) surfaces of
bulk stoichiometry. Although bismuth molybdate catalysts
◦
are typically used at or below 500 C, they are sometimes
processed (i.e., calcined) above this temperature. For this
reason, studies beyond normal operating temperatures were
undertaken. Water vapor is of course normally present
during calcining treatments as well as during selective
oxidation reactions, at the least as a byproduct, and it
is known to enhance the volatility of Mo [22–30]. Our
results demonstrate that this volatility plays a central role
in the evolution of the structure and chemistry of the
Bi2Mo3O12(010) surface during thermal treatments.
The Bi2Mo3O12 samples were treated for predetermined
periods of time between 2 and 300 h. The principal at-
mospheres investigated were dry air (BOC) and air saturated
with H2O at 24 1 ◦C (∼ 2.3% H2O). The dry air was used
as received and had an H2O content of less than 10 ppm. The
air–2.3% H2O mixture was prepared by passing the dry air
through two fritted H2O bubblers connected in series. Fol-
lowing reaction, the samples were quickly cooled to room
temperature under a flow of the test gas. Cooling was has-
tened by lifting the reaction tube out of the furnace. By doing
so, it took less than 5 min for the samples to cool to below
2. Methods
2.1. Bi2Mo3O12 crystal growth
The (010)-oriented Bi2Mo3O12 single crystal plates
(∼ 5 × 4 × 1 mm3) used in this study were cleaved from
larger samples grown by flux and directional solidification
approaches [31,32]. The plates were optically transparent
and had a yellow hue. In a typical flux growth, 200–300 g of
◦
200 C, a temperature where prolonged treatments had no
discernible (with AFM) effect on the Bi2Mo3O12(010) sur-
face.

S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
153
2.3. Surface characterization
X-ray photoelectron spectroscopy (XPS) (AXIS Ultra,
Kratos Analytical, Manchester, UK) was used to probe the
chemical compositions of cleaved and treated
Bi2Mo3O12(010) surfaces. Of particular interest was the rel-
ative composition of Bi and Mo near the surface (i.e., the
atomic Bi/Mo ratio). Following the desired treatments, the
single crystals were immediately transferred to the load-lock
of the XPS unit. Subsequently, XPS analysis was performed
using monochromatized Al-Kα radiation. For any particu-
lar treatment condition, three or four different single crystal
surfaces were analyzed. In the results section, the range of
Bi/Mo ratios measured on the surfaces are reported. To en-
hance the surface sensitivity, the XPS samples were tilted
(∼ 70◦) for analysis. Following XPS, the samples’ surface
morphologies were characterized with AFM.
Once the samples had cooled to room temperature, they
were removed from the reaction tube and the exposed (010)
and (010) surfaces were immediately characterized using
atomic force microscopy (AFM) (Digital Instruments, Santa
Barbara, CA) in the ambient. The surfaces were examined
over the course of several hours, and at least three areas
on four different crystals were studied. There were no
perceptible (with AFM) changes in surface structure over
the analysis period or even over extended (several days)
exposures to the ambient. The AFM observations were made
in the contact mode using Si3N4 probes with V-shaped
cantilevers and pyramidal tips. The vertical calibration of
the AFM scanner was checked periodically using steps on
cleaved MoO3(010) surfaces as standards [33,34].
X-ray diffraction (XRD) was used to characterize the
phase composition of treated single crystal surfaces. The
(010) and/or (010) surface layers of the single crystals were
selectively sampled, either by scraping the surfaces with
a razor blade or by extraction with adhesive tape. Due to
the small quantities of material available, the XRD samples
were mounted on glass slides using double-sided adhesive
tape. XRD analysis was performed on a θ–θ diffractometer
(Rigaku, Tokyo, Japan) using Cu-Kα radiation supplied by a
rotating anode generator operated at 60 kV and 300 mA. In
some cases, parallel XRD experiments were also conducted
with coarse Bi2Mo3O12 powders (∼ 0.25 g). The starting
powders were prepared by crushing single crystals with an
Al2O3 mortar and pestle.
3. Results
3.1. The structure of the Bi2Mo3O12(010) and (010)
cleavage surfaces
The morphologies of Bi2Mo3O12(010) and (010) cleav-
age surfaces were indistinguishable. Both were character-
ized by atomically flat terraces separated by straight steps
(see Fig. 1). The distribution of steps across cleaved sur-
faces varied. Some regions had step populations similar to
those depicted in Fig. 1, while others presented singular flat
terraces as large as 100 × 100 µm2, the upper limit of the
AFM scanner. Cleavage surfaces exhibited two limiting step
Fig. 1. Topographic AFM images of the limiting step morphologies on Bi Mo O (010) and (010) cleavage surfaces. (a) Straight steps with edges parallel to
2
3 12
ꢀ100ꢁ and heights that are an integer multiple of b/2. (b) Serrated terraces characteristically bounded by half-unit-cell steps accompanied by small triangular
terraces (inset) bounded by quarter-unit-cell steps. The black-to-white contrasts in (a) and (b) correspond to topographic ranges of approximately 30 and 80 Å,
respectively.

154
S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
morphologies that, along with singular terraces, could usu-
ally be found in multiple regions on any particular surface.
Most often, terraces were bounded by straight steps that ex-
tended for hundreds of micrometers along ꢀ100ꢁ, as shown
in Fig. 1a. The heights of these steps were invariably an inte-
ger multiple of 5.8 0.5 Å, which corresponds to one-half of
the b lattice parameter of Bi2Mo3O12 (b/2 = 5.75 Å) [12].
In most cases, straight steps parallel to ꢀ100ꢁ were single-
unit-cell steps (height = b = 11.5 Å).
Terraces with a triangular or serrated morphology were
also formed during cleavage (see Fig. 1b). These ter-
races were characteristically bounded by half-unit-cell steps
(height = b/2), but they were often accompanied by smaller
triangular terraces bounded by quarter-unit-cell steps
(height = b/4). The b/4-steps/terraces were localized at the
protruding tips of the serrated terraces and usually extended
less than 400 nm before transitioning to half-unit-cell steps
(inset of Fig. 1b). During the course of study, multiple re-
gions on at least 20 different cleavage surfaces were char-
acterized, and steps whose heights were not an integer mul-
tiple of b/2 were only observed in the vicinity of serrated
terraces. Even in areas with high densities of these terraces,
as in Fig. 1b, terraces bounded by b/4-steps composed less
than 1% of the total (010) terrace area. XPS analysis of (010)
and (010) cleavage surfaces indicated near-surface compo-
sitions consistent with the bulk stoichiometry of Bi2Mo3O12
(i.e., Bi/Mo = 2/3 = 0.667). Six cleavage surfaces from dif-
ferent crystals were analyzed, and all exhibited Bi/Mo ratios
between 0.65 and 0.68.
(a)
The crystal structure of Bi2Mo3O12 (space group =
P21/c) has been described by Van Den Elzen and Rieck
[12] and can be viewed as a distorted Scheelite with ordered
Bi vacancies. The structure contains three distinct Mo sites,
each of which is nominally coordinated by five O. In each
case, four of the Mo–O bonds are short (1.68 to 1.91 Å),
while the fifth is considerably longer (> 2.13 Å). As a
result, the Mo coordination is usually regarded as tetrahedral
[12–15]. There are two Bi sites. Each has four close O
neighbors (at distances of 2.12 to 2.34 Å) and four that
are much further away (2.61 to 2.94 Å). The Bi and Mo
are concentrated in layers parallel to (010) that intersect
b at approximately 1/8, 3/8, 5/8, and 7/8 (see Fig. 2a).
Adjacent layers are bridged and held together by metal–
oxygen, primarily Bi–O, bonds. (010) cleavage is expected
to occur between these cation layers at one or more of the
planes that intersect b at ∼ 1/4, 1/2, 3/4, and 1. The fact that
XPS data indicate a Bi/Mo ratio ∼ 2/3 is consistent with
this expectation, since each cation layer has a Bi/Mo ratio
of 2/3. Due to the 21 screw axis along ꢀ010ꢁ, the planes at
1/4 and 3/4 are equivalent, though rotated 180◦with respect
to one another. The planes at 1/2 and 1 are related in the
same manner, but they are not equivalent to those at 1/4 and
3/4. Thus, there are two potential cleavage systems, 1/2–1
and 1/4–3/4, each composed of equivalent planes separated
by b/2.
(b)
Fig. 2. (a) Bonding model of the Bi Mo O structure viewed along
ꢀ100ꢁ with the potential cleavage planes indicated. In this representation,
the Mo coordination is assumed to be tetrahedral, while that of Bi is
assumed to be eightfold. (b) Ionic space-filling model of the bulk terminated
2
3 12
Bi Mo O (010) surface formed through cleavage at the 1/2–1 system. In
2
3 12
−2
this representation, the radius of O is assumed to be 1.4 Å. For clarity, the
radii of Bi and Mo have been slightly exaggerated to 1.1 and 0.6 Å,
respectively.
+3
+6
Although there are two viable cleavage systems, our
AFM observations indicate that cleavage is dominated by
one system. This is because steps that were not an integer
multiple of b/2 (the distance between equivalent planes in
each cleavage system) in height were encountered in only
small populations during AFM studies. Furthermore, when
such steps (i.e., the b/4-steps) were encountered, they were
localized and extended less than 400 nm before transitioning
to half-unit-cell (b/2) steps. Taken together, these character-
istics point to a single, dominant cleavage system. Conse-
quently, the terraces on cleaved Bi2Mo3O12(010) surfaces
will, with the exception of the small population bounded by
b/4-steps, have the same atomic termination. Note, however,

S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
155
that the terminations of terraces separated by steps that are
an odd multiple of b/2 will be rotated 180◦ with respect to
one another. Terraces on (010) will be related in the same
manner, but their termination will differ from that on (010),
as will be described shortly.
tures, as noted earlier, or in their behavior during treatments
in dry air and air–2.3% H2O. Therefore, we will simply re-
fer to the (010) and (010) collectively as (010) through the
remainder of the text.
While AFM observations indicate that cleavage is domi-
nated by one cleavage system, they do not allow us to dis-
tinguish between the two candidates. Consideration of the
bonding across the cleavage planes does, however, permit
distinction. 1/2–1 cleavage would nominally require break-
ing fewer bonds than 1/4–3/4 cleavage. For each unit cell
cleaved, resulting in the creation of opposing (010) and
(010) surface cells, eight Bi–O and two Mo–O bonds would
be broken across 1/2 or 1, while eight Bi–O and four Mo–O
bonds would be broken across 1/4 or 3/4. In both systems
however, the broken Mo–O bonds are associated with the
fifth, distant O (Mo–O > 2.13 Å) around each Mo. These
bonds are presumably broken easily, leaving the tetrahedral
coordination of Mo intact. A clear distinction between the
cleavage systems is evident when the Bi–O bonding is con-
sidered. The bonds broken at 1/2–1 are on average longer
and, therefore, expected to be weaker than those at 1/4–
3/4. All eight of the Bi–O bonds at 1/2–1 can be viewed
as long and weak, with lengths of 2.60 to 2.79 Å. In con-
trast, cleavage at 1/4–3/4 would require breaking a num-
ber of short Bi–O linkages. Four are long (2.67 to 2.94 Å),
but the other four are much shorter (2.22 to 2.24 Å). Hence,
cleavage is favored at the 1/2–1 system. In the rare instances
where terraces bounded by b/4-steps occurred, cleavage
must have also propagated along the 1/4–3/4 system for
short distances. With the identification of the cleavage sys-
tem, a model for the bulk-terminated Bi2Mo3O12(010) sur-
face can be proposed (see Fig. 2b). Based on the bonding
arguments outlined above, Mo at the surface is tetrahedrally
coordinated by O, while each Bi has two dangling bonds.
Cleavage at 1/4–3/4 would yield similar coordination envi-
ronments, except the broken Bi–O bonds would be shorter.
As noted earlier, the terminations of the opposing (010)
and (010) cleavage surfaces are expected to be different. The
cleavage plane that intersects b at 1 (and 1/2) contains a
center of inversion. Therefore, the crystal elements across
the plane are related through the combination of a mirror
and a 180◦ rotation, and the terminations of (010) and (010)
are mirror images. They are, however, nonsuperimposable
mirror images (i.e., enantiomers), because their respective
surface cells are oblique (plane group = P1). Though un-
favorable, cleavage at 1/4–3/4 would also lead to enan-
tiomorphic surfaces, because the cleavage plane coincides
with the c glide plane in the Bi2Mo3O12 structure. Since
only relatively weak bonds are ruptured during cleavage, sur-
face reconstructions that might destroy the enantiomorphic
relationship between (010) and (010) are not anticipated
[33–37]. Regardless, the opposing surfaces will be related
by a mirror and would be expected to behave identically in
achiral environments, such as air–H2O [38]. Through this
study, differences were not observed in their cleavage struc-
3.2. Evolution of Bi2Mo3O12(010) in air–H2O
(T ꢀ 500 ◦C)
The Bi2Mo3O12(010) cleavage surface was not modified
in any manner detectable with AFM during dry air anneals
in excess of 100 h at 400 ◦C. Within 2 h at 500 ◦C,
morphological changes were apparent, and surfaces were
marked by irregularly shaped islands and pits with lateral
dimensions less than 300 nm (see Fig. 3a). The heights of the
islands above (and pits below) adjacent terraces varied, even
across any particular one of these features, but ranged from
3 to 6 Å (∼ b/4–b/2). Due to the general roughness of the
surfaces, it was often difficult to identify the transitions from
distinct islands (or pits) to terraces [39]. While the region in
Fig. 3a principally contained islands, pits were prominent
in other areas, indicating that the Bi2Mo3O12(010) surface
was eroded or etched during the treatment. This etching
was nonspecific, as terraces and steps were affected with
essentially equal probability. The morphologies of (010)
surfaces treated for 4 h at 500 ◦C in dry air were not
significantly different from those of surfaces treated for 2 h.
XPS analysis of surfaces treated for 4 h indicated a slight
increase in the Bi concentration (relative to the cleavage
surface); the Bi/Mo ratios of the surfaces all fell between
0.69 and 0.74.
In contrast to surfaces treated in dry air, Bi2Mo3O12(010)
surfaces treated in air–2.3% H2O were altered during treat-
ments at 400 ◦C. These surfaces also exhibited morphologi-
cal changes consistent with erosion or etching, but the etch-
ing was specific, as only steps were affected. Etching was
initially evident in the “rounding-off” of the protruding tips
of serrated terraces and in the absence of the associated b/4-
terraces that were observed on cleavage surfaces (Fig. 1b).
This rounding was confined to within ∼ 500 nm of the ter-
◦
race tips after 48 h at 400 C. Etching was accelerated at
500 ◦C, and changes in step morphology were apparent af-
ter only 2 h (see Fig. 3b). These changes included rounding
of the step tips as well as step recession (advance of a step
toward its upper terrace). Flat (010) terraces appeared un-
changed after 2 h in air–2.3% H2O at 500 ◦C and were indis-
tinguishable from those on fresh cleavage surfaces. Surfaces
◦
treated for 4 h at 500 C in air–2.3% H2O exhibited simi-
lar step-specific etching. Despite the rather modest change
in surface morphology during the treatment, XPS analysis
indicated significant Bi enrichment near the surface. Three
different surfaces were analyzed, and all exhibited Bi/Mo
ratios between 0.76 and 0.80.
With extended reactions at 500 ◦C, morphological chan-
ges in the Bi2Mo3O12(010) surface were more pronounced.
Within 48 h in dry air, the irregular islands and pits of
variable height (Fig. 3a) gave way to micrometer-sized step

156
S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
◦
Fig. 3. (a) Topographic AFM image of a Bi Mo O (010) surface following treatment for 2 h at 500 C in dry air. (b) AFM deflection image of a
2
3 12
◦
Bi Mo O (010) surface treated for 2 h at 500 C in air–2.3% H O. The black-to-white contrast in (a) corresponds to a topographic range of 60 Å.
2
3
12
2
loops (islands) that were a uniform half unit cell (b/2) in
height (see Fig. 4a). Islands persisted through treatments up
to 120 h (the longest time investigated), but surface pits were
only rarely observed. The (010) terraces that separated the
islands were usually decorated with fine deposits 3–5 Å high
that made the surfaces appear “grainy” in AFM images. Due
to their small size, the lateral dimensions of these deposits
could not be reliably assessed [39]. Etching at steps was
also apparent (see Fig. 4b). Steps receded irregularly but
as discrete units that were b/2 in height (i.e., as half-unit-
cell steps). As a result, the straight steps characteristic of the
cleavage surface were transformed into irregular half-unit-
cell steps. Small deposits were also found in the vicinity
of steps that had receded. The deposits were often aligned
in rows along ꢀ100ꢁ, most likely marking the positions of
the receding steps at earlier points in time. Whether at steps
posits (see Fig. 4c). The pits were composed of half-unit-cell
step loops and invariably encircled the deposits, which had
heights of up to 20 Å after 72 h. While this pitting could
only have resulted from local material loss, the loss was in-
congruent, since some matter was left behind in the surface
deposits.
In contrast to flat terraces, steps were prone to modifica-
tion and etching at 500 ◦C in air–2.3% H2O. This resulted in
step recession, once again as discrete half-unit-cell steps, as
shown in Fig. 4d. The morphologies of adjacent steps of-
ten alternated from irregular and curved to comparatively
straight. The straight steps were usually decorated with de-
posits that had heights of 20–40 Å after 72 h. This alternating
step arrangement probably resulted from the dissociation of
straight single-unit-cell steps (which were characteristic of
the cleavage surface) into half-unit-cell steps, with the up-
per half receding while the lower remained in place. The
steps receded in an irregular fashion and left surface deposits
and, on some terraces, irregular half-unit-cell islands in their
wake. This combination of step recession and deposit ac-
cumulation indicates, once again, that the Bi2Mo3O12(010)
◦
or on flat terraces, etching in dry air at 500 C led to an
increase in the step density on Bi2Mo3O12(010) as well as
a complete change in step morphology, i.e., from straight to
irregular and curved. XPS revealed that etching also resulted
in a gradual increase in the surface Bi concentration with
reaction time. Samples analyzed after 72 h at 500 ◦C in dry
air exhibited Bi/Mo ratios of 0.72–0.76.
◦
surface is incongruently etched away at 500 C. In many
cases, deposits were located on terraces and islands bounded
by receding steps, but far from the steps themselves. This
suggests that steps had swept through these regions previ-
ously and, therefore, at least one b/2-thick layer of the sur-
face had already been etched away.
Acicular precipitates were usually apparent within 72 h
at 500 ◦C in air–2.3% H2O and can be seen in Fig. 4d.
These features had heights of 40–100 Å and were often ori-
ented at an angle of ∼ 40◦ with respect to the Bi2Mo3O12
ꢀ100ꢁ axis. The precipitates occurred in groups or clusters
that were encompassed by pits. The pits were comprised of
There was significant region-to-region variability in the
morphology of Bi2MO3O12(010) surfaces treated for ex-
◦
tended periods at 500 C in air–2.3% H2O. This variabil-
ity appeared to be influenced by the initial structure of the
cleavage surface, in particular the density of steps. Follow-
ing reactions of as long as 144 h, flat (010) terraces far
(ꢁ 10 µm) from steps often appeared unmodified over ar-
eas as large as 50 × 50 µm2. Intermittently, however, the
terraces were interrupted by isolated pits and, with reac-
tions in excess of 24 h, accompanying nanometer-scale de-

S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
157
◦
Fig. 4. Topographic AFM images of Bi Mo O (010) surfaces treated for 72 h at 500 C in dry air (a) and (b) and air–2.3% H O (c) and (d). The black-to-white
2
3
12
2
contrasts in (a)–(d) correspond to topographic ranges of 30, 60, 40, and 100 Å, respectively.
concentric half-unit-cell step loops and had depths of up to
6b (∼ 70 Å). In general, the widths and depths of the pits in-
creased with the total volume of precipitates they enveloped.
The surface deposits and the precipitates interacted differ-
ently with steps. Receding steps and islands often had sharp
and discontinuous changes in their curvature. These discon-
tinuities almost invariably coincided with the occurrence of
surface deposits, indicating that the deposits pinned the steps
and hindered their migration. In contrast, the acicular pre-
cipitates appeared to locally enhance step recession, partic-
ularly in directions away from the precipitates themselves.
Though areas such as those in Fig. 4c and 4d were
common on Bi2Mo3O12(010) surfaces treated in excess of
48 h at 500 ◦C in air–2.3% H2O, the microstructures of the
surfaces were dominated by a population of much larger
precipitates, which had lateral dimensions of up to 25 µm
after 72 h (see Fig. 5). The precipitates were embedded
in the (010) surface but often had portions that rose above
(50–100 nm) adjacent terraces. The distribution of these
precipitates was easily studied using optical microscopy. The
precipitates were generally dispersed uniformly across the
(010) surfaces of treated crystals and occupied 20–30% (by

158
S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
tube where the flowing gas exited the furnace. Thus, both
the Bi2Mo3O12 single crystals and powders lost Mo through
volatilization in air–2.3% H2O at 500 ◦C. The nucleation and
growth of a second phase (Bi2MoO6) richer in Bi indicates
that the extent of Bi loss was small relative to that of Mo
in air–2.3% H2O. In dry air, on the other hand, secondary
phases were not detected on (010) surfaces (via AFM or
optical microscopy) or in powders (via XRD) treated for
up to 120 and 250 h, respectively. Furthermore, there was
no indication of MoO3 (or other material) accumulation at
the reactor exhaust. Hence, Mo and Bi volatility in dry air
were both small compared to that of Mo in air–2.3% H2O.
Material was, however, still removed from the (010) surface
during treatments in dry air, as evidenced by incongruent
terrace etching and step recession, and the surface became
enriched with Bi, indicating that the volatility of Mo in dry
air was greater than that of Bi.
3.3. Evolution of Bi2Mo3O12(010) in air–H2O
(T = 600 ◦C)
Fig. 5. Topographic AFM image of
a
precipitate formed on
a
The modifications introduced to the Bi2Mo3O12(010)
surface at 600 ◦C were similar to those at 500 ◦C. Surfaces
were again marked by incongruent etching and the array of
accompanying morphological changes, including step reces-
sion, half-unit-cell step loops (pits and islands), and deposits.
There were, however, some noteworthy differences. First,
the rate at which these features developed was accelerated
at the higher temperature. Additionally, the step morphol-
◦
Bi Mo O (010) surface during a 144-h treatment at 500 C in air–2.3%
2
3 12
H O. The precipitates developed into an overlayer within ∼ 250 h.
2
XRD analysis of these overlayers demonstrated that the precipitates were
γ -Bi MoO .
2
6
area) of the surfaces within 72 h. Enrichment was, however,
usually evident near the edges of the crystals. The average
size of the precipitates increased with reaction time, as did
the fraction of the surface they covered, until a virtually
continuous precipitate overlayer/coating developed within
approximately 250 h at 500 ◦C. The overlayers were fragile
and often delaminated and crumbled on cooling, which made
it difficult to measure their thickness, but they were at most
10 µm thick (after 350 h).
◦
ogy was more regular at 600 C, as the edges of islands,
pits, and receding steps were smooth, with an almost cir-
cular morphology. Finally, flat (010) terraces were etched
◦
to a greater extent at 600 C, both in the presence and ab-
sence of H2O. Terraces treated in dry air (see ◦Figs. 6a and
7a) were uniformly modified after 2 h at 600 C and were
characterized by half-unit-cell step loops (pits and islands)
and deposits with heights of up to 100 Å. Surfaces treated
in air–2.3% H2O exhibited more variability. The region in
Fig. 6b is characteristic of terraces that eroded extensively
over 2 h in air–2.3% H2O. In other areas (see Fig. 7b), fewer
pits and deposits were encountered, often at concentrations
comparable to those observed at 500 ◦C.
Micrometer-sized precipitates were apparent after only
2 h at 600 ◦C on Bi2Mo3O12(010) surfaces treated in either
air–2.3% H2O or dry air. The precipitates were uniformly
distributed across the (010) surfaces, although enrichment
near the edges of the crystals was again evident. The precip-
itates that developed in dry air were characteristically acic-
ular and had lengths of up to 30 µm after 2 h (see Fig. 7a).
The precipitates that formed in air–2.3% H2O (see Fig. 7b)
were more equiaxed and usually had lateral dimensions in
excess of 150 µm. The image in Fig. 7b shows only one
corner of a trapezoid-shaped precipitate that was more than
200 µm long. Irrespective of the atmosphere, XPS analy-
sis indicated elevated Bi concentrations at Bi2Mo3O12(010)
surfaces treated for 2 h at 600 ◦C. Surfaces treated in dry air
X-ray diffraction (XRD) analysis of samples scraped
from single crystal (010) surfaces treated for 200 and 350 h
◦
at 500 C in air–2.3% H2O revealed the presence of two
crystalline phases, Bi2Mo3O12 and γ -Bi2MoO6. Given the
fact that micromete◦r-sized Bi2MoO6 precipitates developed
within 48 h at 500 C in air–2.3% H2O, it is not surprising
that XPS indicated enhanced Bi concentrations at (010)
surfaces treated for or in excess of 48 h. For example,
surfaces treated for 72 h exhibited Bi/Mo ratios of 0.85–
0.96. It should, of course, be kept in mind that these
XPS data only provide an indication of the average near-
surface composition. Obviously, the local Bi/Mo ratio in
regions with large Bi2MoO6 precipitates (as in Fig. 5) would
be expected to exceed that in regions free of precipitates
(Fig. 4c).
Parallel experiments with Bi2Mo3O12 powders yielded
results consistent with the single crystal studies. After 200 h
◦
at 500 C in air–2.3% H2O, three phases were detected in
powder beds with XRD, Bi2Mo3O12 and small amounts of
Bi2MoO6 and MoO3. During powder experiments, MoO3
crystals were also deposited on the walls of the reaction

S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
159
◦
Fig. 6. Topographic AFM images of terraced regions on Bi Mo O (010) surfaces treated for 2 h at 600 C in dry air (a) and air–2.3% H O (b). The
2
3
12
2
black-to-white contrasts in (a) and (b) correspond to topographic ranges of 110 and 60 Å, respectively.
exhibited Bi/Mo ratios of 0.73–0.77, while those treated in
air–2.3% H2O had Bi/Mo ratios of 0.79–0.86. The precipi-
tates grew rapidly in air–2.3% H2O, and an overlayer devel-
oped within 24 h. XRD analysis of these overlayers revealed
that the precipitate phase was β-Bi2Mo2O9. This assignment
was also supported by parallel powder experiments. After
24 h at 600 ◦C in air–2.3% H2O, Bi2Mo2O9 was easily de-
tected with XRD. MoO3 crystals were deposited at the reac-
tor exhaust during both single crystal and powder treatments
in excess of 24 h.
Precipitate growth was slow in dry air, and
Bi2Mo3O12(010) surfaces treated for up to 100 h at 600 ◦C
did not yield enough of the precipitate phase for it to be iden-
tified with XRD. Similarly, phase changes were not detected
in Bi2Mo3O12 powders treated for up to 100 h at 600 ◦C in
dry air. Energy dispersive X-ray analysis performed within
◦
Fig. 7. AFM deflection images of precipitates formed on Bi Mo O (010) surfaces during 2-h anneals at 600 C in dry air (a) and air–2.3% H O (b).
2
3
12
2
In air–2.3% H O, the precipitates developed into an overlayer within ∼ 24 h. XRD analysis of these overlayers demonstrated that the precipitates were
2
β-Bi Mo O .
2
2 9

160
S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
a scanning electron microscope (SEM) indicated that the
precipitates had a Bi/Mo stoichiometry close to 1/1, but
the possibility that the underlying Bi2Mo3O12 matrix con-
tributed to the measured composition could not be ruled out.
Hence, the precipitates were enriched in Bi but not necessar-
ily Bi2Mo2O9.
600 ◦C), as are other significant changes in the structure and
chemistry of the Bi2Mo3O12(010) surface.
◦
Within 2 h at or above 500 C, the morphology of the
Bi2Mo3O12(010) surface is altered as a result of incongruent
etching reactions that proceed in dry air and air–2.3% H2O.
In both atmospheres, etching is marked by the removal of
material (step recession and/or surface pitting) and the ac-
cumulation of surface deposits. Given the dissimilar volatil-
ities of Mo and Bi, surface pitting and step recession can
be attributed to local Mo volatilization and the concomitant
decomposition of Bi2Mo3O12. Since Bi volatility is essen-
tially negligible, decomposition/etching leaves the surface
enriched in Bi, as verified with XPS. Initially, the resid-
ual Bi is likely present as surface adatoms, but it eventu-
ally condenses with O (and perhaps some Mo) to form the
observed deposits. This mechanism is certainly consistent
with the close spatial association between deposits and re-
ceding steps and pits on surfaces treated in both air–2.3%
H2O (Figs. 4c, d, and 6b) and dry air (Figs. 4b, 6a, and 7a).
While the deposits are undoubtedly enriched in Bi, their ex-
act chemical makeup is unclear. They may well be Bi2MoO6
or Bi2Mo2O9 nuclei (depending on the temperature), but
they could also be a metastable intermediate, which then re-
acts with Bi2Mo3O12 or the Mo-rich vapor to form the pre-
cipitates.
One would expect there to be mechanistic differences
between etching in the presence and absence of H2O.
For example, Mo volatilization is expected to proceed via
MoO3 in dry air, but as MoO2(OH)2 in air–2.3% H2O
[23,25]. This might in part account for the fact that etching
primarily affects steps in air–2.3% H2O, while both steps
and terraces are etched in dry air. The formation and
desorption (volatilization) of MoO2(OH)2 must presumably
be preceded by H2O adsorption, whereas MoO3 desorption
can proceed independent of an adsorption step. Therefore,
the specificity of etching in air–2.3% H2O might reflect
a structural sensitivity of H2O chemisorption. However,
both MoO3 and MoO2(OH)2 desorption should be favored
at steps over terraces, since this requires breaking fewer
bonds and circumvents the energy barrier associated with
the nucleation of an isolated pit on a singular terrace (i.e.,
two-dimensional or layer-by-layer evaporation) [40–45].
The fact that flat (010) terraces are uniformly etched in
dry air indicates that, at some point during the treatment,
the vapor pressure of MoO3 in the reactor is below that
required for the onset of layer-by-layer evaporation [40–45].
Rather than a sustained nonspecific (layer-by-layer) process,
etching could just proceed in a nonspecific manner early in
the treatment. As the Bi2Mo3O12 samples are first heated
to the treatment temperature, the driving force (i.e., vapor
undersaturation) for evaporation and pit nucleation is at
its highest, since the initial vapor pressures of MoO3 and
Bi2O3 in the reactor are effectively zero. As the treatment
progresses, the vapor pressures will rise and the driving force
will be reduced, potentially to the point that pit nucleation
is no longer feasible. Such a transition—from layer-by-
4. Discussion
The results of our AFM, XRD, and XPS experiments
clearly demonstrate that Bi2Mo3O12 loses Mo through
volatilization above 500 ◦C in dry air and air–2.3% H2O. Not
surprisingly, this volatility leads to changes in the structure
and chemistry of the (010) surface. The most conspicuous
consequence of Mo loss is the precipitation of new Bi-rich
bismuth molybdates, with Bi2MoO6 and Bi2Mo2O9 forming
in air–2.3% H2O at 500 ◦C and 600 ◦C, respectively. Though
Mo volatilization proceeds more slowly in dry air, Bi-rich
precipitates are formed within only 2 h at 600 ◦C. The obser-
vation that H2O enhances Mo volatility is of course not sur-
prising, as it has been linked to Mo volatilization from Mo-
containing alloys, silicides, and oxides alike [22–30]. This
is usually attributed to the formation of volatile MoO2(OH)2
(g) through direct reaction with H2O [22–30]. Although the
present study has addressed more moderate temperatures
and environmental conditions, the results are fully consistent
with those of Zhang and co-workers [25], which showed that
Bi volatilization from Bi2Mo3O12 (as well as Bi2MoO6 and
Bi2Mo2O9) powders was negligible in dry (N2) and humid
(H2O) atmospheres below ∼ 700 ◦C and that Mo volatiliza-
tion was accelerated by H2O.
That volatile Mo loss from Bi2Mo3O12 in air–2.3% H2O
results in the precipitation of Bi2MoO6 at 500 ◦C and
Bi2Mo2O9 at 600 ◦C is in agreement with the Bi2O3–MoO3
phase diagram proposed by Egashira and co-workers [9].
The formation of these phases indicates that Bi2Mo3O12 is
unstable under the conditions examined, undoubtedly due
to the disparate volatilities of Mo and Bi. Our conclusion
that Bi2Mo3O12 is unstable is not necessarily inconsistent
with the results of prior bulk studies, even though these
studies have generally concluded that the phase is stable
in air and under typical reaction conditions [8–17]. This
is because the precipitates we observed on single crystal
surfaces at comparable temperatures and over comparable
durations (usually less than 100 h) would not have been
detected in powders with bulk structural probes, such as
XRD and Raman and infrared spectroscopies [8–17]. For
example, Snyder and Hill [8] did not detect phase changes in
Bi2Mo3O12 powders with in situ Raman spectroscopy over
24 h at 600 ◦C in dry air. Similarly, we detected no changes
with XRD over 100 h under the same conditions. Only
through protracted anneals can sufficient volume fractions
of the second phases be obtained to enable identification
with XRD. Of course, decomposition and precipitation are
initiated over much shorter periods (e.g., within 2 h at

S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
161
layer to step/defect-controlled evaporation—could explain
the transition from small islands and pits of variable (b/4–
b/2) height (Fig. 3a) to uniform half-unit-cell (b/2) islands
(Figs. 4a and 4b) during treatments at 500 ◦C. Subsequently,
the half-unit-cell islands must in fact erode preferentially
from their bounding steps, since new surface pits are only
rarely observed at long exposures. The change in surface
morphology that accompanies the transition is probably
indicative of coarsening of the initial (transient) morphology
(∼ 2 h, Fig. 3a) to a lower energy configuration (∼ 72 h,
Fig. 4a).
The rarity of terrace etching in air–2.3% H2O indicates
that the vapor pressure of MoO3, which is related to the
vapor pressure of MoO2(OH)2, is sufficiently high so that
layer-by-layer evaporation is largely suppressed [40–45].
This behavior is somewhat counterintuitive because Mo
volatilizes more rapidly in air–2.3% H2O than in dry air, in-
dicating that the driving force (vapor undersaturation) for pit
nucleation should be greater in air–2.3% H2O. The differ-
ence could however be explained if H2O disproportionately
enhances the rate at which Mo can volatilize from steps com-
pared to terraces, for example by lowering the activation en-
ergy for desorption. Such a mechanism is consistent with
the fact that steps are etched in air–2.3% H2O at as low as
400 ◦C. Rapid volatilization from steps, defects (e.g., dislo-
cations), and the crystal edges as the samples are heated to
the treatment temperature could raise the pressure of MoO3
to the point that layer-by-layer evaporation is not feasible.
As the treatment temperature is increased, the vapor under-
saturation required for layer-by-layer evaporation should de-
crease, consistent with the increased occurrence of pits on
singular terraces in air–2.3% H2O at 600 ◦C.
(010) surfaces would be expected to be controlled by de-
fects, such as steps and oxygen vacancies.
Acknowledgments
This work was supported by the Department of Materials
Science and Engineering of the Massachusetts Institute of
Technology (MIT). The authors gratefully acknowledge the
technical assistance of Mr. J. Chen, Mr. R. Li, and Ms. L.
Metzger and the use of the shared experimental facilities
of the National Science Foundation—Materials Research
Science and Engineering Center (NSF-MRSEC) at MIT.
References
[1] J. Haber, in: R. Grasselli, J. Brazdil (Eds.), Solid State Chemistry in
Catalysis, American Chemical Society, Washington, DC, 1985, p. 1.
[2] D.A. Jefferson, J.M. Thomas, M.K. Uppal, R.K. Grasselli, J. Chem.
Soc. Chem. Commun. 594 (1983).
[3] A.W. Sleight, in: A.K. Cheetham, P. Day (Eds.), Solid State Chemistry
Compounds, Clarendon Press, Oxford, 1992, p. 166.
[4] D.J. Buttrey, D.A. Jefferson, J.M. Thomas, Philos. Mag. A 53 (1986)
897.
[5] P.N. Rylander, in: J.R. Anderson, M. Boudart (Eds.), Catalysis Science
and Technology, Springer, New York, 1983, p. 2.
[6] T.P. Snyder, C.G. Hill Jr., Catal. Rev. Sci. Eng. 31 (1989) 43.
[7] N. Arora, G. Deo, I.E. Wachs, A.M. Hirt, J. Catal. 159 (1996) 1.
[8] T.P. Snyder, C.G. Hill Jr., J. Catal. 132 (1991) 536.
[9] M. Egashira, K. Matsuo, S. Kagawa, T. Seiyama, J. Catal. 58 (1979)
409.
[10] L.D. Krenzke, G.W. Keulks, J. Catal. 64 (1980) 295.
[11] J.F. Brazdil, D.D. Suresh, R.K. Grasselli, J. Catal. 66 (1980) 347.
[12] A.F. Van Den Elzen, G.D. Rieck, Acta Crystallogr. B 29 (1973) 2433.
[13] P.L. Gai, J. Solid State Chem. 49 (1983) 25.
Although there are differences in the manner in which
the Bi2Mo3O12(010) surface is etched in dry air and air–
2.3% H2O, the surface evolves to have similar morphologi-
cal characteristics in the two environments. For example, Mo
loss results in Bi enrichment and the development of Bi-rich
deposits. While steps on the cleavage surface are characteris-
tically straight, steps on treated surfaces are curved. Further-
more, steps are always an integer multiple of b/2 in height.
Terraces bounded by smaller steps, resulting either from
cleavage (Fig. 1b) or short exposu◦res at 500 ◦C (Fig. 3a), are
annealed out within 72 h at 500 C. Recalling the symme-
try and bonding arguments outlined with regard to cleavage
(Section 3.1), this points to a consistent Bi2Mo3O12(010)
termination. All the terraces on (010) will have the same
ideal termination, while those on (010) will be a mirror im-
age. We assume this is the same termination favored by 1/2–
1 cleavage (Fig. 2b), as it affords the lowest possible broken
bond density. Since Mo sites are generally thought to play a
role in reactant chemisorption on bismuth molybdates, it is
interesting to note that this terrace termination presents Mo
that are tetrahedrally coordinated by O and effectively bond
saturated. Hence, if coordinatively unsaturated Mo atoms are
required for chemisorption, the reactivities of the (010) and
[14] P.L. Gai, Catal. Rev. Sci. Eng. 34 (1992) 1.
[15] I. Matsuura, R. Schut, K. Hirakawa, J. Catal. 63 (1980) 152.
[16] D.A.G. Van Oeffelen, J.H.C. Van Hooff, G.C.A. Schuit, J. Catal. 95
(1985) 84.
[17] J. Kumar, E. Ruckenstein, J. Solid State Chem. 31 (1980) 41.
[18] D. Carson, G. Coundurier, M. Forissier, J.C. Vedrine, J. Chem. Soc.
Faraday Trans. 79 (1983) 1921.
[19] L.E. Briand, A.M. Hirt, I.E. Wachs, J. Catal. 202 (2001) 268.
[20] I.E. Wachs, Catal. Today 27 (1996) 437.
[21] C.-B. Wang, Y. Cai, I.E. Wachs, Langmuir 15 (1999) 1223.
[22] T. Millner, J. Neugebauer, Nature 163 (1949) 601.
[23] G.R. Belton, A.S. Jordan, J. Phys. Chem. 69 (1965) 2065.
[24] T. Lim, S.W. Weller, J. Catal. 108 (1987) 175.
[25] L. Zhang, D. Liu, B. Yang, J. Zhao, Appl. Catal. A Gen. 117 (1994)
163.
[26] M.K. Meyer, A.J. Thom, M. Akinc, Intermetallics 7 (1999) 153.
[27] M. Akinc, M.K. Meyer, M.J. Kramer, A.J. Thom, J.J. Huebsch, B.
Cook, Mater. Sci. Eng. A 261 (1999) 16.
[28] R.L. Smith, G.S. Rohrer, J. Catal. 163 (1996) 12.
[29] R.L. Smith, G.S. Rohrer, J. Catal. 180 (1998) 270.
[30] R.L. Smith, G.S. Rohrer, J. Catal. 184 (1999) 49.
[31] T. Chen, J. Cryst. Growth 20 (1973) 29.
[32] A. Grandin De l’eprevier, D.A. Payne, J. Cryst. Growth 51 (1981) 377.
[33] R.L. Smith, G.S. Rohrer, J. Solid State Chem. 124 (1996) 104.
[34] R.L. Smith, G.S. Rohrer, K.S. Lee, D.-K. Seo, M.-H. Whangbo, Surf.
Sci. 367 (1996) 87.
[35] G.S. Rohrer, W. Lu, R.L. Smith, A. Hutchinson, Surf. Sci. 292 (1993)
261.

162
S.V. Yanina, R.L. Smith / Journal of Catalysis 213 (2003) 151–162
[36] T. Oshio, Y. Sakai, S. Ehara, J. Vac. Sci. Technol. B 12 (1994) 2055.
[37] C. Noguera, Physics and Chemistry at Oxide Surfaces, Cambridge
Univ. Press, Cambridge, UK, 1996.
[38] C.F. McFadden, P.S. Cremer, A.J. Gellman, Langmuir 12 (1996) 2483.
[39] R.L. Smith, G.S. Rohrer, in: D.A. Bonnell (Ed.), Scanning Tunneling
Microscopy and Spectroscopy: Theory Techniques and Applications,
2nd ed., VCH, New York, 2001.
[40] W.K. Burton, N. Cabrera, Discuss. Faraday Soc. 5 (1949) 33.
[41] W.K. Burton, N. Cabrera, Discuss. Faraday Soc. 5 (1949) 40.
[42] F.C. Frank, Discuss. Faraday Soc. 5 (1949) 48.
[43] J.P. Hirth, Acta Metall. 7 (1959) 755.
[44] J.B. Hudson, Surface Science: An Introduction, Wiley, New York,
1992.
[45] W.W. Mullins, G.S. Rohrer, J. Am. Ceram. Soc. 83 (2000) 214.