XPS study of the size effect in ethene epoxidation on supported silver catalysts

Article abstract of DOI:10.1039/a608414a

Supported silver catalysts (Ag/α-Al₂O₃) with different particle sizes (100-1000 A?) have been prepared and studied by XPS. It has been shown that the increase in the ethene epoxidation rate with silver particle size (size effect) is

Full text of DOI:10.1039/a608414a

View Article Online / Journal Homepage / Table of Contents for this issue
XPS study of the size e†ect in ethene epoxidation on supported silver
catalysts
Valery I. Bukhtiyarov,*¤ Igor P. Prosvirin, Ren I. Kvon, Svetlana N. Goncharova and
Bair S. BalÏzhinimaev
Boreskov Institute of Catalysis, L avrentieva prosp., 5, Novosibirsk, 630090, Russia
Supported silver catalysts (Ag/a-Al O ) with di†erent particle sizes (100È1000 Ó) have been prepared and studied by XPS. It has
2
3
been shown that the increase in the ethene epoxidation rate with silver particle size (size e†ect) is accompanied by a decrease in
the Ag 3d
binding energy value. The variation in E (Ag 3d ) cannot be explained in terms of a change in the metallic
5@2
b
5@2
character of the silver (initial state e†ect), but seems to be determined by the di†erential charging of the supported silver. The
latter e†ect originates from the easier screening of surface charge induced by photoemission on the silver particles owing to their
internal conductivity, rather than that on the dielectric support. The correlation between the internal conductivity of supported
silver particles and their catalytic activity is discussed in the context of possible mechanisms for ethene epoxidation.
particle size is correlated with the speciÐc catalytic activity of
these catalysts. Furthermore, we will present data on the inÑu-
Introduction
Studying small silver particles supported on a-alumina is of
great importance because of their uniqueness as catalysts for
epoxidation of ethene. Decreasing the Ag particle size has an
economic incentive, but one must take into account the exis-
tence of a size e†ect on the catalytic activity of the small sup-
ported particles.
ence of caesium, which is the main promoter in commercial
silver catalysts.
Experimental
A VG Escalab high-pressure electron spectrometer was used
to carry out this investigation. Before beginning the experi-
ments, the spectrometer was calibrated against E (Au 4f ) \
The dependence of catalytic activity on the sizes of metal
particles (size e†ect) is one of the main features of supported
catalysts, and, therefore, there have been a large number of
investigations devoted to the study of this problem.1h10 It has
been shown that the rates of a great number of catalytic reac-
tions change at very small sizes of supported particles (\20È
50 Ó).1h4 A number of authors have tried to explain such
behaviour by the metalÈnon-metal transition in small metal
clusters. Indeed, physical methods of analysis such as XPS,
UPS, STM and EXAFS indicate that the deviation of di†erent
spectral characteristics from those of the bulk are observed for
sizes of metal particles \50 Ó.2,5h7 A few studies have
revealed a correlation between the deviation of the electronic
properties of the supported metals from bulk ones and their
catalytic activity.2,8h10
Amongst the metal supported catalysts, silver occupies a
unique position. The change of its catalytic properties takes
place at much larger sizes of the supported metal particles
than is usually observed.11h13 Despite some discrepancies one
can state that the sharp increase in the speciÐc catalytic activ-
ity of the silver catalysts occurs at sizes above 100È500 Ó. It
seems that these changes in epoxidation rate can not be
caused by a metalÈnon-metal transition. Indeed, this tran-
sition occurs for silver particles within a typical size range as
for other metals.5h7 One must therefore suppose other reasons
cause the observed variation in the catalytic properties of sup-
ported silver. To study this problem, we have used X-ray
photoelectron spectroscopy (XPS) which has proven to be a
very useful technique in the characterization of supported par-
ticles.
b
7@2
84.0 eV and E (Cu 2p ) \ 932.6 eV.
3@2
Ag/Al O supported catalysts with various silver particle
b
2
3
sizes and an Ag(111) single crystal were used as samples in this
study. To prepare the Ag/Al O model catalysts we used a-
2
3
alumina with a speciÐc area of 7 m2 g~1 and a solution of a
silver complex with ethanolamine having a molar ratio
Ag : NH C H OH \ 1 : 4 prepared by adding ethanolamine
2 2
4
to an aqueous solution of silver nitrate.14 Impregnation of the
a-alumina with silver amine complexes followed by drying,
reduction of the silver particles at T \ 360 K in vacuo and
calcination at T \ 510 K led to the formation of metallic
silver particles. This conclusion is based on X-ray synchrotron
di†raction radiation (XRSDR) and transmission electron
microscopy (TEM) data taken for the same catalysts which
indicate that the phase composition of all samples corre-
sponds to a-alumina and metallic silver.15 Cs-promoted cata-
lysts were prepared by impregnating the supported Ag
samples with Cs solution followed by drying at 360 K in vacuo
and calcination at 510 K. This catalyst preparation method
has been described in more detail previously.14 It should be
noted that to avoid undesired impurities, all initial reagents
for the preparation of catalysts and supports were of special
purity grade or synthesized under conditions to exclude
sample contamination. In consequence, XPS survey spectra of
our samples contained only lines characteristic of the support
(aluminium and oxygen), silver and carbon. A typical survey
spectrum of a supported catalyst is shown in Fig. 1 (curve 1).
Variation in the Ag particle size from ca. 100 to 1000 Ó was
achieved by increasing the amount of supported silver by
repeating the impregnation procedure. The Ag content was
determined by atomic absorption and was varied from 0.4 to
13.8 mass%. The size of silver crystallites was determined both
by oxygen chemisorption and by transmission electron
microscopy. The histograms of these catalysts had a variation
We report an XPS study of Ag/Al O catalysts prepared
2
3
with a narrow size distribution of the supported silver par-
ticles and both valence-band and core-level spectra are
analysed. The variation of XP spectral characteristics with
¤ E-mail: vib=catalysis.nsk.su
J. Chem. Soc., Faraday T rans., 1997, 93(13), 2323È2329
2323

View Article Online
Fig. 1 Survey spectra of Ag/Al O supported catalyst (1) and Ag(111) single crystal (2)
2
3
in particle size of \40%, i.e. a reasonably uniform size dis-
tribution was observed.14,15
and T \ 470 K and subsequent annealing in vacuo up to 970
K. After a few repetitions of this cycle, no contamination was
registered in the survey XPS spectra (Fig. 1, curve 2) and the
surface was characterized by a sharp (1 ] 1) low-energy elec-
tron di†raction (LEED) pattern.
The rate of ethylene oxide formation was measured by the
Ñow-recirculation method at P \ 1 atm and T \ 500 K. The
initial reaction mixture consisted of 2% ethene, 7% oxygen
and nitrogen as the balance. Table 1 presents results on cata-
lytic testing of Ag/Al O catalysts taken from the ref. 14,
Results
Fresh unpromoted catalysts
2
3
where the procedure of the catalytic activity measurement and
all characteristics for the synthesized catalysts were described
thoroughly. Here, we use only those data which demonstrate
the inÑuence of the mean particle size on the rate of ethene
epoxidation and selectivity towards the key product, ethylene
oxide. It should be noted that both the activity and the selec-
tivity change with time. Activation of the catalysts occurred
within a few hours, and afterwards a steady state was attained.
We have studied both the fresh (before testing their catalytic
properties) and activated (after attainment of the steady state)
catalysts.
Fig. 2 shows valence band spectra of silver and an Ag 3d
core level spectra for three freshly prepared catalysts with
5@2
mean sizes of supported Ag particles of 160, 560 and 1000 Ó.
These sizes have been chosen in accordance with the catalytic
data presented in Table 1, which indicate that a sharp increase
in speciÐc catalytic activity by more than an order of magni-
tude occurs for these catalysts in the size range 300È500. Here-
after, we use the terms small, medium and large for Ag
particles with sizes of 160, 560 and 1000 Ó, respectively. The
valence band of the pure support, shown for comparison, is
superimposed on the silver signal, and therefore it has been
subtracted from the catalyst spectra in order to obtain the
valence band of the supported silver.
One can see that increasing the sizes of the Ag particles
from 160 to 1000 Ó causes changes in the silver valence-band
spectra. These are determined by the variation in the silver
concentration for the di†erent samples: from 0.5 to 13.8
mass%, respectively. Unfortunately, the silver mass concentra-
tion can not be used to recalibrate the spectra, since most of
silver atoms do not contribute to the XPS signal. This state-
ment is based on the fact that the XPS analysis depth (20È30
Ó) is much smaller than the particle sizes used in this study
([100 Ó). To tackle this problem we have used the relative
All samples of the supported catalysts were ground and
mounted on the standard holder by means of double-sided
adhesive tape. Al-Ka radiation (hl \ 1486.6 eV) was used to
record both valence-band and core-level spectra. To calculate
the charging e†ect for non-conductive samples such as
Ag/Al O we used a method of internal standardization, the
2
3
Al 2p line being taken as an internal reference with a binding
energy of 74.5 eV.16
The Ag(111) single crystal (Monocrystal Co., USA) with a
diameter of 12 mm and thickness of 2 mm was welded to two
thin tungsten wires that allowed us to use a heating rate of 2
K s~1. The temperature was measured with a PtÈPt/Rh ther-
mocouple welded to the edge of the crystal. The Ag(111)
crystal was cleaned by a standard high-vacuum cycle: ionic
etching followed by O adsorption for 10 min at P \ 10 Pa
2
intensities of the core-level Ag 3d
same Figure. It is seen that the Ag 3d
sample with small Ag particles is six and twelve times less
than that for the medium and large particles, respectively.
spectra presented in the
5@2
intensity for the
5@2
Table 1 Variation of the rates of ethene epoxidation [R(C H O)]
2
4
and selectivity (S) towards ethylene oxide with mean silver particle
size [d(Ag)] for unpromoted Ag/Al O supported catalysts
In Fig. 3 the valence-band and Ag 3d spectra of the sup-
2
3
5@2
ported silver with small and medium particles are compared
d(Ag)/Ó
R(C H O)/1017 molecules m~2 s~1
S (%)
2
4
(the spectra for the small particles being multiplied by a factor
of six). The 560 Ó silver particles are large enough to lead to
an increase in the rate of ethene epoxidation (Table 1). One
can see that both the shape and width of the silver valence
bands for both samples are identical. The position and full
160
220
400
560
1000
0.18
0.21
2.0
6.3
4.8
7
8
22
47
41
width at half maximum (FWHM) of the Ag 3d spectra also
5@2
2324
J. Chem. Soc., Faraday T rans., 1997, V ol. 93

View Article Online
Fig. 2 Valence-band di†erence spectra (top) and Ag 3d core level
5@2
spectra (bottom) of fresh supported Ag catalysts with di†erent mean
sizes of metal particles. (1) 160 Ó, (2) 560 Ó and (3) 1000 Ó. Also
shown as a dashed curve is the valance band spectrum from clean
Al O .
Fig. 3 Valence-band di†erence spectra (top) and Ag 3d
core-level
5@2
spectra (bottom) of fresh supported Ag catalysts with mean sizes of
metal particles of 160 Ó (È È È) and 560 Ó (ÈÈ). The spectra of the
former is multiplied by a factor of six.
2
3
coincide for these fresh catalysts. It should be noted however
that the value of the Ag 3d binding energy for the sup-
5@2
ported silver (367.9 eV) in the samples is less than the corre-
sponding value for bulk silver (368.2 eV).16,17 The same value
of 368.2 eV was measured by us for the Ag(111) single crystal.
sample with large particles and so its Ag 3d spectra are not
5@2
discussed.
Thus, the data for the activated samples (Fig. 4) indicate
that the activation changes the position of the Ag 3d
spectra, the nature of these changes depends on the silver par-
The coincidence of Ag 3d core-level binding energies for the
5@2
fresh catalysts with various particle sizes seems to indicate
5@2
that no correlation is observed between XPS data and cata-
lytic properties of these catalysts in the ethene epoxidation
reaction. It should, however, be recalled that the measurement
of the reaction rate is carried out after reaching the steady
state, i.e. after some activation of the fresh catalysts under
reaction conditions.
Activated unpromoted catalysts
Ag 3d
spectra for the activated samples with small and
5@2
medium silver particles are shown in Fig. 4 (dashed curves),
the corresponding spectra of the fresh samples are also shown
for comparison (solid curves). After activation, a shift of the
Ag 3d spectrum to higher E (up to 368.3 eV) is observed
5@2
b
for catalysts with small particles. The FWHM of the spectra
remain unchanged, consistent with a shift of the whole core-
level line. The increase in E for Ag 3d
above the value
b
5@2
characteristic for metallic silver (368.2 eV) conÐrms that the
lower binding energies observed for the fresh catalysts are
genuine, and are not caused by a mistake in the reference level
value used [E (Al 2p) \ 74.5 eV]. Distinctive behaviour is
b
revealed for the catalyst with medium Ag particles: activation
Fig. 4 Ag 3d core-level spectra of supported Ag catalysts with dif-
5@2
does not change the Ag 3d spectrum (Fig. 4). Similarly, no
changes were observed in the spectral characteristics of the
ferent mean sizes of metal particles measured before (ÈÈ) and after
5@2
(È È È) activation. (1) 160 Ó and (2) 560 Ó.
J. Chem. Soc., Faraday T rans., 1997, V ol. 93
2325

View Article Online
more pronounced size e†ect in ethene epoxidation than pre-
viously reported in the literature.11h13 It should be noted that
the decrease in catalytic activity with decrease in particle size
is accompanied by decrease in the selectivity towards ethylene
oxide (Table 1). This phenomenon is explained by the fact that
the rate of ethylene oxide formation on the catalysts with
small silver particles becomes comparable with the rate of sec-
ondary oxidation of ethene on the support surface.18 The
latter process together with direct total oxidation is
responsible for the formation of CO , a byproduct of ethene
2
oxidation. This suggestion is conÐrmed by an increase in the
selectivity for the catalysts with small silver particles up to
25È30% when the alumina support is replaced by silica.14
These observations indicate that an analysis of the reasons
causing the size e†ect is of great signiÐcance for silver cata-
lysts.
Fig. 5 Variation of ethene epoxidation rate (@) and E (Ag 3d ) (K)
b
5@2
with mean sizes of metal particles measured for unpromoted Ag cata-
lysts
The correlation between the rate of ethene epoxidation and
the value of core-electron level binding energy seems to indi-
cate the existence of a correlation between the catalytic and
electronic properties of these catalysts and their variations
with silver particle size. The particle size at which the elec-
tronic properties of small metal clusters become similar to
those of the bulk metal has been discussed extensively, with
silver having been the subject of a number of these reports. In
particular, it has been shown that like other metals the band-
width of the photoemission spectrum of the supported silver
reaches its bulk value for Ag coverages of H \ 2 ] 1019 atoms
m~2 (i.e. particle size [20 Ó).5,6 Similar conclusions about
the critical size of Ag particles were made by Juska et al.19
who analysed changes in the modiÐed Auger parameter upon
high-vacuum Ag deposition on an alumina surface. Our data
for fresh Ag supported catalysts are in agreement with this.
Indeed, starting with Ag particles [100 Ó in size, we have
observed no variations either in the width of the Ag valence
ticle size. This conclusion is conÐrmed by the data of Fig. 5
which shows the dependence of E (Ag 3d ) on Ag particle
b
5@2
size not only for the above samples, but also for other investi-
gated catalysts. One can see that there is some threshold in
the XPS characteristics of supported silver: small Ag particles
in the activated samples have a much higher E (Ag 3d
values than the larger ones. We have juxtaposed the values of
)
b
5@2
the ethene epoxidation reaction rate and E (Ag 3d ) for
b
5@2
several activated catalysts in Fig. 5. One can see a very good
correlation between the catalytic activity and the XPS data:
the lower the value of E (Ag 3d ), the higher the rate of
b
5@2
ethylene oxide formation.
Cs promoted catalysts
Analogous comparisons of the rate of ethene epoxidation and
the binding energies of Ag 3d vs. caesium content for Cs
band or in the position or FWHM of the Ag 3d
(Fig. 3), i.e. all the fresh catalysts studied have identical elec-
spectra
5@2
5@2
promoted catalysts, as shown in Fig. 6, conÐrm this corre-
lation. Indeed, when the caesium is doped in quantities larger
than optimal a sharp decrease in the reaction rate is accompa-
nied by an increase in E (Ag 3d ) up to 368.3 eV. Thus, the
tronic properties.
The lower value of the Ag 3d binding energy observed for
5@2
the fresh samples (367.9 eV) compared with that for bulk silver
b
5@2
correlation between the catalytic activity of the supported
(368.2 eV) can originate either from an initial state e†ect, when
the chemical state of silver changes from metallic, or from a
di†erential charging e†ect. Indeed, workers who have studied
not only metallic silver, but also silver oxides by XPS,20,21
have demonstrated that, in contrast to many other metals, the
silver catalysts and the value of E (Ag 3d ) seems to be
b
5@2
general. However, before using this correlation it is necessary
to understand what determines a decrease or an increase in
the Ag 3d
binding energies and, especially, their variation
5@2
with Ag particle sizes (Fig. 5). The inÑuence of caesium on the
adsorptive and catalytic properties of silver will be discussed
elsewhere.
oxidation of silver atoms decreases the Ag 3d
5@2
energy value: the E (Ag 3d ) values for Ag O and AgO are
5@2
binding
b
2
367.7 and 367.4 eV, respectively. On the other hand, metal
crystallites of between 50 and several hundred Ó can exhibit
the solid-state properties of a conductive metal that should
provide easier compensation of the charge on their surface
than that on the surface of a dielectric support. In conse-
quence, the charging potential (/) induced by photoemission
will be less for the XPS lines of a metal compared with the
charging potential on a dielectric support and so-called di†er-
ential charging of a more conducting phase compared with a
non-conducting one will appear; the XPS spectra of a metal
will be negatively shifted by di†erential charging. The exis-
tence of di†erential charging for metal crystallites supported
on a dielectric support has been revealed and thoroughly
analysed by Barr22,23 who has shown that this shift can reach
1 eV. It is obvious that we have such a situation for our
Ag/Al O samples.
Discussion
The preparation of Ag/Al O supported catalysts with a
narrow particle size distribution has allowed us to measure a
2
3
2
3
XPS parameters which do not depend on charging should
be used to measure its contribution (*E ) in the total varia-
F
tion of E . One such methods is the application of a modiÐed
b
Auger parameter that represents the sum of a core-level
binding energy and the kinetic energy of the corresponding
Auger peak.23h26 Since the charging potential increases the
core level binding energy and decreases the Auger kinetic
energy by the same value, their sum is not changed with
Fig. 6 Variation of ethene epoxidation rate (@) and E (Ag 3d ) (K)
b
5@2
with Cs concentration measured for promoted Ag catalyst with 560 Ó
metal particles
2326
J. Chem. Soc., Faraday T rans., 1997, V ol. 93

View Article Online
charging. This parameter for supported silver given by a \
E (Ag 3d ) ] E (Ag MNN). It should be also noted that in
b
5@2
k
the case of silver the Auger parameter is very sensitive to the
chemical state of silver: changing from metallic silver to its
compounds is accompanied by a change of a of ca. 2
eV.16,24,25
The corresponding Ag MNN Auger spectra of supported
Ag for the fresh samples are centred at E \ 358.4 eV which
k
gives a value of a of 726.3 eV. Comparison of this value with
the known data for metallic Ag (a \ 726.0È726.3 eV) and for
Ag oxides and other silver compounds (a \ 724.0È724.5
eV)16,24,25 indicates that the fresh samples contain metallic
silver and, hence, the lower Ag 3d binding energies are most
5@2
likely to originate from the di†erential charging e†ect.
Further conÐrmation of silver di†erential charging has been
obtained by means of valence-band spectra which are sensitive
both to the initial state and to relaxation e†ects.27,28 Fig. 7
compares the valence-band spectra of supported silver in a
fresh sample and bulk Ag. For the bulk silver sample, we used
an Ag(111) single crystal that had a good electrical contact,
and, hence, electrochemical equilibrium with the spectrometer.
One can see that the supported Ag has a valence band spec-
trum very similar to that of the single crystal, except for a shift
of 0.8 eV to lower binding energies (Fig. 7). Indeed, a shift of
the Ag(111) valence band spectrum by this value results in the
coincidence of both shape and width of these spectra (Fig. 7)
conÐrming our conclusion about the metallic state of Ag in
the fresh catalysts.
The existence of di†erential charging in this system was also
suggested by Juska et al.19 who studied Ag evaporation in
vacuo onto a Al O surface. They observed that E (Ag 3d
)
2
3
b
5@2
which for small Ag coverages is higher than the bulk value,
decreases to below the latter with increasing concentration of
deposited silver and this di†erence is equal to 1.0 eV for
H(Ag) [ 1019 atoms m~2.19 Such behaviour contrasts with
that of other metals (e.g. Pd and Cu) where the binding energy
value of the metal core levels trends to the bulk value.27,29,30
Taking into also account the considerable variation of Ag
Fig. 8 Ag 3d (top) and Ag MNN (bottom) spectra of a supported
5@2
silver catalyst with Ag particle size of 160 Ó measured before (1) and
after (2) activation
3d
binding energies (up to 1 eV) observed for commercial
5@2
catalysts31 one can state that di†erential charging is a general
property of supported silver.
Then, it would be logical to suggest that the observed shift
of the Ag 3d binding energy caused by the activation of the
Auger line. In consequence, the Auger parameters coincide in
both cases, and this shows the initial state of Ag in both
samples is identical. The sensitivity of a to changes in the
relaxation e†ect (*E \ 1*a)23,24 allows us to conclude that
5@2
fresh sample with small silver particles (Fig. 4) originates also
R
2
from the di†erential charging e†ect. The Ag MNN Auger
spectra for fresh and the treated catalyst samples shown in
Fig. 8 conÐrm this. Indeed, one can see that the shift of the
XPS line (Fig. 8, top) is accompanied by the same shift of the
relaxation shows essentially no contribution to the variation
in the Ag 3d spectra. This is not surprising, since the relax-
5@2
ation shift is usually observed at much smaller sizes for sup-
ported metal particles (\20 Ó)24,25 than the sizes used here
([100 Ó). Thus, all presented data indicate that the studied
samples contain metallic silver particles, and both lowering of
E (Ag 3d ) values for the fresh samples and the shift of Ag
b
5@2
3d spectra for the treated samples with small Ag particles
5@2
arise from the variation of the di†erential charging value of
the supported silver. The variation of the di†erential charging
value most likely originates from changes in the internal con-
ductivity of the supported silver particles: the higher the inter-
nal conductivity of the Ag particle, the more e†ective is the
compensation of its surface charge induced by photoemission,
and the lower is the apparent binding energy value.
Comparison of the Ag 3d
spectra for the fresh catalyst
5@2
and Ag(111) single crystal made in Fig. 9 conÐrms the latter
statement. Indeed, after artiÐcial alignment of the Fermi levels
of both the samples, the spectrum of the single crystal with
high conductivity is centred at lower binding energy than that
of the supported silver. Thus, the increase in E (Ag 3d ) for
b
5@2
Ag particles of size \400 Ó (Fig. 5) indicates a lower internal
conductivity compared to larger particles.
The existence of some transition in the conducting proper-
ties of supported Ag particles for sizes above 200È300 Ó (Fig.
5) is also indicated from the NMR data:32h34 the so-called
Fig. 7 Valence-band spectra of supported Ag catalyst with 560 Ó
metal particles (ÈÈ) and an Ag(111) single crystal (È È È)
J. Chem. Soc., Faraday T rans., 1997, V ol. 93
2327

View Article Online
ethene epoxidation. However, additional experiments are
necessary to conÐrm this hypothesis.
Summary
We have revealed a correlation between the changes in the
catalytic activity of supported silver catalysts and E (Ag 3d
with Ag particle size. Relatively large Ag particles exhibit
lower binding energies and a higher rate of ethene epoxidation
)
b
5@2
than smaller (\500 Ó) particles. These di†erences in E (Ag
b
3d ) values are most likely to result from the variations in
5@2
di†erential charging of the supported silver which originate
from the higher conductivity of Ag particles compared to the
supporting surface. Measurements on an Ag(111) single crystal
and comparison of its XPS spectra with those of the sup-
ported silver allowed us not only to prove the di†erential
charging e†ect but also to determine its value. It is the di†er-
ential charging that separates the silver particles with di†erent
surface potential and demonstrates the change in electronic
properties of the silver with particle size. The revealed corre-
lation between the internal conductivity of supported silver
particles and their catalytic activity in ethene epoxidation is a
general phenomonen, and can be used to characterize other
silver catalysts.
Fig. 9 Ag 3d
spectra of a supported silver catalyst (2) and an
Ag(111) single crystal (1). The spectrum of Ag(111) is shown after shift-
5@2
ing by the di†erential charging value (0.8 eV) to lower E .
b
Knight shift disappears upon decreasing the Ag particle size
below 500 Ó. The disappearance of the Knight shift which
results from a hyperÐne interaction of the Ag nuclei with con-
duction electrons is revealed by drastic broadening of 109Ag
NMR lines from Ag atoms located in the surface and sub-
surface layers of Ag particles.32h35 Theoretical calculation of
the population of s-levels, their densities near the Fermi level
and the Knight shifts for the surface and subsurface layers
shows that for silver particles the inÑuence of the surface pro-
pagates unexpectedly deeply into the inner layers of the Ag
crystal (up to 200È250 Ó, or 70È80 layers from the surface).34
This behaviour is quite di†erent from that for Pt particles,
where the signal from the metal bulk was observed for con-
siderably smaller mean particle sizes.35 Therefore, both XPS
and NMR data show that the conductive properties of Ag
particles change signiÐcantly at a few 100 Ó level.
There are two possible reasons for the variation of the con-
ductivity of silver in an Ag(111) single crystal, catalysts with
large Ag particles and catalysts with small particles. One
possibility is changes of the inner silver morphology, the inÑu-
ence of which on the di†erential charging of metal crystallites
has been shown by Barr.22,23 Alternatively, the presence of
any contaminants in the supported Ag particles could also
a†ect the conductivity of silver. In particular we should take
into account oxygen because of the well known ability of
silver to dissolve a large amount of oxygen into the bulk or
The authors would like to acknowledge G. N. Kryukova and
A. L. Chuvilin for electron microscopy measurement, V. Yu.
Gavrilov for adsorption measurements and V. P. Ivanov for
SIMS data. Furthermore, V. I. Bukhtiyarov wishes to thank
A. F. Carley for valuable discussions concerning this paper.
References
1
2
3
G. C. Bond, Surf. Sci., 1985, 156, 966.
M. Che and C. O. Bennett, Adv. Catal., 1989, 36, 55.
P. Briot, A. Auroux, D. Jones and M. Primet, J. Catal., 1990, 59,
141.
4
5
6
7
B. Tardy, C. Noupa, C. Leclercq, J. C. Bertolini, A. Hoareau, M.
Treilleux, J. P. Faure and G. Nihoul, J. Catal., 1991, 129, 1.
R. C. Baetzold, M. G. Mason and J. F. Hamilton, J. Chem. Phys.,
1980, 72, 366.
G. K. Wertheim, S. B. DiCenzo and D. N. E. Buchanan, J. Phys.
Chem., 1986, 33, 5384.
C. N. R. Rao, H. K. N. Aiyer, T. Arunarkavalli and G. U. Kulk-
arni, Catal. L ett., 1994, 23, 37.
X. P. Xu and D. W. Goodman, Catal. L ett., 1994, 24, 31.
D. W. Goodman, Surf. Sci., 1994, 300, 837.
8
9
10 L. V. Nosova, M. V. Stenin, Yu. N. Nogin and Yu. A. Ryndin,
Appl. Surf. Sci., 1992, 55, 43.
11 J. K. Lee, X. E. Verykios and R. Pitchai, Appl. Catal., 1989, 50,
171.
subsurface layers.36,37 Furthermore, O is one of the com-
ponents of the reaction mixture.
Whichever of the factors, morphology or contamination, is
the more signiÐcant, the revealed correlation between the
ethene epoxidation rate and E (Ag 3d ) means that samples
12 S. Cheng and A. ClearÐeld, J. Catal., 1985, 94, 455.
13 S. R. Seyedmonir, J. K. Plischke, M. A. Vannice and H. W.
Young, J. Catal., 1990, 123, 534.
14 S. N. Goncharova, E. A. Paukshtis and B. S. BalÏzhinimaev, Appl.
Catal., 1995, 126, 67.
15 S. V. Tsybulya, G. N. Kryukova, S. N. Goncharova, A. N.
Shmakov and B. S. Balzhinimaev, J. Catal., 1995, 154, 194.
16 C. D. Wagner, W. M. Riggs, L. E. Davis, J. F. Moulder and G. E.
Muilenberg, Handbook of X-ray Photoelectron Spectroscopy,
Perkin-Elmer, Eden Prarie, MN, 1979.
2
b
5@2
of Ag with higher conductivity show higher activity in ethyl-
ene oxide formation. This conclusion is in accordance with the
data of Campbell38 who has shown that Ag single crystals
have an activity greater by two orders of magnitude than sup-
ported silver. Obviously this is related to features of the
mechanism of ethene epoxidation on silver. As shown in our
previous studies,39h41 to form ethylene oxide both “ionicÏ, and
so-called “covalentÏ, oxygen should coexist at the silver
surface. The “covalentÏ oxygen is incorporated into ethene
providing the formation of ethylene oxide, while the “ionicÏ
oxygen is necessary to create Ag` ions (or surface silver oxide)
as adsorption centres for ethene molecules.41 Formation of
surface silver oxide requires the transfer of electron density
from the silver to O atoms and higher conductivity of the Ag
particles will facilitate this. More e†ective adsorption of
ethene on these centres will in turn increase the total rate of
17 G. Johansson, J. Hedman, A. Berndtsson, M. Klasson and R.
Nilsson, J. Electron Spectrosc. Relat. Phenom., 1973, 2, 295.
18 D. A. Bulushev, E. A. Paukshtis, Yu. N. Nogin and B. S.
BalÏzhinimaev, Appl. Catal., 1995, 123, 301.
19 T. Juska, Z. Bastl, T. Derer and G. Kiss, Poverkhnost (in
Russian), 1987, 10, 96.
20 G. SchoŽ n, Acta Chem. Scand., 1973, 27, 2623.
21 J. S. Hammond, S. W. Gaarenstroom and N. Winograd, Anal.
Chem., 1975, 47, 2193.
22 T. L. Barr, Crit. Rev. Anal. Chem., 1991, 22, 229.
23 T. L. Barr, J. V ac. Sci. T echnol. A, 1989, 7, 1677.
24 C. D. Wagner, Anal. Chem., 1975, 47, 1201.
25 C. D. Wagner, C. D. Gale and R. H. Raymond, Anal. Chem.,
1979, 51, 466.
2328
J. Chem. Soc., Faraday T rans., 1997, V ol. 93

View Article Online
26 V. I. Bukhtiyarov, I. P. Prosvirin and R. I. Kvon, J. Electron
Spectrosc. Relat. Phenom., 1996, 77, 7.
27 R. C. Baetzold, Surf. Sci., 1981, 106, 243.
28 V. Vijayakrishnan, A. Chainani, D. D. Sarma and C. N. R. Rao,
J. Phys. Chem., 1992, 96, 8679.
29 Yu. A. Ryndin, L. V. Nosova, A. I. Boronin and A. L. Chuvilin,
Appl. Catal., 1988, 42, 131.
30 A. F. Carley, M. K. Rajumon and M. W. Roberts, J. Solid State
Chem., 1993, 106, 156.
31 J. Li, Z. Zhang, L. Cai, J. Xu, C. Tang, Q. Zhong, Ya. Xiong and
K. Huang, Proc. 9th Int. Congr. Catal., Calgary, 1988, vol. 4, p.
1680.
32 V. M. Mastikhin, I. L. Mudrakovsky, S. N. Goncharova, B. S.
Balzhinimaev, S. P. Noskova and V. I. Zaikovsky, React. Kinet.
Catal. L ett., 1992, 48, 425.
33 J. K. Plischke, A. J. Benesi and M. A. Vannice, J. Phys. Chem.,
1992, 96, 3799.
34 V. M. Mastikhin, S. N. Goncharova, V. M. Tapilin, V. V. Ters-
kikh and B. S. Balzhinimaev, J. Mol. Catal., 1995, 9, 175.
35 H. E. Rhodes, P-K. Wang, H. T. Stokes and C. Slichter, Phys.
Rev. B, 1982, 26, 3559.
36 C. Backx, J. Moolhuysen, P. Geenen and R. A. van Santen, J.
Catal., 1981, 72, 364.
37 J. K. Plischke and M. A. Vannice, Appl. Catal., 1988, 42, 255.
38 C. T. Campbell, J. Catal., 1985, 94, 436.
39 V. I. Bukhtiyarov, A. I. Boronin and V. I. Savchenko, Surf. Sci.,
1990, 232, L205.
40 V. I. Bukhtiyarov, A. I. Boronin, I. P. Prosvirin and V. I. Sav-
chenko, J. Catal., 1994, 150, 268.
41 V. I. Bukhtiyarov, I. P. Prosvirin and R. I. Kvon, Surf. Sci., 1994,
320, L47.
Paper 6/08414A; Received 16th December, 1996
J. Chem. Soc., Faraday T rans., 1997, V ol. 93
2329