Lawrence Berkeley National Laboratory
Recent Work
Title
Dynamic restructuring drives catalytic activity on nanoporous gold-silver alloy catalysts.
Permalink
https://escholarship.org/uc/item/7vm6x77c
Journal
Nature materials, 16(5)
ISSN
1476-1122
Authors
Zugic, Branko
Wang, Lucun
Heine, Christian
et al.
Publication Date
2017-05-01
DOI
10.1038/nmat4824
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
ARTICLES
PUBLISHED ONLINE: 19 DECEMBER 2016 | DOI: 10.1038/NMAT4824
Dynamic restructuring drives catalytic activity on
nanoporous gold–silver alloy catalysts
Branko Zugic
1
, Lucun Wang
1
, Christian Heine
2
, Dmitri N. Zakharov
3
, Barbara A. J. Lechner
2
,
Eric A. Stach
3
, Juergen Biener
4
, Miquel Salmeron
2
, Robert J. Madix
5
and Cynthia M. Friend
1,5
*
Bimetallic, nanostructured materials hold promise for improving catalyst activity and selectivity, yet little is known about the
dynamic compositional and structural changes that these systems undergo during pretreatment that leads to ecient catalyst
function. Here we use ozone-activated silver–gold alloys in the form of nanoporous gold as a case study to demonstrate
the dynamic behaviour of bimetallic systems during activation to produce a functioning catalyst. We show that it is these
dynamic changes that give rise to the observed catalytic activity. Advanced in situ electron microscopy and X-ray photoelectron
spectroscopy are used to demonstrate that major restructuring and compositional changes occur along the path to catalytic
function for selective alcohol oxidation. Transient kinetic measurements correlate the restructuring to three types of oxygen
on the surface. The direct influence of changes in surface silver concentration and restructuring at the nanoscale on oxidation
activity is demonstrated. Our results demonstrate that characterization of these dynamic changes is necessary to unlock the
full potential of bimetallic catalytic materials.
B
imetallic catalysts present unique opportunities for tuning
product selectivity and increasing energy efficiency in the
chemicals industry. The dynamic response of bimetallic
materials to their environment
1–3
strongly indicates that changes
in structure and surface composition control catalytic function,
demonstrating the need for the investigation of such materials using
in situ techniques. Gold-based alloys
4–8
, in particu lar, have emerged
as promising catalysts for s ele ctive alcohol oxidation, esterif ication,
and amidization reactions.
Nanoporous gold (npAu) is a well-defined, bimetallic, single-
phase solid solution alloy that is not complicated by lattice
mismatch, intermetallic compounds, or metal-support interactions.
It is highly active and selective for oxygen-assisted coupling of both
alcohols and amines
9,10
. A key step in these O-assisted reactions
is the dissociation of O
2
to adsorbed O, which is required for the
initiation of reaction. Although the gold-rich surface is nee ded for
high reaction selectivity, pure metallic gold does not activate O
2
with a me asurable rate
11,12
, whereas this dilute Ag–Au alloy (npAu)
does. Activation of the npAu by flowing ozone and subsequently
a mixture of methanol and O
2
is required for reproducible and
robust catalytic function
9,10,13
. The c atalyst activated in this way
has unique properties, being inactive for sustained, catalytic CO
oxidation
13
(although npAu in other forms is)
14,15
but highly active
for selective oxidative alcohol coupling. This difference makes the
following results distinct from the recent studies by Fujita et al.
16
that
probed structural rearrangements of npAu during CO oxidation on
as-prepared npAu materials.
A critical question addressed here is how and why the ozone
activation process changes the c atalyst. We demonstrate that critical
structural and compositional changes at the surface of npAu (due to
the ozone activation and subsequent reduction by methanol) govern
the development of its unique catalytic activity by using in situ
X-ray and electron microscopy techniques and transient kinetic
measurements. All experiments were performed under similar
conditions of pressure, gas composition, and temperature so that
they could be directly correlated to provide critical details about
the catalyst structure and composition. We show that in situ ozone
pretreatment creates a silver-rich oxide layer on the npAu surface
that is highly reactive to both CO and CH
3
OH to yield the non-
selective oxidation product CO
2
. Removal of this reactive oxygen
by CO leads to restructuring that generates a gold–silver alloy
and a distinct oxygen species stabilized at Ag–Au alloy sites. This
species can be removed by CH
3
OH to exclusively form selective
oxidation products (its removal causes further migration of Ag
into the subsurface) and regenerated by exposure of the alloy
surface to O
2
. The creation of this species is strongly linked to Ag
migration facilitated by ozone treatment, w hich leads to a material
that is functionally distinct from npAu that is highly active for
CO oxidation. This study clearly demonstrates that the behaviour
of alloy catalysts can be tuned using carefully constructed catalyst
pretreatment methods. In addition, this work demonstrates how
catalyst activation—the transformation of a material into a robust
and long-lived catalyst—is critical to the development of efficient
catalytic processes.
Results and discussion
Exposure of ozone-treated npAu to a reactant stream of methanol
(6.5%) and oxygen (20%) under catalytic oxidation conditions
results in the transient production of CO
2
(combustion) b efore
selective coupling to methyl formate commences (Fig. 1). The initial
reaction in the flow reactor is the near-complete combustion of
methanol to CO
2
as the temperature approaches 150
◦
C. The onset
of CO
2
is sudden and accompanied by a release of heat, causing the
temperature to overshoot, thus suggesting an autocatalytic process.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
1
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, USA.
2
Materials Science Division, Lawrence
Berkeley National Laboratory, Berkeley, California 94720, USA.
3
Center for Functional Nanomaterials, Brookhaven National Laboratory, Upton,
New York 11973, USA.
4
Nanoscale Synthesis and Characterization Laboratory, Lawrence Livermore National Laboratory, Livermore, California 94550, USA.
5
Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA.
*
e-mail: friend@fas.harvard.edu
558 NATURE MATERIALS | VOL 16 | MAY 2017 | www.nature.com/naturematerials
NATURE MATERIALS DOI: 10.1038/NMAT4824
ARTICLES
0246810
0
20
40
60
80
100
0
100
200
Time (h)
CH
3
OH conversion and
product selectivity (%)
Temperature (
°
C)
Methyl formate
CO
2
CH
3
OH conversion
Figure 1 | Activation of npAu for selective methanol oxidation. Final
activation of ozone-treated npAu occurs by flowing the methanol/O
2
reaction mixture over the catalyst while the temperature is ramped to the
operating temperature of 150
◦
C. During activation, combustion first
predominates followed by a switch to selective formation of methyl formate
with nearly 100% selectivity. The solid line indicates the point at which a
temperature of 150
◦
C is reached, and the corresponding onset of reaction.
The activity for selective production of methyl formate remains stable for at
least several weeks (see ref. 13.).
After this initial combustion phase, methyl formate becomes
the dominant product, eventually reaching >95% selec tivity. The
activity and selectivity are sustained for a long period—up to several
weeks
13
. Furthermore, the three different forms of nanoporous Au
used—ingots, thin foils, and hollow shells—all exhibit the same
activation behaviour and catalytic activity
13
. The correspondence
in catalytic behaviour demonstrates that the catalytic behaviour is
dictated by the nano- and atomic-scale behaviour of the material
and not the larger-scale architecture.
Combustion of methanol is also initially predominant when the
ozone-treated npAu catalyst is exposed to very short pulses (<1 ms)
of either methanol alone or a methanol–CO mixture (Fig. 2a,c and
Supplementary Fig. 1). The rapid transition from CO
2
to selective
oxidation indicates that there are two types of oxygen species on the
surface—one responsible for combustion and the other for parti al
oxidation. Due to extremely short contact times in the pulse reactor,
the dominant selective oxidation product is formaldehyde, rather
than methyl formate. In the flow reactor, the same catalyst yielded
methyl formate with high selectivity, as previously reported. It is
well known that the first step in the catalytic c ycle for methanol
self-coupling is activation of methanol to formaldehyde
17,18
; the
transient experiments probe the initial step in t he reaction pathway
that governs the steady-state reaction. The difference in product
selectivity in the titration with methanol pulses and in steady flow
is related to the large difference in contact times (10
−4
s versus
10
−1
s, respectively) with the catalyst and will be addressed in a
separate paper
19
.
In a sep arate experiment, the surface oxygen resulting from
O
3
-treatment of the npAu surface was first titrated with 100 pulses of
CO to remove the surface O responsible for combustion. Exposure
to methanol pulses then directly yields partial oxidation products
and not the initial methanol combustion (indicated by the dashed
line in Fig. 2b). This result indicates that the chemical behaviour
of the adsorbed O is not sensitive to the method of reduction of
the more reactive oxygen that promotes combustion. In addition, a
pause during the continuous pulsing of either CO or CH
3
OH has
minimal effect on the evolution of the different products as the
pulses are resumed, indicating that the different states of surface
oxygen from O
3
treatment are not inter-convertible.
When the O
3
-treated npAu catalyst is first exposed to pulses
of a mixture of CO and CH
3
OH simultaneously, both reactants
are completely consumed and CO
2
is the only product observed
(Fig. 2c,d). After most of t he surface oxygen (95%) is reacted away
through CO
2
production, the CO conversion rapidly drops to zero.
Concomitantly, selective reaction of methanol to formaldehyde and
methyl formate commences without CO
2
formation. The pattern of
methanol reactivity is essentially the same whether it is pulsed alone
or in the presence of CO (Fig. 2).
These results clearly demonstrate that there are two chemically
distinct types of oxygen species formed during the O
3
treatment of
the catalyst and that only the minority oxygen species, present on
the surface at low concentrations, is responsible for the selective
oxidation of methanol. Furthermore, the catalyst faci litates the
dissociation of O
2
after complete reaction of the ‘selective’ oxygen
via reaction with methanol. Exposure of this material to dioxygen
produces adsorbed O, but only from reaction with the minority
sites
19
; this adsorbed oxygen is unreactive with CO but reacts
readily with methanol to give selective oxidation products. This is
in agreement with previous studies of ozone-activated npAu
13
.
The primary morphological change observed by in situ
aberration-corrected transmission electron microscopy (TEM)
after ozone treatment is the formation of an amorphous thin film
oxide on the surface of the npAu (Fig. 3 and Supplementary Fig. 2).
The oxide covers >80% of the npAu surface (Fig. 3b,c) with a
thickness of 1.1 ± 0.1 nm. This oxide is not present in as-prepared
npAu (that is, not treated with ozone). Additionally, islands (Fig. 3c)
and embedded oxides (Fig. 3d) are observed on and within the
npAu ligaments, respectively. The islands are two atomic layers in
height and 2–6 nm in diameter. The observable inter-planar spacing
within these structures is ∼0.31 nm (compared with 0.24 nm for
the (111) spacing of metallic Au and Ag), w hich suggests formation
of metal oxides (Supplementary Table 1)
20,21
. These findings are in
line with previous studies of textured Au(111) surfaces that indicate
that t reatment with ozone results in the formation of chemisorbed
oxygen species and three-dimensional gold oxides of ∼1.5 nm in
thickness
22
. The disordered, embedded oxides also observed by
TEM (Fig. 3d) are ∼1–2 nm deep and ∼6–8 nm in diameter.
In addition to these structural changes, high-angle annular dark-
field scanning transmission electron microscopy/electron energy-
loss spectroscopy (HAADF-STEM/EELS) analysis clearly shows that
the ozone treatment causes the aggregation of Ag oxides at the npAu
surface. This is seen in the HAADF-STEM image (Fig. 3d) as regions
of lower density (that is, lower Z contrast). Further analysis by EELS
(Fig. 3e and Supplementary Fig. 3) shows that these regions are hig h
in Ag and O concentration. By contrast, little to no Ag is detected in
the npAu away from these regions.
All of the amorphous oxide layer is removed by in situ reduction
of the O
3
-treated npAu with CO (0.1 torr CO at 150
◦
C for 30 min)
in the environmental TEM (E-TEM Fig. 4a–c). Since catalytic CO
oxidation is not sustained, we infer that the oxide is not reformed
under operating catalytic conditions. The removal of the oxide
layer is accompanied by formation of smal l, irregular nanoparticles
on highly stepped regions of npAu (Fig. 4b,c). The newly formed
particles are irregularly shaped with large variations in the atomic
spacing (Fig. 4b). The larger lattice spacing (∼0.32 nm), determined
by Fourier transform analysis, corresponds to that of gold and silver
oxides (Supplementary Table 1).
Further reduction of the npAu surface by methanol results in
the growth of crystalline nanopart icles at the surface to 2–2.5 nm
in diameter (Fig . 4d–f). In addition to the increase in diameter,
these structures appear more ordered and show more distinct
facets, as shown in Fig. 4f. The particle aspect ratio (height to
width ratio at the base of the par ticle) also changes from 0.7 after
NATURE MATERIALS | VOL 16 | MAY 2017 | www.nature.com/naturematerials
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
559
ARTICLES
NATURE MATERIALS DOI: 10.1038/NMAT4824
0
5
10
15
0 50 100 150 200
CH
3
OH pulsing
MS intensity
0
5
10
15
MS intensity
CO
2
HCHO
HCOOCH
3
× 10
HCOOCH
3
× 10
CH
3
OH pulsing
CO
pulsing
Number of CO or CH
3
OH pulses
CO
2
HCHO
0
1
2
10
15
0 100 200 300 400 500 600
0.0
0.2
0.4
0.6
0.8
1.0
CO
2
MS intensity
HCHO
HCOOCH
3
×
10
Simultaneous pulsing
of CO and CH
3
OH
CO
CH
3
OH
Normalized MS intensity
Number of CO and CH
3
OH pulses
0 50 100 150 200
Number of CO or CH
3
OH pulses
0 100 200 300 400 500 600
Number of CO and CH
3
OH pulses
a
d
c
b
Figure 2 | Pulsed experiments over ozone-treated npAu. Transient (pulsed) experiments over O
3
-treated npAu at 150
◦
C demonstrate that there are two
chemically distinct oxygen species present on the surface: one responsible for combustion to CO
2
and the other for selective CH
3
OH oxidation to HCHO
and HCOOCH
3
. a, During CH
3
OH pulsing over O
3
-treated npAu, complete combustion to CO
2
is observed initially; af ter ∼25 pulses, combustion rapidly
diminishes and selective CH
3
OH oxidation to formaldehyde is observed; the reaction ceases after ∼200 pulses. b, Non-selective oxygen species can be
removed by titration of highly reactive oxygen on O
3
-treated npAu by exposure to 75 pulses of CO; CH
3
OH pulsing thereafter yields only selective
oxidation products. c,d, Titration of surface oxygen on O
3
-treated npAu by simultaneously pulsing methanol and CO onto O
3
-treated npAu yields an
identical pattern of reactivity—initial reaction of both CO and CH
3
OH to CO
2
, followed by reaction of only CH
3
OH to selective oxidation products (the
pulse sizes of CO and CH
3
OH are reduced by 50% relative to a and b to keep total reactant per pulse constant).
CO treatment to 0.3 after CH
3
OH exposure. These changes may
indicate a higher propensity for the particles to realloy into the bulk
at low oxygen concentrations at the surface. Furthermore, there is
a clear alignment of the crystal planes in the par ticle and ligament,
suggesting that the particles are more metallic in nature.
The oxidation and enrichment of Ag at the npAu surface after
ozone treatment is demonstrated using ambient-pressure X-ray
photoelec tron spectroscopy (AP-XPS; Fig. 5 and Supplementary
Tables 2–4). Ess entially all Ag and 80% of Au within the 0.5 nm
sampling depth (t hat is, inelastic mean free path through Au)
23
are
initially oxidized by ozone at 150
◦
C (Fig. 5a and Supplementary
Table 3). Immediately following the ozone treatment, the Ag/Au
atomic ratio is 0.46 (a 30% increase relative to fresh npAu), even
though the bulk npAu Ag concentration is only ∼3 at% based
on energy dispersive X-ray spectroscopy
13
. Furthermore, all of the
detectable silver on the surface is f ully oxidized to AgO (ref. 24).
The presence of chemically distinct oxygen species after ozone
treatment is also detected using the AP-XPS. The O1s binding
energy is in agreement with previously reported oxygen states
for bulk Au and Ag oxides (Fig. 5a and Supplementary Table 4,
O
x
species)
22,25–27
. Following reduction by CO (0.1 torr, 150
◦
C),
∼90% of this oxidic oxygen is removed (Fig. 5b), which correlates
with the CO pulse titration results described above and with the
restructuring of the materials in E-TEM. These results further
confirm that CO readily reacts with t he oxidic oxygen. A new
oxygen species (O
sel
) with a binding energy of 531.5 eV appears in
conjunction with the reduction (Fig. 5b and Supplementary Fig . 4).
This oxygen species persists until exposure to CH
3
OH (Fig. 5c)
and it is attributed to the second state of O that leads to selective
oxidation of methanol. There is also a residual state of oxygen (O
res
)
with a binding energy of 532 eV that remains in the surface region
even after the 30 min exposure to methanol at 150
◦
C; this species is
attributed to oxygen in the subsurface
28
. D epth profiling of the npAu
over the course of these experiments indicates that the oxidized
Ag and Au species are present only at the outer surface region and
that realloying of Ag into the bulk takes place during the sequential
reduction by CO and CH
3
OH (Supplementary Fig. 5).
Reduction of the O
3
-treated npAu also leads to a significant
rearrangement of gold and silver within the surface region such
that the Ag/Au ratio decreases significantly from 0.46 to 0.26 af ter
reduction of the material by CO (Figs 5b and 6). A corresponding
increase in the abundance of undercoordinated Au (Au
u/c
), metallic
Au (Au
0
), and Ag-alloyed Au (Au
alloy
) is also observed (Fig. 6 and
Supplementary Information)
29
. Nevertheless, s ome Ag
2
O remains
and the ratio of the Au
alloy
to Ag
alloy
is found to b e ∼1. Subsequent
reduction by CH
3
OH (Fig. 5c) results in a further reduction of silver
content in the surface (to an Ag/Au ratio of 0.11), a further increase
in the undercoordinated gold sites, and the complete removal of the
560
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE MATERIALS | VOL 16 | MAY 2017 | www.nature.com/naturematerials
NATURE MATERIALS DOI: 10.1038/NMAT4824
ARTICLES
iv
v
iii
ii
f
Ag
O
Au
400 500 600
Energy loss (eV)
Counts (a.u.)
b
e
5 nm
Oxide
c
5 nm
Island
a
d
b
c
50 nm
AgO
x
2 nm
ii
iii
iv
v
i
20 nm
i
Figure 3 | E-TEM analysis of npAu after ozone treatment. In situ aberration-corrected TEM images demonstrate that npAu is oxidized by O
3
treatment
in a flow reactor at 150
◦
C for 1 h. a, Low-magnification image of the material showing the pore and ligament structure. b, High-resolution image of a
representative spot on the material showing the resolved lattice of the ligaments and a layer of amorphous oxide after O
3
treatment. c, In some areas,
crystalline islands are visible on the npAu surface in the high-magnification images. d, Ex situ aberration-corrected TEM image showing the presence of
amorphous oxides embedded into the npAu ligament. e, HAADF-STEM image showing areas of varying contrast on the npAu surface, corresponding to
features seen by TEM in d. f, EELS analysis of various spots from e, showing that the darker areas are regions of high Ag and O concentration relative to the
rest of the npAu surface.
b
c
a
bc
5 nm
f
g
h
h
5 nm 5 nm 2 nm
1 nm 1 nm
0.32 nm
0.32 nm
d
e
i
j
d
i
0.27 nm
Figure 4 | E-TEM analysis of npAu during CO and CH
3
OH exposure. In situ removal of the oxide layer and formation of nanoparticles due to reduction of
ozone-treated npAu by CO and CH
3
OH is observed in aberration-corrected E-TEM images. a–c, Exposure of O
3
-treated npAu to 0.1 torr CO at 150
◦
C for
30 min leads to removal of the oxide film and precipitation of defective nanoparticles (1–1.6 nm in diameter) with an expanded lattice. d, Masked fast
Fourier transform of the region marked in b, showing the two types of lattice spacing observed. e, Inverse fast Fourier transform of d, showing the oxidic
0.32 nm spacing corresponding to the particle in b. f–h, Highly crystalline metallic nanoparticles form following further reduction by 0.1 torr CH
3
OH at
150
◦
C for 30 min. i, Masked fast Fourier transform of the region marked in h. j, Inverse fast Fourier transform of i, showing a typical Au(111)-type
arrangement. All images were obtained at 25
◦
C and under vacuum conditions after reduction treatments; no beam eects were observed during imaging.
Note: the observed particle density and shape on the npAu surface accounts for ∼11% of the npAu surface area after CO reduction and ∼27% after
CH
3
OH reduction.
NATURE MATERIALS | VOL 16 | MAY 2017 | www.nature.com/naturematerials
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
561
