You are on page 1of 8

ARTICLES

PUBLISHED ONLINE: 19 SEPTEMBER 2016 | DOI: 10.1038/NCHEM.2607

Adsorbate-mediated strong metalsupport


interactions in oxide-supported Rh catalysts
John C. Matsubu1, Shuyi Zhang2,3, Leo DeRita1, Nebojsa S. Marinkovic4, Jingguang G. Chen4,5,
George W. Graham2,3, Xiaoqing Pan2,6 and Phillip Christopher1,7,8*

The optimization of supported metal catalysts predominantly focuses on engineering the metal site, for which physical
insights based on extensive theoretical and experimental contributions have enabled the rational design of active sites.
Although it is well known that supports can inuence the catalytic properties of metals, insights into how metalsupport
interactions can be exploited to optimize metal active-site properties are lacking. Here we utilize in situ spectroscopy and
microscopy to identify and characterize a support effect in oxide-supported heterogeneous Rh catalysts. This effect is
characterized by strongly bound adsorbates (HCOx) on reducible oxide supports (TiO2 and Nb2O5) that induce oxygen-
vacancy formation in the support and cause HCOx-functionalized encapsulation of Rh nanoparticles by the support. The
encapsulation layer is permeable to reactants, stable under the reaction conditions and strongly inuences the catalytic
properties of Rh, which enables rational and dynamic tuning of CO2-reduction selectivity.

I
dentication of the optimal active sites on supported metal cata- O2 (refs 21,22), which creates a situation in which only a small frac-
lysts often focuses on engineering the composition or geometry tion of the metal sites is inuenced by the partially reduced support,
of the metal site to maximize the reaction rate or control the reac- specically at the metalsupport interface2326. The poisoning or
tion selectivity15. Much less is known about how metalsupport receded SMSI overlayer structures that exist under catalytic reaction
interactions can be exploited to control the reactivity of hetero- conditions have curtailed the use of SMSI overlayers to increase the
geneous metal-oxide-supported metal catalysts, although the reactivity or control the selectivity on supported metal catalysts.
support characteristics can inuence catalytic reactivity or selectivity Here we demonstrate an SMSI encapsulation state that forms
as considerably as the characteristics of the metal6,7. Demonstrated with the treatment of TiO2- and Nb2O5-supported Rh nanoparticles
mechanisms of support effects on metal reactivity include small- in CO2H2 (CO2-rich) environments at temperatures of 150300 C.
cluster stabilization8, charge transfer912, support participation in In situ spectroscopy and microscopy show that the high coverage of
catalysis7,13,14 and oxide encapsulation of metal nanoparticles1519. adsorbates (HCOx) on the support induces oxygen-vacancy
Encapsulation of metal nanoparticles by reducible oxide-support formation, which drives migration of the HCOx-functionalized
overlayers is the only mechanism by which supports can affect cat- support onto the metal. This adsorbate-mediated SMSI (A-SMSI)
alysis at a majority of active sites on metal particles with diameters encapsulation state is stabilized against re-oxidation by H2O and
larger than 12 nm (ref. 12); this was designated by Tauster et al.15 modies the reactivity of all the remaining exposed Rh sites, and
as strong metalsupport interactions (SMSIs). The SMSI appears to be comprehensive in covering Rh but amorphous
encapsulation state forms as a result of high-temperature H2 and permeable to reactants. Formation of the A-SMSI state
treatment of reducible oxide-supported Pt-group metals, which induces a selectivity switch in the CO2-reduction reaction from
causes a reduction of the oxide support to substoichiometric CH4 production on bare Rh particles to CO production in the
oxygen concentrations and induces oxide migration on top of A-SMSI state, which effectively renders Rh less active for CH
the metal nanoparticles. It has been hypothesized that bonding bond formation. Our results show that the A-SMSI state represents
between the cationic support-metal atoms and the metal-catalyst a powerful support effect that enables the rational manipulation
surface makes migration of the support onto the metal of metal-catalyst reactivity.
thermodynamically favourable20.
The excitement that surrounds the discovery of SMSI overlayers Results
was stoked by the suggestion that the SMSI state could be used to Effect of in situ pre-treatments on CO2-reduction catalysis. In a
tune metal-catalyst reactivity via the partial decoration of metals recent analysis of TiO2-supported Rh (Rh/TiO2) catalysts for the
by oxide overlayers16. However, the SMSI encapsulation state reduction of CO2 by H2 , we observed a dynamic decrease in the
rarely nds an intermediate conguration in which partial metal rate of CH4 production and increase in the rate of CO production
coverage by the oxide allows interaction with a majority of the when operating at CO2:H2 ratios greater than 1 (ref. 5). Based on
exposed metal sites. Instead, the oxide overlayer either covers all an assignment of CO production that occurs at isolated Rh atoms
the metal sites, which renders catalysts inactive, or retreats off the and CH4 production that occurs at Rh structurenanoparticle
metal because of re-oxidation of the reduced support by H2O or surfaces, the reactivity change was attributed to Rh-nanoparticle

1
Department of Chemical and Environmental Engineering, University of California, Riverside, California 92521, USA. 2 Department of Chemical Engineering
and Materials Science, University of California, Irvine, Irvine, California 92697, USA. 3 Department of Materials Science and Engineering, University of
Michigan, Ann Arbor, Michigan 48109, USA. 4 Department of Chemical Engineering, Columbia University, New York 10027, USA. 5 Chemistry Department,
Brookhaven National Laboratory, Upton, New York 11973, USA. 6 Department of Physics and Astronomy, University of California, Irvine, Irvine, California
92697, USA. 7 Program in Materials Science, University of California, Riverside, Riverside, California 92521, USA. 8 UCR Center for Catalysis, University of
California, Riverside, Riverside, California 92521, USA. * e-mail: christopher@engr.ucr.edu

120 NATURE CHEMISTRY | VOL 9 | FEBRUARY 2017 | www.nature.com/naturechemistry

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE CHEMISTRY DOI: 10.1038/NCHEM.2607 ARTICLES
a 36 b c 12.0
100 100
Reduced

)
SCH4

1
Rate (mmol C2+ h1 gRh
Reduced

20CO2:2H2 treatment
20CO2:2H2
)

CH4 selectivity (%)


CH4 selectivity (%)
Rate (mmol h1 gRh
1

27 75 75 9.0
CH4
Reduction

Reduction
18 50 50 6.0

9 25 25 3.0
20CO2:2H2
CO
0 0 0 0.0
0 6 12 18 24 0 2 4 6 1 3 5 10 15 25

Time (h) Wt% Rh CO2 and H2 concentration (%)

Figure 1 | Control of CO2 reduction selectivity on Rh via catalyst pre-treatment. a, Rate of CO and CH4 production and selectivity to CH4 (SCH4) over
6% Rh/TiO2 measured at 200 C with a feed composition of 1% CO2 , 1% H2 and 98% He, after treatments with pure H2 at 450 C for 4 h (reduction) and
then 20CO2:2H2 at 250 C for 4 h. Between treatments, CH4 selectivity was reversibly controlled from 98 to 11%. b, Effect of 20CO2:2H2 treatment on CH4
selectivity as a function of Rh weight loading (%). c, Effect of 20CO2:2H2 treatment on the C2+ production rate over 2% Rh/TiO2 catalyst as a function of
feed composition with an equimolar CO2 and H2 feed and a balance of He. 20CO2:2H2 treatment consistently decreased the CH4 and C2+ selectivities.
Experimental errors for the rate and selectivity measurements were consistently less than 6%. For plots in which the reactivity data after different treatments
are separated along the x-axis, the CO and CH4 formation rate and CH4 selectivity are in orange, green and purple, respectively. When reactivity or selectivity
after different treatments is plotted on the same x-axis, reduced catalyst data are in red and 20CO2:2H2-treated catalyst data are in blue.

disintegration into isolated Rh atoms. However, the disproportionate reactant partial pressures, and showed a robust behaviour despite
magnitude of changes in CO and CH4 production rates being in conditions at which signicant H2O concentration is
(Supplementary Fig. 1) cannot be rationalized by the disintegration produced (Supplementary Figs 4,68).
mechanism, which suggests that a more-complex physical catalyst The inuence of the 20CO2:2H2 treatment on CO2-reduction
transformation was responsible for the changing selectivity. selectivity was tested for various Rh weight loadings on TiO2
To understand the controllability and mechanism of the (Fig. 1b and Supplementary Figs 914). Consistently, the
observed dynamic change in CO2-reduction selectivity, a series of 20CO2:2H2 treatment decreased CH4 production and increased
oxide-supported Rh catalysts with varying oxide compositions CO production, which was reversible on re-reduction. A low CH4
and Rh weight loadings were synthesized (Supplementary selectivity on catalysts with a lower Rh weight loading after
Information). To explore the effect of various pre-treatments on reduction was caused by high concentrations of isolated Rh-atom
CO2-reduction selectivity, Rh/TiO2 catalysts were reduced at 450 C active sites. The largest selectivity changes were observed for
in H2 , evaluated for reactivity at 1%CO2:1%H2:98%He and 200 C higher Rh weight loadings that predominantly consist of Rh-nano-
(the reaction conditions were chosen for catalyst stability particle active sites, which suggests that the 20CO2:2H2 treatment
(Supplementary Figs 2 and 3) and used throughout the report modied the reactivity of Rh nanoparticles. The negligible change
except when stated otherwise), exposed to various environments in CO production on 0.2% Rh/TiO2 (Supplementary Fig. 9), on
(Supplementary Table 1) and evaluated again for reactivity at 1% which nearly all the catalytic sites are isolated Rh atoms, indicates
CO2:1%H2:98%He and 200 C. Only treatments in CO2:H2 environ- that the 20CO2:2H2 treatment has little effect on the reactivity of iso-
ments with the CO2:H2 feed ratio greater than one induced a rapid lated Rh atoms. The production rates of ethane and propane (C2+
(on the time scale of four hours) selectivity switch from CH4 pro- products) were also suppressed by the 20CO2:2H2 treatment, as
duction on the reduced catalysts to CO production on the treated shown for 2% Rh/TiO2 in Fig. 1c, which is consistent with prior
Rh/TiO2 catalysts. work in which catalysts that exhibited a high CH4 selectivity typi-
After identifying the optimal treatment conditions (20% CO2: cally produced more C2+ products27. The results in Fig. 1 demon-
2% H2:78% He at 250 C for four hours, hereafter the 20CO2:2H2 treat- strate that the reactivity of Rh/TiO2 for the CO2-reduction
ment) that induce a selectivity switch from CH4 to CO production, reaction is dynamically tunable via 20CO2:2H2 and H2 treatments, in
the reversibility was tested by exposing the 20CO2:2H2-treated cat- which the Rh-nanoparticle reactivity after 20CO2:2H2 treatment is
alyst to H2 environments. H2 treatment at temperatures greater consistent with the catalytic behaviour of the more-noble (Pt, Pd, Cu)
than 350 C for four hours restored the original CH4 selectivity of metal catalysts2830.
the reduced catalyst (Supplementary Fig. 4). A complete cycle is
shown in Fig. 1a for 6% Rh/TiO2 , in which the CH4 selectivity Rh structure stability during treatments. Considering that isolated
decreased from 98% after reduction (450 C for four hours) to Rh atoms are selective for CO production and Rh nanoparticles are
11% after 20CO2:2H2 treatment and returned to 98% CH4 selectivity selective for CH4 production, it could be hypothesized that the
after re-reduction (450 C for four hours). The switch in CO2-reduction 20CO2:2H2 treatment disintegrates Rh nanoparticles to form
selectivity was induced by a 40-fold decrease in the CH4 formation rate isolated Rh atoms on the TiO2 surface31,32. In situ X-ray
from 28 to 0.7 mmol CH4 h1 g1 Rh and a tenfold increase in the CO absorption spectroscopy (XAS) following the Rh K edge on a 2%
formation rate from 0.5 to 5 mmol CO h1 g1 Rh, and was completely Rh/TiO2 catalyst during the 20CO2:2H2 treatment showed no
reversible for multiple cycles of reduction and 20CO2:2H2 treat- measurable difference between the reduced and 20CO2:2H2-
ments (Supplementary Fig. 5). Typical H2 conversions measured treated catalyst, with a constant RhRh coordination number
at 1%CO2:1%H2:98%He and 200 C were below 8%. To ensure that (Supplementary Figs 15 and 16). The XAS results were consistent
the change in selectivity after 20CO2:2H2 treatment was not simply with ex situ scanning transmission electron microscopy (STEM)
the result of a lower reactant conversion, the selectivity was compared images of 2% Rh/TiO2 catalysts after reduction and 20CO2:2H2
at identical conversions (4%), shown in Supplementary Fig. 6 (300 C) treatment showed no evidence of Rh structural changes
and Supplementary Fig. 7 (200 C), and large differences were still (Supplementary Fig. 17). Based on the lack of structural changes
observed. The stability of the CO-producing state was tested at in Rh observed by XAS and STEM and disproportionate
varying H2-treatment temperatures, reaction temperatures and changes in the rate of CH4 and CO production (Fig. 1a and

NATURE CHEMISTRY | VOL 9 | FEBRUARY 2017 | www.nature.com/naturechemistry 121

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
ARTICLES NATURE CHEMISTRY DOI: 10.1038/NCHEM.2607

a b
1

RhCO

KM intensity (a.u.) TiO2(HCOx)

Rh2(CO)
TiO2H2O

Reduced

KM intensity (a.u.)
20CO2:2H2 treatment

2,050 1,850 1,650 1,450 1,250


Wavenumber (cm1)
2
c 3 1% H2 60 min
2,080 In situ H2 flow TPD
4
5
Linear CO stretch (cm1)

2,040 Reduced
6 TPD in He

2,000 7
20CO2:2H2
8

1,960
0 20 40 60 200 300 400 2,200 2,000 1,800 1,600 1,400 1,200
Time (min) Temperature (C) Wavenumber (cm1)

Figure 2 | Infrared analysis of selectivity switch. a, In situ DRIFT spectra collected from 2% Rh/TiO2 at the reaction conditions (180 C, 1% CO2 , 1% H2 ,
98% He) for reduced (red), 20CO2:2H2-treated (blue) and re-reduced (black) catalysts. On 20CO2:2H2 treatment, the CO stretching frequency of linear and
bridge Rh-carbonyl groups decreased by 50 cm1, their intensity dropped twofold and TiO2-bound HCOx species appeared. Re-reduction reversed all the
effects of the 20CO2:H2 treatment. b, DRIFT spectra of 2% Rh/TiO2 under the reaction conditions after reduction (spectrum 1), after 20CO2:2H2 treatment
(spectrum 2) and CO2 ow removed (only 1% H2) for 5, 15 and 60 min (spectra 35). Spectra 68 were collected during TPD in He at 213, 312 and 370 C,
respectively, immediately after the collection of spectrum 5. The spectra are presented in Kubelka Munk (KM) units. The vertical dotted line represents
the CO stretch frequency on a reduced 2% Rh catalyst under the reaction conditions. Corresponding increases in the carbonyl stretching frequency with
HCOx desorption suggest they are related. c, Peak positions of the linear CO stretching frequency during the experiments shown in Fig. 2b and Supplementary
Fig. 18. The reduced (red) and 20CO2:2H2-treated (blue) catalysts exhibited a similar decrease in CO stretching frequency in H2 ow, which suggests that the
shift in carbonyl frequency in a did not result from a decrease in CO coverage.

Supplementary Fig. 1), the switch in reactivity induced by 20CO2:2H2 Rh36, a local electric-eld-induced Stark effect37 or coordination of
treatment is attributed to modied RhTiO2 interactions. CO across a metalsupport interface23. 20CO2:2H2 treatment also
introduced a high coverage of formate (HCO2 , 2,973, 2,923, 2,853,
Changes in adsorbates on Rh and TiO2. In situ diffuse-reectance 1,531 and 1,351 cm1) and a bicarbonate-like species (HCO3 ,
infrared Fourier transform spectroscopy (DRIFTS) was used to 1,444 cm1) on the TiO2 surface38, which were only observed
examine the effect of 20CO2:2H2 treatment on the species under reaction conditions following the 20CO2:2H2 treatment.
adsorbed on Rh and TiO2 under the reaction conditions. Re-reduction of the catalyst regenerated spectral characteristics of
Figure 2a shows in situ DRIFT spectra acquired from 2% Rh/TiO2 the freshly reduced catalyst, consistent with the regenerated reactivity
under the reaction conditions (1%CO2:1%H2:98%He and 200 C) shown in Fig. 1a.
after reduction, 20CO2:2H2 treatment and re-reduction, essentially To identify whether the 50 cm1 redshift and decreased inten-
identical to those in Fig. 1a. Under the reaction conditions, the sity of the linear CO stretching mode were simply caused by a
DRIFT spectrum of the reduced catalyst shows CO linearly bound decrease in CO coverage on Rh, CO2 was removed from the reactant
to Rh at the top (2,046 cm1) and bridge (1,880 cm1) sites and a stream to leave H2 to react with the adsorbates. After removal of
low adsorbate coverage on TiO2 , aside from H2O (1,620 cm1), in CO2 from the reactant stream, CO stretches redshifted 3540 cm1
agreement with previous reports33,34. After 20CO2:2H2 treatment, as the CO coverage decreased on both the reduced and
the switch in CO2-reduction selectivity correlated with a 50 cm1 20CO2:2H2-treated catalysts (spectra 14 in Supplementary
redshift in the frequency of the linear- and bridge-bound CO Fig. 18 and spectra 25 in Fig. 2b). This is shown quantitatively in
stretching modes and a twofold decrease in the integrated area of Fig. 2c, in which the reduced and 20CO2:2H2-treated catalysts
the linear-bound CO stretch. The redshift and decreased intensity exhibit a similar decrease in the CO stretching frequency in H2
of the linear CO stretch could be explained by a decrease in the ow. The coverage-dependent redshift in the CO stretching fre-
CO coverage of Rh35 or by physical blocking of Rh sites coupled quency of 3540 cm1 (spectra 14 in Supplementary Fig. 18 and
with the polarization of CO bonds, induced by charge transfer to spectra 25 in Fig. 2b) is consistent with the effect of reduced

122 NATURE CHEMISTRY | VOL 9 | FEBRUARY 2017 | www.nature.com/naturechemistry

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE CHEMISTRY DOI: 10.1038/NCHEM.2607 ARTICLES
a 100 b 150 120

CH4
Reduced
80 120 CO
90

)
CH4 selectivity (%)

CH4 selectivity (%)


Rate (mmol h1 gRh
1
CH4 selectivity
60 90
60
40 60

30
20 30
20CO2:2H2

0 0 0
100 200 300 400 500 Reduced Formic acid Re-reduced
treatment
Temperature of CO2:H2 treatment (C)

Figure 3 | Identifying the mechanism of A-SMSI formation. a, CH4 selectivity on 2% Rh/TiO2 as a function of 20CO2:2H2 treatment temperature, shown in
blue, measured at 200 C with a feed composition of 1% CO2 , 1% H2 , 98% He (100 sccm total). The catalyst was re-reduced after each sequential
20CO2:2H2 treatment, and the reactivity under identical reaction conditions is shown in red. The experiment involved an identical protocol to that for
Supplementary Fig. 5, but with increasing temperature during the 20CO2:2H2 treatment for each cycle. Lines are to guide the eye. b, Rate of CH4 and CO
production and CH4 selectivity after reduction, formic acid treatment and re-reduction on a 2% Rh/TiO2 catalyst measured under reaction conditions of
200 C, 1% CO2 , 1% H2 and 98% He (100 sccm total). Formic acid treatment was 25 sccm He bubbled through formic acid at room temperature for 4 h
with a catalyst-bed temperature of 250 C. The agreement in the temperature at which the 20CO2:2H2 treatment affected CH4 selectivity and formic-acid-induced
oxygen-vacancy formation4042 in a, along with the inuence of formic acid on reactivity in b, suggests an A-SMSI. Experimental errors for the rate and selectivity
measurements were consistently less than 6%.

dipole coupling on nanoparticle-catalyst surfaces35, which indicates Formic-acid-induced oxygen-vacancy formation in TiO2 occurs
that the 20CO2:2H2-treatment-induced 50 cm1 redshift (Fig. 2a) under conditions that resemble the conditions we observed
in the CO stretching frequency was not caused by a change in following 20CO2:2H2 treatment, in which TiO2 is covered by
local CO coverage. HCOx. The temperature range (150300 C) in which 20CO2:2H2
The origin of the 50 cm1 shift in the CO stretching frequency treatment induced a switch in CO2-reduction selectivity agrees with
was probed further by executing a temperature-programmed deso- that of formic-acid-induced TiO2 reduction, in which the TiO2
rption (TPD) in He after the coverage-dependent experiment reduction is limited at a low temperature (100 C) by H2O-
(spectra 68 in Fig. 2b). The HCOx species desorbed from TiO2 desorption kinetics and at high temperature (300 C) by formate
above 300 C with a simultaneous blueshift in frequency of the desorption (Fig. 3a). Furthermore, Fig. 3b shows that, similar to
remaining linearly bound CO by 35 to 1,996 cm1, almost iden- the 20CO2:2H2 treatment, formic acid treatment of 2% Rh/TiO2
tical to the frequency observed at low coverage on the reduced cat- decreased the rate of CH4 production by 14-fold and increased
alyst (Fig. 2c). Catalyst performance was tested after the HCOx the rate of CO production by vefold compared with the reduced
species were desorbed from the 20CO2:H2-treated sample and the catalyst, which was reversible on re-reduction.
reactivity was restored to the behaviour of the original reduced cat- The similarity between the inuence of 20CO2:2H2 and formic
alyst state (Supplementary Fig. 19). Correlation between the removal acid treatments on the reactivity of Rh/TiO2 catalysts suggests the
of HCOx species from TiO2 and the blueshift in the CO stretching existence of an A-SMSI state in which a high coverage of HCOx
frequency, combined with the return of the reduced catalyst reactiv- induces the formation of oxygen vacancies at the TiO2 surface,
ity, demonstrates that the modied reactivity induced by 20CO2:2H2 and thereby causes migration of the support onto Rh. However, it
treatment is mediated by interactions between HCOx and the can be imagined that charge donation from HCOx-functionalized
TiO2 support. reduced TiO2 into Rh, without migration onto Rh, causes the modi-
The decreased intensity of the CO stretching modes on Rh under ed reactivity. To differentiate these effects, a 6% Rh/TiO2 catalyst
reaction conditions after 20CO2:2H2 treatment and the resulting was sintered, which increased the Rh particle sizes from 13 nm
inuence of HCOx-functionalized TiO2 on all the remaining to 1050 nm in diameter (Supplementary Fig. 20). The sintered
RhCO bonds, both in terms of reactivity and spectroscopically, catalyst was evaluated for reactivity after reduction and 20CO2:2H2
are consistent with SMSI overlayer effects19,23. However, the traditional treatment, and CH4 selectivity decreased from 97% after reduction
SMSI encapsulation state for Rh on TiO2 forms at a higher tempera- to 4% after 20CO2:2H2 treatment (Supplementary Fig. 20). The
ture (500 C) and in more-reducing conditions (pure or diluted H2) almost identical inuence of 20CO2:2H2 treatment on the sintered
than the 20CO2:2H2 treatment and recedes off Rh in the moist and unsintered 6% Rh/TiO2 , with large differences in Rh particle
atmosphere of CO2 reduction, which results in little inuence on sizes, strongly supports the mechanism of HCOx-mediated
reactivity21. Furthermore, the SMSI overlayer on Rh/TiO2 only migration of TiOx onto Rh, because the charge-transfer mechanism
blocks the available Rh sites for CO adsorption, but does not without overlayer formation would have been suppressed by the
modify the stretching frequency of CO adsorbed at uncovered Rh increased particle size in the sintered catalyst12. Thus, interaction
sites, in contrast with the observations in Fig. 2a39. between HCOx-functionalized TiO2 and the Rh active sites
must occur locally at a metaloverlayer interface as a result of
HCOx-induced TiO2 reduction. Interestingly, analysis of the coordination between the overlayer and the Rh surface.
decomposition of formic acid on TiO2 has shown that at a high
coverage of formic acid and temperatures greater than 100 C, Physical and chemical nature of the encapsulation state. To
oxygen vacancies on the TiO2 surface form via H2O desorption to visualize directly the 20CO2:2H2-induced TiO2 structural
leave a HCO2-covered, disordered and reduced TiO2x surface4042. transformation, in situ STEM analyses were performed in a closed

NATURE CHEMISTRY | VOL 9 | FEBRUARY 2017 | www.nature.com/naturechemistry 123

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
ARTICLES NATURE CHEMISTRY DOI: 10.1038/NCHEM.2607

a b reactivity and stability of the A-SMSI overlayer in humid


CO2-reduction conditions.
SMSI
overlayer Rh NP Relating A-SMSI to SMSI behaviour. As the traditional SMSI and
A-SMSI states both initiate via support reduction, it was
hypothesized that supports known to form SMSI overlayers would
A-SMSI also form A-SMSI overlayers. Figure 5a,b show CH4- and
Rh NP overlayer
CO-production rates, respectively, after reduction and 20CO2:2H2
treatment of 2% Rh on TiO2 , Nb2O5 , CeO2 and Al2O3 (see
3 nm 5 nm TiO2 Supplementary Figs 2527). No signicant change in CH4- or
CO-production rates was observed on Al2O3- and CeO2-
Figure 4 | Visualizing SMSI and A-SMSI states. a,b, In situ STEM images of supported Rh, as Al2O3 , an irreducible support, does not exhibit
sintered 6% Rh/TiO2 after treatment in 5% H2 and 95% N2 at 550 C for SMSI behaviour and CeO2 only exhibits SMSI behaviour under
10 min (a), which induced the formation of a TiOx SMSI crystalline bilayer, extremely harsh conditions21,22. Rh/Nb2O5 responded to
containing exclusively Ti3+, on the surface of Rh nanoparticles (NPs), and 20CO2:2H2 treatment similarly to Rh/TiO2 , exhibiting suppressed
after treatment in 20CO2:2H2 at 250 C for 3 h (b), which caused the CH4 production and increased CO production, which agrees with
formation of an amorphous A-SMSI overlayer, containing a mixture of Ti3+ the known similar SMSI behaviour of Nb2O5 and TiO2. In situ
and Ti4+, on the surface of Rh NPs. The images were collected at DRIFTS analyses agreed with the changes in reactivity, in which
atmospheric pressure. The samples were pre-sintered at 800 C in N2 to only for Nb2O5 were the CO stretching frequencies on Rh
increase the Rh particle size from 13 nm to 1050 nm and improve the signicantly redshifted because of 20CO2:2H2 treatment (see
contrast in the micrographs. Supplementary Figs 2830). In contrast to TiO2 , Nb2O5 SMSI
overlayers on Rh formed in response to high-temperature H2
gas cell at atmospheric pressure, as described before43,44, and treatments are stable in CO2-reduction reaction environments,
executed during reduction and 20CO2:2H2 treatment of the which renders metal nanoparticles inactive20. The stable SMSI
sintered 6% Rh/TiO2 catalyst. After treatment at 550 C in H2 , overlayer on Nb2O5 enables a direct comparison of the effects of
conditions known to form the traditional SMSI state, a crystalline SMSI and A-SMSI on CO2-reduction reactivity. Figure 5c shows a
bilayer of TiOx quickly formed as a conformal coating on large comparison of the impact of SMSI and A-SMSI formation on the
crystalline Rh particles, with Ti exclusively in the Ti3+ oxidation reactivity of 2% Rh/Nb2O5. SMSI formation suppressed CH4
state (Fig. 4a)26,4548. In contrast, after 20CO2:2H2 treatment for production on the Rh nanoparticles by 40-fold, but left CO
three hours at 250 C an amorphous overlayer on Rh was production on isolated Rh atoms unchanged. The formation of
observed to form (Fig. 4b and Supplementary Figs 2123). In situ the A-SMSI state decreased CH4 formation by about sixfold;
electron energy-loss spectroscopy (EELS) measurements with a however, the CO production increased by 2.5-fold, attributed to
1.01.5 spot size focused at various locations on the overlayer the HCOx-functionalized NbOx overlayer effect on the reactivity
directly proved that Ti was present in the amorphous overlayer of Rh nanoparticles.
formed from 20CO2:2H2 treatment (Supplementary Fig. 24). Ti in
the amorphous overlayer was found to exist in a combination of Discussion
Ti3+ (30%) and Ti4+ (70%) oxidation states, quantied with It is worth summarizing the evidence for the participation of HCOx
standard spectra of Ti3+ and Ti4+ using the multiple linear least adsorbates on the support in the formation and stabilization of the
squares method (Supplementary Information). The in situ STEM A-SMSI overlayer and in the modication of Rh reactivity. The cor-
and EELS analyses of 20CO2:2H2-treated Rh/TiO2 catalysts relation between the modied Rh reactivity and the appearance of
conclusively demonstrate the existence of an A-SMSI overlayer on HCOx on the support measured by in situ DRIFTS, similarity
Rh. Differences in Ti oxidation state for the traditional SMSI and between the effects of 20CO2:2H2 and formic acid treatment on
A-SMSI overlayers are hypothesized to result from the presence of Rh/TiO2 reactivity and in situ STEM/EELS results corroborate the
HCOx in the A-SMSI overlayer and to be related to the unique role of HCOx in mediating A-SMSI overlayer formation. Based on

a 50 b 30 c 20
Reduced Reduced CH4
)

)
1
Rate (mmol CH4 h1 gRh

1
Rate (mmol CO h1 gRh

40 20CO2:2H2 24 20CO2:2H2 16 CO
1
Rate (mmol h1 gRh )

30 18 12 A-SMSI

20 12 8
SMSI
10 6 4

0 0 0
Al2O3 CeO2 TiO2 Nb2O5 Al2O3 CeO2 TiO2 Nb2O5 300 C red. 400 C red. 20CO2:2H2

Figure 5 | Relating SMSI and A-SMSI behaviour. a,b, CH4 (a) and CO (b) production rates on 2% Rh on various supports after reduction and after
20CO2:2H2 treatment under standard reaction conditions (200 C, 1% CO2 , 1% H2 , 98% He). The supports were chosen for known differences in SMSI
behaviourAl2O3 and CeO2 do not exhibit SMSI behaviour at reasonable conditions, whereas TiO2 and Nb2O5 exhibit SMSI behaviour. The lack of change in
reactivity for Al2O3- and CeO2-supported Rh 20CO2:2H2 treatment and the large changes in CO and CH4 production for the TiO2- and Nb2O5-supported
samples suggests that A-SMSI overlayer formation is related to the reducibility of the support and characteristics of oxygen-vacancy formation. c, Rate of
CH4 and CO production on 2% Rh/Nb2O5 after 300 C reduction (red.), 400 C reduction and 300 C reduction followed by 20CO2:2H2 treatment. This
gure shows the reactivity differences between the SMSI and A-SMSI states for Rh/Nb2O5 , because the Nb2O5 SMSI state blocks the Rh NPs from driving
the CH4 formation, which leaves isolated Rh atoms able to produce CO, and the Nb2O5 A-SMSI state alters the Rh NP reactivity towards producing CO.
Experimental errors for the rate and selectivity measurements were consistently less than 6%.

124 NATURE CHEMISTRY | VOL 9 | FEBRUARY 2017 | www.nature.com/naturechemistry

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE CHEMISTRY DOI: 10.1038/NCHEM.2607 ARTICLES
Low temperature reduction Support reduction CO2-reduction conditions

c A-SMSI OL intact
b A-SMSI-permeable OL
HCOx
HCOx

HC
O
x
O

HC

HC
HC

Ox
Rh

Ox
Rh
250 C, 4 h,
20CO2:2H2 TiO2 or Nb2O5
TiO2 or Nb2O5 All three
a Exposed Rh catalyst
tested at
200 C, e Robust impermeable SMSI OL
1% CO2,
Rh 1% H2,
d SMSI-impermeable OL 98% He
TiO2 or Nb2O5 Rh

Nb2O5
500 C, 1 h, Rh
H2 f Receded SMSI OL
TiO2 or Nb2O5

Rh

TiO2

Figure 6 | SMSI and A-SMSI overlayer structure and behaviour. a, Schematic showing bare Rh particles on TiO2 or Nb2O5 with exposed Rh sites that favour
CH4 production. b, 20CO2:2H2-treated catalyst that forms a permeable A-SMSI overlayer composed of TiOx species and HCOx. c, The stable A-SMSI
overlayer under CO2-reduction conditions modies the Rh catalytic behaviour. d, Rh/(TiO2 or Nb2O5) after treatment with high-temperature (500 C) H2
forms an impermeable SMSI overlayer. A crystalline bilayer structure observed for TiO2 has yet to be conrmed for the Nb2O5 SMSI state. e, Reactivity
results suggest there is a stable NbOx SMSI overlayer that is impermeable in CO2-reduction conditions, which completely suppresses the reactivity of
Rh nanoparticle surfaces. f, TiOx SMSI overlayer recedes off Rh when exposed to CO2-reduction conditions because of H2O-induced re-oxidation of TiOx to
TiO2. The catalytic behaviour in f is nearly identical to that in a. OL, overlayer.

the TPD results in Fig. 2b,c and on the identication of Ti4+ in the metal. For SMSI overlayers formed with TiO2, the overlayer is oxi-
A-SMSI overlayer by in situ EELS, we hypothesize that HCOx dized rapidly in the humid environment of the CO2 + H2 reaction,
species coordinate with Ti in the overlayer, and thereby stabilize which causes recession of the support off the metal and a negating
the overlayer under the reaction conditions by decreasing the inuence on the catalytic reactivity of the underlying metal. SMSI
driving force for re-oxidation by H2O. Finally, although there is overlayers formed with Nb2O5 are stable under humid reaction con-
no direct evidence that HCOx species contribute to the modication ditions and thus suppress the reactivity of catalytic nanoparticles.
of Rh reactivity, it is clear, based on the particle-size-independent However, the A-SMSI overlayers derived from TiO2 and Nb2O5
inuence of the overlayer on Rh reactivity, that the A-SMSI effect are stable under humid reaction conditions, which enables the
on Rh reactivity is a local effect at the Rh surface A-SMSI overlayer to inuence strongly the reactivity of the catalytically
overlayer interface. active metal.
We hypothesize mechanisms by which A-SMSI overlayer for- In summary, it has been demonstrated that HCOx adsorbates on
mation could induce the observed switch in CO2-reduction selectiv- TiO2- and Nb2O5-supported Rh catalysts can induce oxygen-
ity. The downshift in CO stretching frequency could suggest that the vacancy formation in the support and drive the formation of an
CO bond is signicantly polarized, which would decrease the A-SMSI overlayer on Rh. The A-SMSI overlayer is porous (which
CORh bond strength and so allow CO to desorb before being enables access for the reactants to interact with the Rh surface)
hydrogenated2830. Another plausible mechanism is that local and stable under humid reaction conditions. The existence of the
modication of the Rh electronic structure caused by coordination A-SMSI overlayer locally modies the reactivity of the underlying
with the A-SMSI overlayer decreases the surface coverage of atomic Rh-nanoparticle surface, which opens new avenues for tuning and
hydrogen, which minimizes the driving force for CO hydrogen- controlling the reactivity of supported metal catalysts.
ation30. Finally, the transformation in the local active site environ-
ment could also change the CO2 reduction mechanism, enabling Methods
the overlayer to play a direct role as an active site7. Catalyst synthesis. All catalysts were synthesized via an aqueous wetness-
impregnation approach. The necessary mass of (Rh(NO3)3xH2O (Sigma-Aldrich
The formation mechanisms and characteristics of the A-SMSI No. 83750)) for each weight loading was dissolved in an evaporation dish with
overlayers proposed here, and the traditional SMSI overlayer distilled water. Supports were mixed with the aqueous Rh solution, dried at 95 C for
described by Tauster et al.15,16, are schematically depicted in four hours, ground with a mortar and pestle, and calcined in a tube furnace at 450 C
Fig. 6. In both overlayers, migration of the support onto the cataly- in air for four hours. To induce Rh particle-size growth, a 6% Rh/TiO2 sample was
tically active metal is induced by the oxygen-vacancy formation in heated in N2 at 800 C for three hours. All treatments to the catalysts were performed
in situ prior to kinetic and characterization analyses.
the support. For the SMSI overlayer formation, H2 treatment at
500 C induces oxygen-vacancy formation in the support, which Catalyst testing. CO2-reduction reaction rates and selectivity were measured using
drives the formation of an impermeable, crystalline and fully 1520 mg of catalyst at 200 C in a 0.25 inch (6.35 mm) outer diameter (o.d.)
reduced metaloxide overlayer on at facets of metal particles. borosilicate-packed bed reactor operated at atmospheric pressure and running under
Formation of the A-SMSI overlayer is mediated by the high coverage differential reactor conditions (conversion of limiting reagent <8%) with the efuent
quantied by online gas chromatography (GC (SRI MG #3)). CO2 was separated
of HCOx on the support at 150300 C, which causes oxygen- from other gases using a Hayesep D column, whereas H2 , N2 , CH4 and CO were
vacancy formation and drives the formation of a porous and par- separated using a molecular sieve (MS13X) packed column. The separated gases
tially reduced metaloxide overlayer on the catalytically active were then quantied using a ame ionization detector and a thermal conductivity

NATURE CHEMISTRY | VOL 9 | FEBRUARY 2017 | www.nature.com/naturechemistry 125

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
ARTICLES NATURE CHEMISTRY DOI: 10.1038/NCHEM.2607

detector. The total reactant ow rate in all the studies was 100 standard cubic cm per 14. Green, I. X., Tang, W., Neurock, M. & Yates, J. T. Jr. Spectroscopic observation of
minute (sccm). Specics regarding reactant concentrations and pre-treatment dual catalytic sites during oxidation of CO on a Au/TiO2 catalyst. Science 333,
conditions are given in the Supplementary Information. 736739 (2011).
15. Tauster, S. J., Fung, S. C. & Garten, R. L. Strong metalsupport interactions.
DRIFTS characterization. DRIFTS measurements were executed to identify how Group 8 noble metals supported on TiO2. J. Am. Chem. Soc. 100,
the various treatments explored in this study impact the species adsorbed on the 170175 (1978).
catalyst surface. Generally, catalysts were loaded into a Harrick Praying Mantis high- 16. Tauster, S. J. Strong metalsupport interactions. Acc. Chem. Res. 20,
temperature reaction chamber (ZnSe windows) mounted within a Thermo Scientic 389394 (1987).
Praying Mantis diffuse-reectance adapter, set inside a Thermo Scientic Nicolet 17. Dulub, O., Hebenstreit, W. & Diebold, U. Imaging cluster surfaces with atomic
iS10 FT-IR spectrometer with a MCT-A detector. Measurements were taken with resolution: the strong metalsupport interaction state of Pt supported on
128 scans, 4 cm1 resolution and a 3040 standard litres per minute N2 purge of the TiO2(110). Phys. Rev. Lett. 84, 36463649 (2000).
spectrometer and Praying Mantis adapter box. Typically, 10 mg of the catalyst 18. Datye, A. K., Kalakkad, D. S., Yao, M. H. & Smith, D. J. Comparison of metal
sample was packed on top of 100 mg of -alumina. The reactor efuent was routed support interactions in Pt/TiO2 and Pt/CeO2. J. Catal. 155, 148153 (1995).
to the GC to monitor the reactivity of the catalyst and ensure consistent behaviour 19. Haller, G. L. & Resasco, D. E. Metalsupport interaction: group VIII metals and
with experiments performed in the glass reactors. reducible oxides. Adv. Catal. 36, 173235 (1989).
20. Sakellson, S., McMillan, M. & Haller, G. L. EXAFS evidence for direct metal
STEM and EELS characterization. All STEM analysis was executed on a JEOL metal bonding in reduced Rh/TiO2. J. Phys. Chem. 90, 17331736 (1986).
3100R5 with a double Cs corrector operated at 300 kV. The 2% Rh/TiO2 samples 21. Deleitenburg, C. & Trovarelli, A. Metalsupport interactions in Rh/CeO2 ,
were suspended in methanol via sonication and dropcast onto lacy carbon-on- Rh/TiO2 , and Rh/Nb2O5 catalysts as inferred from CO2 methanation activity.
copper grids for ex situ analysis and a special SiN heater chip for in situ observation,
J. Catal. 156, 171174 (1995).
which utilized a Protochips Atmosphere system. The purity of gases used in the
22. Uchijima, T. SMSI effect in some reducible oxides including niobia. Catal. Today
in situ experiment was 99.9995%. All the reported temperatures are based on the
28, 105117 (1996).
Protochips calibration.
23. Boffa, A., Lin, C., Bell, A. T. & Somorjai, G. A. Promotion of CO and CO2
EELS was performed to identify the existence of Ti in the amorphous A-SMSI
hydrogenation over Rh by metal oxides: the inuence of oxide Lewis acidity and
overlayer and characterize the valance state. EELS data were collected in situ when
reducibility. J. Catal. 149, 149158 (1994).
the sample was sandwiched between two SiN membranes, with a total thickness of
80 nm, using the Gatan #965 Quantum Imaging Filter. The spot size was 1.01.5 24. Vannice, M. A. & Sen, B. Metal-support effects on the intramolecular selectivity
in diameter and a typical acquisition time was ten seconds. The Ti valence was crotonaldehyde hydrogenation over platinum. J. Catal. 115, 6578 (1989).
quantied by using multiple linear least square tting with Digital Micrograph. 25. Shi, X. Y. et al. Real-space observation of strong metalsupport interaction:
state-of-the-art and whats the next. J. Microsc. 262, 203215 (2016).
XAS characterization. XAS experiments were performed at Beamline 2-2 at the 26. Bernal, S. et al. Some contributions of electron microscopy to the
Stanford Synchrotron Radiation Light source. A double-crystal Si (220) characterisation of the strong metalsupport interaction effect. Catal. Today 77,
monochromator was used to scan X-ray energy from 100 to 200 eV and from 200 385406 (2003).
to 1,300 eV relative to the Rh K edge (23,220 eV) for X-ray absorption near-edge 27. Porosoff, M. D., Yan, B. & Chen, J. G. Catalytic reduction of CO2 by H2 for
spectroscopy and extended X-ray ne-structure (EXAFS) spectra, respectively. synthesis of CO, methanol and hydrocarbons: challenges and opportunities.
About 20 mg of 2% Rh/TiO2 catalyst was placed in 2.4 mm o.d. quartz tubes and Energy Environ. Sci. 9, 6273 (2016).
into a resistance-heating-capable Claussen cell49 with gas owing through the 28. Porosoff, M. D. & Chen, J. G. Trends in the catalytic reduction of CO2 by
powder. Rh foil was placed between the transmission and reference X-ray detectors hydrogen over supported monometallic and bimetallic catalysts. J. Catal. 301,
and measured simultaneously with the nanoparticle samples for X-ray energy 3037 (2013).
calibration and data alignment. Data processing and analysis were performed using 29. Panagiotopoulou, P., Kondarides, D. I. & Verykios, X. E. Selective methanation
the IFEFFIT package50. The amplitude and phase photoelectron-scattering functions of CO over supported noble metal catalysts: effects of the nature of the metallic
for the rst nearest neighbour photoelectron path were calculated using the FEFF6 phase on catalytic performance. Appl. Catal. A. 344, 4554 (2008).
program51 and used to t the EXAFS equation in R-space52,53. 30. Avanesian, T., Gusmo, G. S. & Christopher, P. Mechanism of CO2 reduction by
H2 on Ru(0001) and general selectivity descriptors for late-transition metal
Received 24 December 2015; accepted 5 August 2016; catalysts. J. Catal. http://dx.doi.org/10.1016/j.jcat.2016.03.016 (in the press).
published online 19 September 2016 31. Solymosi, F., Bnsgi, T. & Novk, . Effect of NO on the CO-induced disruption
of rhodium crystallites. J. Catal. 112, 183193 (1988).
32. Serna, P. & Gates, B. C. Zeolite-supported rhodium complexes and clusters:
References switching catalytic selectivity by controlling structures of essentially molecular
1. Christopher, P. & Linic, S. Engineering selectivity in heterogeneous catalysis: species. J. Am. Chem. Soc. 133, 47144717 (2011).
Ag nanowires as selective ethylene epoxidation catalysts. J. Am. Chem. Soc. 130, 33. Karelovic, A. & Ruiz, P. Mechanistic study of low temperature CO2 methanation
1126411265 (2008). over Rh/TiO2 catalysts. J. Catal. 301, 141153 (2013).
2. Studt, F. et al. Discovery of a NiGa catalyst for carbon dioxide reduction to
34. Henderson, M. A. & Worely, S. D. An infrared study of the hydrogenation
methanol. Nat. Chem. 6, 320324 (2014).
of carbon dioxide on supported rhodium catalysts. J. Phys. Chem. 89,
3. Calle-Vallejo, F. et al. Finding optimal surface sites on heterogeneous catalysts by
14171423 (1985).
counting nearest neighbors. Science 350, 185189 (2015).
35. Lundwall, M. J., McClure, S. M. & Goodman, D. W. Probing terrace and step
4. Holewinski, A., Idrobo, J.-C. & Linic, S. High-performance AgCo alloy catalysts
sites on Pt nanoparticles using CO and ethylene. J. Phys. Chem. C 114,
for electrochemical oxygen reduction. Nat. Chem. 6, 828834 (2014).
79047912 (2010).
5. Matsubu, J. C., Yang, V. N. & Christopher, P. Isolated metal active site
concentration and stability control catalytic CO2 reduction selectivity. J. Am. 36. Brabec, L. & Novkov, J. Ship-in-bottle synthesis of anionic Rh carbonyls in
Chem. Soc. 137, 30763084 (2015). faujasites. J. Mol. Catal. A. 166, 283292 (2001).
6. Behrens, M. et al. The active site of methanol synthesis over Cu/ZnO/Al2O3 37. Deshlahra, P., Conway, J., Wolf, E. E. & Schneider, W. F. Inuence of dipole
industrial catalysts. Science 336, 893897 (2012). dipole interactions on coverage-dependent adsorption: CO and NO on Pt(111).
7. Graciani, J. et al. Highly active copperceria and copperceriatitania catalysts Langmuir 28, 84088417 (2012).
for methanol synthesis from CO2. Science 345, 546550 (2014). 38. Bando, K. K., Sayama, K., Kusama, H., Okabe, K. & Arakawa, H. In-situ FT-IR
8. Farmer, J. A. & Campbell, C. T. Ceria maintains smaller metal catalyst particles study on CO2 hydrogenation over Cu catalysts supported on SiO2 , Al2O3 , and
by strong metalsupport bonding. Science 329, 933936 (2010). TiO2. Appl. Catal. A 165, 391409 (1997).
9. Ioannides, T. & Verykios, X. E. Charge transfer in metal catalysts supported on 39. Haller, G. L. et al. Geometric and electronic effects of SMSI in group VIIITiO2
doped TiO2: a theoretical approach based on metalsemiconductor contact systems. In Proc. 8th International Congress on Catalysis Vol. 5 (ed. Ertl, G.)
theory. J. Catal. 161, 560569 (1996). 135144 (1984).
10. Bruix, A. et al. A new type of strong metalsupport interaction and the 40. Henderson, M. A. Complexity in the decomposition of formic acid on the
production of H2 through the transformation of water on Pt/CeO2(111) and TiO2(110) surface. J. Phys. Chem. B 101, 221229 (1997).
Pt/CeOx/TiO2(110) catalysts. J. Am. Chem. Soc. 134, 89688974 (2012). 41. Diebold, U. The surface science of titanium dioxide. Surf. Sci. Rep. 48,
11. Campbell, C. T. Catalystsupport interactions: electronic perturbations. 53229 (2003).
Nat. Chem. 4, 597598 (2012). 42. Morikawa, Y. et al. First-principles theoretical study and scanning tunneling
12. Lykhach, Y. et al. Counting electrons on supported nanoparticles. Nat. Mater. microscopic observation of dehydration process of formic acid on a TiO2(110)
15, 284288 (2015). surface. J. Phys. Chem. B 108, 1444614451 (2004).
13. Saavedra, J., Doan, H. A., Pursell, C. J., Grabow, L. C. & Chandler, B. D. The 43. Zhang, S. et al. Dynamic structural evolution of supported palladiumceria
critical role of water at the goldtitania interface in catalytic CO oxidation. coreshell catalysts revealed by in situ electron microscopy. Nat. Commun. 6,
Science 345, 15991602 (2014). 7778 (2015).

126 NATURE CHEMISTRY | VOL 9 | FEBRUARY 2017 | www.nature.com/naturechemistry

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NATURE CHEMISTRY DOI: 10.1038/NCHEM.2607 ARTICLES
44. Zhang, S. et al. Revealing particle growth mechanisms by combining Acknowledgements
high-surface-area catalysts made with monodisperse particles and electron P.C. acknowledges funding from the University of California, Riverside, and the National
microscopy conducted at atmospheric pressure. J. Catal. 337, 240247 (2016). Science Foundation (NSF), Grant No. CHE-1301019. G.W.G. and X.P. acknowledge the
45. Bowker, M. et al. Model catalyst studies of the strong metalsupport interaction: NSF, Grants No. CBET-1159240 and No. DMR-0723032. XAS measurements were
surface structure identied by STM on Pd nanoparticles on TiO2(110). J. Catal. performed on Beamline 2-2, which was supported in part by the Synchrotron Catalysis
234, 172181 (2005). Consortium, US Department of Energy Grant No. DE-SC0012335. A. V. Dudchenko is
46. Liu, J. J. Advanced electron microscopy of metalsupport interactions in acknowledged for his efforts in Arduino automation of the packed-bed reactor
supported metal catalysts. ChemCatChem 3, 934948 (2011). experimental apparatus.
47. Logan, A. D., Braunschweig, E. J., Datye, A. K. & Smith, D. J. Direct observation
of the surfaces of small metal crystallites: rhodium supported on TiO2. Langmuir
4, 827830 (1988).
Author contributions
J.C.M. and P.C. developed the project, analysed the data and wrote the paper. J.C.M.
48. Zhang, S. et al. Dynamical observation and detailed description of catalysts
performed all the catalyst synthesis, catalyst testing and DRIFTS analysis. L.D. assisted with
under strong metal-support interaction. Nano Lett. 337, 45284534 (2016).
FTIR data collection. S.Z. performed STEM experiments. G.W.G. and X.P. assisted with
49. Chupas, P. J. et al. A versatile sample-environment cell for non-ambient X-ray
STEM data analysis. N.M. and J.G.C. performed and analysed the XAS experiments.
scattering experiments. J. Appl. Crystallogr. 41, 822824 (2008).
P.C. oversaw the project.
50. Newville, M. IFEFFIT: interactive XAFS analysis and FEFF tting. J. Synchrotron
Radiat. 8, 322324 (2001).
51. Zabinsky, S. I., Rehr, J. J., Ankudinov, A., Albers, R. C. & Eller, M. J. Additional information
Multiple-scattering calculations of X-ray-absorption spectra. Phys. Rev. B 52, Supplementary information is available in the online version of the paper. Reprints and
29953009 (1995). permissions information is available online at www.nature.com/reprints. Correspondence and
52. Beneld, R. E. Mean coordination numbers and the non-metal-metal transition requests for materials should be addressed to P.C.
in clusters. J. Chem. Soc. Faraday Trans. 88, 11071110 (1992).
53. Sasaki, K. & Marinkovic, N. in X-Ray and Neutron Techniques for Nanomaterial Competing nancial interests
Characterization (ed. Kumar, C. S. S. R.) Ch. 6 (Springer, 2016). The authors declare no competing nancial interests.

NATURE CHEMISTRY | VOL 9 | FEBRUARY 2017 | www.nature.com/naturechemistry 127

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

You might also like