Articles

https://doi.org/10.1038/s41929-022-00745-y

Modulating the dynamics of Brønsted acid sites

on PtWOx inverse catalyst
Jiayi Fu1,2,11, Shizhong Liu2,11, Weiqing Zheng2,11, Renjing Huang   2,3, Cong Wang2, Ajibola Lawal   2,4,
Konstantinos Alexopoulos2, Sibao Liu   5, Yunzhu Wang2, Kewei Yu1,2, J. Anibal Boscoboinik   2,6,
Yuefeng Liu   7, Xi Liu   8, Anatoly I. Frenkel   2,9,10, Omar A. Abdelrahman   2,4, Raymond J. Gorte2,3,
Stavros Caratzoulas2 and Dionisios G. Vlachos   1,2 ✉

Brønsted acid sites on the oxide overlayers of metal–metal oxide inverse catalysts are often hypothesized to drive selective
C–O bond activation. However, the Brønsted acid site nature and dynamics under working conditions remain poorly under-
stood due to the functionalities of the constituent materials. Here we investigate the formation and the dynamics of Brønsted
acid and redox sites on PtWOx/C under working conditions. Density functional theory-based thermodynamic calculations and
microkinetic modelling reveal a complex interplay between Brønsted acid and redox sites and potentially fast catalyst dynamics
at comparable timescales to the Brønsted acid catalysed dehydration chemistry. Combining in situ characterization and probe
chemistry, we demonstrate that the density of Brønsted acid sites on the PtWOx/C inverse catalyst could be modulated by up
to two orders of magnitude by altering the reaction parameters and by the chemistry itself. We elicit an order of magnitude
increase in the acid-catalysed dehydration average reaction rate by periodic hydrogen pulsing.

M
etal–metal oxide (M–MO) catalysts are central to
petrochemical, fine chemical, pharmaceutical and
biomass upgrade reactions1–10 owing to cooperative
catalysis among multiple active sites. For example, synthesizing
selectively high-value C3–C6 terminal diols and aromatics over
noble metal catalysts from biomass-derived polyols and furanics
by hydrodeoxygenation is challenging. The former molecules
require cleaving secondary C–O bonds instead of accessible pri-
mary hydroxyls2,3,6,10; the latter entail activating the primary C–O
bonds without reducing the aromatic groups5,7,11,12 that interact
strongly with extended noble metal surfaces. Surprisingly, deco-
rating the surface of supported noble metals (for example, Pt, Ir,
Rh) with reducible oxides (WOx, MoOx, ReOx) enables the selec-
tive production of many chemicals, such as terminal diols, and also
C–O bond scission in m-cresol while preserving aromaticity2–9 and
the ring-opening of tetrahydrofurfuryl alcohol5,6. These multifunc-
tional catalysts consist of metal sites, redox sites and Brønsted and
Lewis acid sites.
It has been suggested that the Brønsted acid sites activate the sec-
ondary C–O bond and stabilize the cationic intermediates towards
the formation of diols from biomass-derived substrates such as
glycerol and tetrahydrofurfuryl alcohol2,3,6,8,13. Establishing struc-
ture–activity relationships is complicated by the intimate coupling
of Brønsted acidity and catalyst reducibility. Specifically, in situ
Brønsted acid formation is often accompanied by partial reduction
of the oxide, leading to redox centres (vacancy formation)14. Such
redox centres have been proposed to activate the primary hydroxyl
groups in aromatic compounds5,15,16; others have attributed the
activity to Brønsted acid sites on partially reduced WOx (ref.14).
Therefore, it is crucial to understand the dynamics and coupling
of acid and redox sites under various reaction environments. Aside
from the ambiguity on the active sites of these materials, the active
sites are taken to be stationary.
In this Article, we investigate Brønsted acid site dynamics on the
core–shell metal/metal oxide PtWOx/C inverse catalyst as a case
study. The catalyst is synthesized by atomic layer deposition (ALD).
Using density functional theory (DFT) calculations, microkinetic
modelling (MKM) and experimental kinetics with probe molecules,
we elucidate the structure of the WOx overlayer, the catalyst phase
behaviour and the dynamics of the acid sites under various work-
ing conditions. Unlike the metal and the oxide that are not acidic
under our conditions, we demonstrate Brønsted acid catalysis by the
synergy of the elements and identify water dissociation, hydrogen
spillover and oxide reduction as crucial contributors to the dynam-
ics and density of Brønsted acid sites, leading to profound changes
in the dehydration rate of probe reactions. While the acidity of these
materials can be strong compared to microporous aluminosilicate
zeolites, their site density is modulated by the chemistry, the oper-
ating conditions, the catalyst pretreatment and the longitudinal
position in a reactor. The fundamental insights provide methods
for modulating the sites in the selective C–O bond activation pav-
ing numerous routes to catalyst design. Using the dehydration of
tert-butanol as a case study, we demonstrate that periodic pulsing
of hydrogen allows the PtWOx/C catalyst to achieve alternate states
with about one order of magnitude higher activity than under con-
stant hydrogen feed.

Department of Chemical and Biomolecular Engineering, University of Delaware, Newark, DE, USA. 2Catalysis Center for Energy Innovation, University
1

of Delaware, Newark, DE, USA. 3Department of Chemical and Biomolecular Engineering, University of Pennsylvania, Philadelphia, PA, USA. 4Department
of Chemical Engineering, University of Massachusetts Amherst, Amherst, MA, USA. 5Key Laboratory for Green Chemical Technology of Ministry of
Education, School of Chemical Engineering and Technology, Tianjin University, Tianjin, China. 6Center for Functional Nanomaterials, Brookhaven National
Laboratory, Upton, NY, USA. 7Dalian National Laboratory for Clean Energy, Dalian Institute of Chemical Physics, Chinese Academy of Science, Dalian,
China. 8In-situ Center for Physical Science, School of Chemistry and Chemical Engineering, Shanghai Jiao Tong University, Shanghai, China. 9Department
of Material Science and Chemical Engineering, Stony Brook University, Stony Brook, NY, USA. 10Division of Chemistry, Brookhaven National Laboratory,
Upton, NY, USA. 11These authors contributed equally: Jiayi Fu, Shizhong Liu, Weiqing Zheng. ✉e-mail: vlachos@udel.edu

144 Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal

Nature Catalysis
a b
d e
WO3 monolayer W L1
WO3 trimer
1.5
WO2 trimer
Normalized µ(E )
1.0
0.5
Simulation
0 0
12,080 12,120 12,160
Energy (eV)
Fig. 1 | Model structures and XANES spectra. a, WO3 monolayer on Pt(111). b, W3O9 cluster on Pt(111). c, W3O6 cluster on Pt(111). Colour code: blue, Pt;
grey, W; red, O. d, Simulated W L1 XANES spectra of model structures. μ(E), absorption coefficient; E, photon energy of the incident X-ray. e, Experimental
W L1 XANES spectra of WO3 standard, WOx/C-1R and PtWOx/C-1R. Reduction at 400 °C. The scaled raw data without background removal are shown in
Supplementary Fig. 1. 1R indicates 1 ALD cycle.
Results
Morphology, coordination and oxidation state of WOx/Pt.
Establishing a representative structural model is important.
Although tetrahedral WOx clusters on various oxide surfaces have
been reported5, the structure of WOx on noble metals under reduc-
tive conditions has not been explored. We investigated three struc-
tural models: a monolayer on Pt(111) with W atoms coordinating
to five oxygen atoms (Fig. 1a), as previously reported17; a tetrahedral
W3O9 cluster with two terminal W centres and a bridge W centre5,
representative of oligomeric clusters on Pt(111) (Fig. 1b); and a
W3O6 trimer representing a reduced structure (Fig. 1c). Figure 1d
shows simulated X-ray absorption near edge structure (XANES)
spectra (W L1 edge). The W3O9 cluster model shows an intense
pre-edge feature, characteristic of tetrahedral W(VI) centres—this
feature is absent from the monolayer model18,19. Our experimen-
tal W L1 XANES spectra display a similar intense pre-edge fea-
ture at 12,095 eV for PtWOx/C-1R and a smaller pre-edge peak for
WOx/C-1R and WO3 (Fig. 1e), indicating more tetrahedrally coor-
dinated W centres on Pt than on carbon or a bulk oxide. The smaller
pre-edge peak is attributed to the distorted octahedral W centres.
Reducing the W(VI) trimer to a W(IV) trimer decreases the calcu-
lated pre-edge feature, also seen in the experimental spectrum of the
reduced PtWOx/C-1R (Fig. 1e). The scanning transmission electron
microscopy (STEM) characterization6 and the XANES spectra sug-
gest a sub-monolayer of WOx. Thereby, we conclude that the WOx
overlayer in PtWOx/C-1R is distinct from WO6 units in bulk WO3
and WO2 and is faithfully represented by a W3O9 trimer on a Pt(111)
slab (Fig. 1b).
Formation mechanisms of Brønsted acid sites. We have exam-
ined three formation mechanisms of Brønsted acid sites on the
PtWOx catalyst, including H2 dissociation on Pt followed by H dif-
fusion and attachment to a terminal or bridge O of the WOx cluster
(hereinafter referred to as the spillover mechanism), direct H2
Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal
Articles
c
WO3 W L1
WOx /C-1R (fresh)
1.5
PtWOx /C-1R (fresh)
PtWOx /C-1R (reduced)
Normalized µ(E )
1.0
0.5
Experiment
12,080 12,120 12,160
Energy (eV)
dissociation on WOx and direct H2O dissociation on WOx.
The dissociative adsorption of H2 on Pt has an enthalpy
change of −0.95 eV and is non-activated. Upon dissociation, H
diffusion from Pt onto the W3O9 cluster and formation of an acidic
terminal hydroxyl group (Supplementary Fig. 2a) has an activation
barrier of 0.15 eV and releases 0.08 eV of enthalpy (driving force
for spillover). Various terminal sites for the OH group have been
investigated; two examples are shown in Supplementary Fig. 3a and
Supplementary Fig. 3b. Bridging W–OH–W sites (Supplementary
Fig. 3c) are energetically less favoured than terminal ones by 1.10 eV.
The direct H2 dissociation on the WO3 monolayer has a prohibitively
high barrier of 1.97 eV (transition state shown in Supplementary
Fig. 2b), consistent with experimental reports that hydrogen
activation over WOx–C and WOx–ZrO2 catalysts requires tempera-
tures over 200 °C (refs. 6,20). The transition state for proton forma-
tion by dissociative water adsorption is shown in Supplementary
Fig. 2c. The reaction is exothermic by −0.10 eV and requires only
0.27 eV of activation. H2O dissociation creates a proton on a ter-
minal O site and a hydroxyl on a W site (the bridging OH is not
energetically favoured).
Pt-assisted self-inhibited catalyst dynamics. Figure 2a illustrates
pathways encountered in the inverse-catalyst dynamics under vari-
ous working conditions. As we show below, the extent of these sur-
face reactions governs the active site densities. The facile Pt-assisted
reduction of WO3 on several oxides below 400 °C was demonstrated
using temperature-programmed reduction (TPR) and X-ray photo-
electron spectroscopy (XPS)21,22. By contrast, bulk WO3 (without Pt)
remains fully oxidized at the same conditions23. Figure 2b depicts
the reduction/oxidation cycle on the W3O9 trimer, including H2
activation, H spillover, WOx reduction, the reverse WOx hydration
and the first-principles predicted energy profile of the cycle. Pt acti-
vates H2 and protonates W3Ox (Supplementary Fig. 4). Additional
H spillover can reduce W3Ox by H2O elimination (Supplementary
145
Articles Nature Catalysis

a c
Protonation

Brønsted acid site +1

Oxidation state –1 W3O7H3 W3O7H1
0.10
Hydrogen pressure

Reduction-I
0.08
Brønsted acid site +0 W3O5
0.06

PH (bar)
Oxidation state –2
0.04

2
Reduction-II
0.02
Brønsted acid site –1
Oxidation state –1 0 W3O9
100 150 200 250 300
Temperature (°C)
Oxidation-I d
0.10 0.86
Brønsted acid site +0

Brønsted acid density

Oxidation state +2 0.08

Oxidation-lI 0.06

PH (bar)
Water pressure

0.04

2
Brønsted acid site +1
Oxidation state +1
0.02
Hydration
0
Brønsted acid site +2 100 150 200 250 300 0
Oxidation state +0 Temperature (°C)
e
0.66
0.10

Dehydration

Redox centre density

0.08

0.06
PH (bar)

Brønsted acid site –2

Temperature

Oxidation state +0 0.04

2

Deprotonation
0.02
Brønsted acid site –1 0
Oxidation state +1 100 150 200 250 300 0
Temperature (°C)

b 1

0
TS2
Potential energy (eV)

W3O6 + H2 TS1

–1
W3O9 + H2

–2

–3
Reaction coordinate

Fig. 2 | Pathways, energy profiles of W3Ox reduction and phase behaviour. a, Possible reactions on W3Ox. Colour code: blue, Pt; grey, W; red, O; white, H.
b, Energy profiles of Brønsted acid and redox site formation by H spillover to W3O9 and W3O6 and oxide reduction. TS, transition state. c, Ab initio phase
diagram for the W3Ox/Pt(111) model. Each square represents the most abundant state under the corresponding conditions. Colour code: black squares,
W3O9; grey squares, W3O5; orange squares, W3O7H1; red squares, W3O7H3. d, Microkinetic heatmap of Brønsted acid site density (number of OHs per
W) at equilibrium. e, Microkinetic heatmap of redox centre density (number of Ws with oxidation state ≤ +4 per W) at equilibrium. Larger clusters and
heterogeneous distribution of cluster sizes provide a continuum spectrum of catalyst states. Water partial pressure is 0.1 bar. PH2, hydrogen partial pressure.

Fig. 5), forming an oxygen vacancy. The reduction of W3O9 to adsorption followed by the notably activated H2O dissociation
W3O8 is thermodynamically downhill and requires minor activa- (0.93 eV). In summary, catalyst re-oxidation is energetically harder
tion (0.3 eV). The reverse process of W3O8 oxidation entails H2O than reduction.

146 Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal

Nature Catalysis Articles
a
t=0s t = 10–2 s t=1s t = 102 s

0.10

0.08

0.06
PH (bar)
2

0.04

0.02

0
100 150 200 250 300 100 150 200 250 300 100 150 200 250 300 100 150 200 250 300
Temperature (°C)

b
t=0s t=1s t = 103 s t = 106 s

0.10

0.08

0.06
PH (bar)
2

0.04

0.02

0
100 150 200 250 300 100 150 200 250 300 100 150 200 250 300 100 150 200 250 300
Temperature (°C)

Fig. 3 | Predicted dynamics of the catalyst state. a, Catalyst state versus time (W3O9 at t = 0). b, Phase behaviour for W3O5 at t = 0. Water partial pressure
is 0.1 bar, surface-to-volume ratio = 10−20 cm−1. Black, W3O9; grey, W3O5; light grey, W3O8; orange, W3O7H1; light orange, W3O8H1; red, W3O7H3.

Notably, the kinetics of WOx reduction depends strongly on
the oxidation state of WOx. Figure 2b compares the electronic
energies of reduction of W3O9 and W3O6, chosen as representative
examples. Reduction of W3O6 is not as thermodynamically driven
as that of W3O9. Desorption of H2O (the last step of the cycle) is
energetically costly.
Figure 2b depicts the energy profiles for the first protonation.
We conducted comprehensive DFT calculations for subsequent
protonations on WOx from W3O4 to W3O9 (Supplementary Table 1
and Supplementary Fig. 6 present all reactions and models, respec-
tively). The enthalpy change for the protonation of W3O9 to W3O9H1
is −0.08 eV; for the protonation of W3O9H2 to W3O9H3 it is 0.43 eV
(reactions 64 and 67, respectively, Supplementary Table 1). This trend
holds for reduced clusters as well. For example, protonation of W3O5
by H spillover becomes endothermic (reaction 5, Supplementary
Table 1); H prefers binding to Pt rather than W3O5. As a result, and
depending on conditions, not all terminal oxygens are protonated.
The protonation slows down as it proceeds, that is, it is self-inhibited.
In summary, H spillover first protonates the terminal oxygens
of the catalyst and eventually eliminates them in the form of water,
creating redox centres, whereas water hydroxylates and re-oxidizes
the catalyst. Interestingly, the protonation and reduction are
self-inhibited; WOx does not become either fully reduced or fully
protonated. Instead, it dynamically responds to the reaction condi-
tions, controlling site identity and density, as discussed below. The
protonation, reduction and oxidation kinetics depend on the degree
of catalyst protonation and oxidation.
Phase diagram and dynamic behaviour. To examine the active
sites under working conditions, we calculated the ab initio phase
diagram (Fig. 2c). In the absence of hydrogen, W3O9 remains
oxidized (the black region in Fig. 2c) and does not possess either
Brønsted acid or redox sites. With increasing hydrogen pressure,
the catalyst is first partially reduced to W3O7H1 over a broad tem-
perature range, then protonated to W3O7H3 at low temperatures or
reduced to W3O5 at high temperatures with a concomitant decrease
in Brønsted acid density (Fig. 2d) and an increase in redox centre
density (Fig. 2e). The degree of protonation and reduction is consis-
tent with the DFT energy profiles (Fig. 2b).
To unravel the catalyst dynamics, we built a discrete state-based
microkinetic model24 to simulate the catalyst state when it is
exposed to different conditions. When the catalyst is not trapped
in a metastable state, the long-time kinetic behaviour (Fig. 3a) coin-
cides with the thermodynamic behaviour, as happens here (Fig. 2c).
High H2O partial pressure leads to more Brønsted acid sites at high
temperatures because equilibrium favours the dissociative adsorp-
tion of H2O (Supplementary Fig. 7).
Figure 3a shows the predicted predominant catalyst state
as a function of time when the catalyst is exposed to different
environments starting from the fully oxidized state, W3O9.
At lower temperatures, Brønsted acid sites form quickly by
hydrogen spillover and mild reduction occurs below 140 °C.
Further catalyst reduction occurs slowly at 150–200 °C and accel-
erates above 200 °C, due to the higher energy barriers for the
reduction of W3O7H1 to W3O5. This analysis has several impli-
cations. First, the in situ formation of Brønsted acid sites is
rapid. Second, a gradual increase in redox site density at the
expense of Brønsted acidity is anticipated at 140–250 °C. Third,
above 250 °C, a swift reduction of WO3 leads to a significant loss of
Brønsted acidity.

Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal 147

Articles
a 1.0 b 6
0.1 bar H2 No H2
0.8
Brønsted acid density
125 °C
Oxidation state
0.6
0.4
150 °C
0.2
175 °C
0 3
0 5 10 15 20
Time (s)
Fig. 4 | Predicted kinetic effect of H2 starting from a fully oxidized W3O9
state. a, Brønsted acid site density (number of OHs per W atom). b,
Oxidation state. Initially, the catalyst is exposed to 0.1 bar H2 and 0.1 bar
H2O. H2 is turned off at 10 s. The oxidation state of W is calculated from the
number of W=O, W–O–W and W–OH bonds.
Expectedly, the kinetic behaviour varies with the initial catalyst
state. Figure 3b depicts the dynamics starting from a reduced state,
W3O5. Unlike W3O9, which suffers rapid reduction under high H2
partial pressure, W3O5 undergoes reduction at a slower rate. First,
the W3O5 domain is replaced with W3O7H1 at 120–250 °C; below
120 °C the catalyst is oxidized to W3O7H3 after 1,000 s. Up to this
point, the kinetic catalyst state agrees with the thermodynamic one
except when no H2 is present. It takes 106 s to overcome the kinetic
barriers and fully oxidize W3O8Hx to W3O9. A slightly reduced cata-
lyst would not become fully oxidized at relevant timescales.
To better understand the catalyst dynamics upon exposure to
H2 and H2O, we plot the time-dependent Brønsted acid site density
and oxidation state at three temperatures (Fig. 4). Upon exposure to
H2, the Brønsted acid density increases rapidly at short times owing
to fast H spillover from Pt. Consistent with the phase diagram, the
maximum Brønsted acid density is anticorrelated with temperature.
Interestingly, after reaching a maximum, the Brønsted acid site den-
sity decreases due to the partial reduction of WOx (Fig. 4b). When
H2 is turned off, the Brønsted acid site density decreases sharply at
first and then gradually at lower temperatures (<125 °C). By con-
trast, at 175 °C, the Brønsted acid site density first increases, owing
to the re-oxidation of the redox centres by H2O and eventually
decreases via backward H spillover to Pt and H2 desorption. Overall,
the complex dynamic catalyst behaviour is found to be controlled
by rapidly forming Brønsted acid sites, slower catalyst reduction by
H2 and even slower re-oxidation by H2O. As the amounts of H2 and
H2O change along the reactor, as well as with time and temperature,
the catalyst protonation and oxidation state will change dynamically.
Clearly, a static picture of acidity does not hold for these materials.
In situ catalyst characterization under reducing conditions. To
assess the computational predictions, we investigated the stabil-
ity and the electronic state of PtWOx/C-1R at various reduction
temperatures and hydrogen partial pressures. CO chemisorption
reveals that roughly 32% of the Pt surface is covered by WOx clus-
ters (Supplementary Note 1) and most of those clusters remain on
Pt after reduction and annealing treatments at 400 °C. The W L3
edge energy of PtWOx/C-1R is 10,209.8 eV (Fig. 5a), close to that of
WO3 (10,209.4 eV). Mild reduction at 250 °C with 0.1 bar of H2 leads
to a white line intensity decrease and a slight edge energy reduction
(0.2 eV), suggesting a W(VI) oxidation state (Fig. 5a). Reduction
at 400 °C further reduces WOx to a state between 4+ and 5+, esti-
mated based on the W L3 edge energy (Supplementary Fig. 10).
By contrast, WOx/C-1R is not reduced (Supplementary Fig. 11).
148
Nature Catalysis
XANES at the Pt L2 edge indicates a readily reduced Pt/C to metal
0.1 bar H2 No H2 at 250 °C and a slightly oxidized Pt on PtWOx/C-1R even at 400 °C,
as indicated by the somewhat higher white line peak at 13,276.5 eV
(Supplementary Fig. 12). The results suggest charge transfer from Pt
5 to the WOx overlayer with a stable, partially reduced WOx overlayer.
Experimental results are qualitatively consistent with model predic-
tions (Fig. 2c,e).
In situ XPS was also conducted over WOx/C-1R, PtWOx/C-1R
4 125 °C and 10%PtWOx/C-15R. Due to the high WOx loading of
150 °C 10%PtWOx/C-15R, its W 4f signal was much stronger than that of
175 °C PtWOx/C-1R, allowing a reliable fit of the spectra. Before reduction,
all catalysts were treated under reduced pressure at 250 °C. At 250 °C,
0 5 10 15 20
WOx/C-1R undergoes minimal changes (Fig. 5b), whereas W(V) of
Time (s) 10%PtWOx/C-15R increases slightly (Fig. 5c). At 400 °C, W(VI)
doublets of the WOx/C-1R still dominate; by contrast, the W(V) dou-
blet contribution of the 10%PtWOx/C-15R increases significantly.
The spectrum of PtWOx/C-1R resembles that of 10%PtWOx/C-15R
(Supplementary Fig. 13). Noble-metal-promoted reduction of WOx
to partially reduced 5+ and 4+ states above 400 °C was reported
previously25,26 and is supported by TPR results (Supplementary
Fig. 14). XPS results are qualitatively consistent with the X-ray
absorption spectroscopy (XAS) data and demonstrate the impor-
tance of Pt in modulating the oxidation state of the WOx overlayer
by facilitating partial reduction only.
Exposing the dynamics of acidity using alcohol dehydration. We
have used the dehydration of 2-propanol to correlate the catalyst state
with activity. All samples were pre-reduced at 250 °C for 1 h to remove
any PtOx. Over the metallic sites of Pt/C, 2-propanol undergoes
almost exclusively dehydrogenation to produce acetone (Fig. 6a,b),
consistent with previous work10. Pt and C show negligible dehydra-
tion activity. On the other hand, WOx dehydrates 2-propanol to
propene (Fig. 6b). PtWOx/C-1R exposes multifunctional behaviour:
WOx Brønsted acid sites dehydrate 2-propanol; Pt dehydrogenates
2-propanol to acetone and H2 and subsequently hydrogenates pro-
pene—with the produced H2—to propane. The substantially lower
dehydrogenation activity compared to Pt reveals a large fraction
of Pt sites covered with WOx, in agreement with energy disper-
sive X-ray spectroscopy (EDX) results (Supplementary Fig. 15)6.
Increasing the number of ALD cycles of W on Pt/C increases the
dehydration activity at the expense of metal-catalysed 2-propanol
dehydrogenation and propene hydrogenation.
To eliminate the complication from the in situ-formed hydro-
gen (from 2-propanol dehydrogenation) causing online catalyst
reduction, we investigated tert-butanol dehydration (Fig. 6a). We
compared the PtWOx/C-1R catalysts, reduced at 250 and 400 °C,
with the untreated (fresh) catalyst. Figure 6c depicts a monotonic
decrease in dehydration rate with increasing reduction temperature
by nearly eight times due to the reduction of W(VI) to W(V) and
W(IV) (XANES data; Fig. 5a) and the lowering of the Brønsted acid
site density, consistent with the predicted phase behaviour (Fig. 2c).
To disentangle the interplay of redox and Brønsted acid sites,
we annealed the catalyst at 400 °C in a mixture of N2 and He. The
annealed catalyst is less active than the fresh one (Fig. 6d), due to the
loss of OHs by dehydration, but more active than the reduced cata-
lyst. The surface OH groups are partially regenerated by steaming the
annealed catalyst in 0.1 bar steam at 400 °C (grey square in Fig. 6d).
The initial dehydration activities of the pretreated catalysts (Fig. 6d)
are in satisfactory agreement with the Brønsted acid site densities
measured by reactive gas chromatography with tert-butylamine
(Supplementary Fig. 16). For PtWOx/C-1R, this acid quantification
technique is preferred over Fourier transform infrared spectroscopy
of adsorbed pyridine because of the low Brønsted acid site density
of PtWOx/C-1R and the strong absorption of infrared by the carbon
support. The adsorbed pyridine cannot be observed with Fourier
transform infrared spectroscopy even after a tenfold dilution with
Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal
Nature Catalysis Articles
a b c
6
PtWOx/C-1R (fresh) WOx/C-1R W 4f 10%PtWOx/C-15R W 4f
PtWOx/C-1R (Tr = 250 °C) Vacuum, 250 °C Vacuum, 250 °C

PtWOx/C-1R (Tr = 400 °C)

WO3
WO2

4
0.5 mbar H2, 250 °C 0.5 mbar H2, 250 °C
Normalized µ(E)

Intensity (a.u.)

Intensity (a.u.)
2
0.5 mbar H2, 400 °C 0.5 mbar H2, 400 °C

PtWOx/C-1R Fresh 250 °C 400 °C

W ox. state 6+ <6+ 4 to 5+

0
10,205 10,210 10,215 10,220 40 38 36 34 32 40 38 36 34 32
Energy (eV) Binding energy (eV) Binding energy (eV)

Fig. 5 | In situ characterization of the W oxidation state. a, W L1 XANES spectra of PtWOx/C-1R. Reduction in 0.1 bar of flowing hydrogen for 1 h. Pink
(tiny) area, standard deviations of triplicates; Tr, reduction temperature. b, W 4f XPS spectra of WOx/C-1R. c, W 4f XPS spectra of 10%PtWOx/C-15R.
Sample heated under reduced presure at 250 °C for 1 h, cooled down to room temperature overnight and reduced for 1 h in 0.5 mbar of H2 at temperatures
indicated. Black line, experimental data; grey line, background; red line, XPS fitting curve; violet line, W(VI); blue line, W(V).

a b
70
Propene
Propane
60 Acetone

50
Initial yield (%)

40

30

20

10

0
Pt/C WOx /C-1R PtWOx /C-1R PtWOx /C-2R

c d e f
Initial dehydration rate (×10–4 mol s−1 gcatalyst−1)

Initial dehydration rate (×10–4 mol s−1 gcatalyst−1)

Initial dehydration rate (×10–4 mol s−1 gcatalyst−1)

gcatalyst )

5 5
−1

No pretreatment 5 No pretreatment Annealed at 400 °C (0.1 bar H2 cofeed)

Annealed at 400 °C; 2.5 Annealed at 400 °C (no H2 cofeed)

steamed at 400 °C Reduced at 400 °C (0.1 bar H2 cofeed)
4 4 Reduced at 400 °C (no H2 cofeed)
Annealed at 400 °C 4
−1

2.0
Dehydration rate (×10 mol s

3 3 3
1.5
–4

2 2 2 1.0

1 1 1 0.5

0 0 0 0
No 250 400 0 0.05 0.15 0 0.02 0.10 0 2 4 6 8
reduction Partial pressure of water Partial pressure of hydrogen Time on stream (h)
Reduction temperature (°C) cofeed (bar) cofeed (bar)

Fig. 6 | Probing active sites with alcohols. a, Reaction schemes. b, Reaction of 2-propanol. Catalysts reduced in 0.1 bar H2 at 250 °C for 1 h and purged
with pure He before reaction. Reaction conditions: 4 mol% 2-propanol; 100 ml min−1 N2/2-propanol mixture; 10 mg catalyst; reaction temperature, 250 °C.
c, Reaction of tert-butanol. Catalysts reduced in 0.1 bar H2 for 1 h. Carrier gas, 100 ml min−1 N2/He (1:1) mixture. d, Modulating surface OHs with annealing
and water. Annealing time, 1 h; steaming time, 1 h; water partial pressure, 0.1 bar; carrier gas, 100 ml min−1 N2/He mixture or water/N2/He mixture. e, Effect
of hydrogen partial pressure on the dehydration activity of fresh PtWOx/C-1R. Carrier gas, 100 ml min−1 N2/He mixture or H2/N2/He mixture. f, Effect of
oxidation state on the in situ formation of Brønsted acid sites by hydrogen. Pretreatment time, 1 h; reduction in 0.1 bar H2; carrier gas, 100 ml min−1 N2/He
mixture or H2/N2/He mixture. Reaction conditions for c–f: 140 °C, 4 mol% tert-butanol, 10 mg catalyst. The dehydration rate accounts for the sum
of isobutene and isobutane. The initial dehydration rates are estimated by linearly extrapolating within 1 h on stream to zero reaction time. Error bars
represent the standard deviation from triplicates.

Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal 149

Articles Nature Catalysis

a
Hydrogen ON
Periodic hydrogen pulse

High Brønsted acid density

Hydrogen OFF

b c

3.0 1.5
No H2 cofeed No H2 cofeed

Average dehydration rate (×10–4 mol s−1 gcatalyst−1)

0.1 bar H2 cofeed 0.1 bar H2 cofeed
Dehydration rate (×10–4 mol s−1 gcatalyst−1)

2.5 Oscillatory H2 cofeed Oscillatory H2 cofeed

0.1 bar H2 0.1 bar H2 0.1 bar H2 0.1 bar H2

2.0 1.0

1.5

×9
1.0
0.5

0.5

0
0 4 8 12 0
Time on stream (h) 10–14 h

Fig. 7 | Impact of hydrogen pulsing on reaction rate. a, Schematic of periodic hydrogen pulse for sustaining high Brønsted acid site density.
b, Time-on-stream data of the hydrogen pulse co-feed experiment. PtWOx/C-1R is pre-reduced at 400 °C in 0.1 bar H2 for 1 h prior to reaction. Reaction
conditions, 140 °C; 4 mol% tert-butanol; 10 mg catalyst; carrier gas, 100 ml min−1 H2/N2/He mixture (0.1 bar H2) or N2/He mixture. Each oscillation cycle
lasts for approximately 4 h. The dehydration rate is the sum of formation rates of isobutene and isobutane. c, Average dehydration rates from 10 h to 14 h.

KBr (Supplementary Fig. 17). On the other hand, reactive gas chro- Tuning activity by periodic pulsing of H2. The fundamental
matography is not affected by the opacity of the support, is Brønsted insights from pretreatment and co-feed studies are consistent with
acid site specific and is capable of measuring Brønsted acid site den- the dynamic behaviour of the WOx overlayer depicted in Fig. 2b
sity below 1 µmol H+ gcatalyst−1 (ref. 27). and Fig. 3. Water hydroxylates the oxide and generates Brønsted
Water co-feed experiments (Fig. 6d) over the annealed catalyst at acid sites, slowly oxidizes the reduced WOx overlayer at 140 °C and
140 °C indicate, counterintuitively, an increased initial dehydration retards the dehydration reaction by promoting its reverse. Hydrogen
rate, despite water product enhancing the reverse, equilibrium-limited generates Brønsted acid sites by spillover at shorter times but slowly
dehydration reaction. The lack of an induction period suggests reduces the catalyst at longer reaction times. This dynamic behav-
that water rapidly hydroxylates the WOx and regenerates OHs iour of WOx can be leveraged to increase the reaction rate by pulsing
(Supplementary Fig. 18). Increasing the steam partial pressure from hydrogen (Fig. 7a).
0.05 to 0.15 bar slightly decreases the initial dehydration activity due Figure 7b shows time-on-stream data for the reduced
to either the competitive adsorption of water on the active sites or the PtWOx/C-1R. After an initial increase in the rate when exposed to
reverse reaction. The inhibiting effect of steam is also illustrated by its hydrogen, the catalyst is deactivated in the first 2 h. When the hydro-
removal after 8 h on stream (Supplementary Fig. 18). gen is switched off after 2 h, the activity of the catalyst decreases
The in situ formation of Brønsted acid sites by co-feeding hydro- immediately but gradually increases due to re-oxidization by the
gen is illustrated in Fig. 6e. While catalyst reduction adversely affects in situ formed water. Since the oxidized WOx overlayer can sustain a
the activity, on the fresh catalyst, hydrogen generates Brønsted acid higher Brønsted acid density from hydrogen spillover, the dehydra-
sites; its partial pressure correlates positively with the dehydration tion activity substantially increases when H2 is re-introduced after
activity (Fig. 6e). On the annealed catalyst, hydrogen increases 4 h. This trend repeats itself. Periodic hydrogen pulsing allows the
the initial rate threefold (Fig. 6f). By contrast, although annealed catalyst to reach alternate states with a much higher acid site den-
WOx/C-1R shows comparable dehydration activity (Supplementary sity that cannot be sustained at steady state. The average dehydra-
Fig. 19), hydrogen exposure has no promotional effect, confirming tion rate over the last 4 h is more than 8 times higher than that with
the pivotal role of Pt in forming Brønsted acid sites by hydrogen constant co-fed hydrogen (Fig. 7c). One may expect more than one
spillover. The reduced catalyst (Fig. 6f) shows distinct dynamic order of magnitude increase in catalyst activity upon optimizing the
behaviour under different environments. In hydrogen, Brønsted frequency of periodic pulsing.
acid sites form at short times, but the catalyst is gradually reduced
at longer times and becomes inactive when the hydrogen is turned Discussion
off after 8 h (Supplementary Fig. 20). Without hydrogen, the catalyst Inverse catalysts have been demonstrated in many essential appli-
is initially less active but becomes more active with time on stream cations1–10, but their active sites and dynamics have remained elu-
due to the in situ generated water via the reverse pathway shown in sive. While catalyst reduction by H2 is typical, in many of these
Fig. 2b. These dynamic changes occur at longer timescales of hours. reactions, hydrogen and water are reactants and/or products and

150 Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal

Nature Catalysis
alter the mode of action of the catalyst. Computational modelling,
in situ catalyst characterization and dehydration probe reactions
reveal a complex interplay of water dissociation, hydrogen spillover
and catalyst reduction in controlling the Brønsted acid and redox
site density of PtWOx/C-1R, a prototypical inverse catalyst. The
time-scales of various elementary steps are quite different, giving
rise to complex dynamic behaviour with time-on-stream or location
in a chemical reactor. Brønsted acid sites on PtWOx/C-1R are more
active than the acid sites in microporous materials such as zeolites
(Supplementary Fig. 21). Different from typical Brønsted acid cata-
lysts with localized protons near the oxygens adjacent to the metal
atoms with charge imbalance, whose site density is constant, the
site density here changes with reaction conditions and location in a
reactor. Excitingly, the chemistry itself can regulate the active sites
and change the rate and selectivity for various pathways.
Unlike hydrogen or other reducing agents, in situ generated
water or steaming creates Brønsted acid sites on the WOx overlayer
by direct, modestly activated dissociation without lowering the oxi-
dation state of W, as revealed by XPS (Supplementary Fig. 22). Water
can also slowly re-oxidize the catalyst, forming W–OH centres and
thus increasing the acid density. Finally, it inhibits dehydration by
promoting its backward reaction. Although direct hydrogen disso-
ciative adsorption on WOx is unfavourable below 400 °C, Pt readily
dissociates H2 and facilitates the rapid formation of Brønsted acid
sites by spillover. It can also dehydrogenate a substrate and form
hydrogen in situ. Pt stabilizes the WOx overlayer inverse structure,
by allowing partial but not complete reduction due to self-inhibition
and enables charge transfer to WOx even at 400 °C, as shown by the
EDX (ref. 6) and Pt L2 XANES spectra (Supplementary Fig. 12b).
Finally, hydrogen enables a slower reduction of WOx that increases
the density of redox centres and decreases the Brønsted acid
site density.
The temperature has a dominant effect on the distribution of
active sites. The in situ XANES (Fig. 5a) and reduction studies
(Fig. 6c) show that reducing the catalyst at high temperatures low-
ers the oxidation state to W(V) and W(IV), preventing the genera-
tion of Brønsted acid sites by hydrogen spillover. High-temperature
annealing in an inert atmosphere facilitates surface dehydroxyl-
ation, lowering the Brønsted acid density again without significantly
reducing the oxide. At low temperatures, hydrogen spillover and
hydroxylation by water (Fig. 3d) significantly increase the number
of Brønsted acid sites due to low energy barriers without signifi-
cantly changing the oxidation state.
At intermediate temperatures, slow reduction of PtWOx not only
causes loss of protons but also lowers the activity of WOx by cleaving
the W=O bond, which is essential for stabilizing and delocalizing
the charge-compensating electron28. This slow reduction is observed
using a time-dependent XANES study (Supplementary Note 2). The
state of MOx can easily be tuned by changing the ratio of metal to
the metal oxide21,29,30, the dispersion of M, the surface density of
MOx (refs. 22,31), the distance between M and MO (Supplementary
Note 2), the support reducibility, the reaction temperature and the
hydrogen pressure. In situ XPS characterization (Supplementary
Fig. 24) shows that reducible metal oxides, such as WOx, can easily
be re-oxidized in air to restore the Brønsted acidity.
The preferred Brønsted acid site formation pathway depends
on operating conditions, such as the hydrogen and water partial
pressures and the WOx coverage on Pt. As the coverage of WOx
approaches one monolayer, Brønsted acid sites could form by water
dissociation only, as the hydrogen dissociation will be hindered by
unavailable Pt sites and its high activation energy barrier on the
WOx overlayer. Our study demonstrates that the Brønsted acid site
density can significantly be modulated by changing the pretreat-
ment conditions (temperature, hydrogen pressure, steam pressure)
and by periodic hydrogen pulsing, leading to a significant activity
increase. The idea could in principle be extended to other biomass
Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal
Articles
reactions producing high-value terminal diols and aromatics, where
acid sites have been proposed to play a crucial role. The fundamen-
tal insights from this study could enable the development of more
efficient inverse catalysts. While these catalysts are far from static
and are inherently complex, simple ab initio thermodynamic and
microkinetic models in conjunction with probe reactions and char-
acterization form a blueprint for understanding and manipulating
the Brønsted versus redox properties of the catalyst and the under-
lying chemistry.
Methods
DFT calculations. Spin-polarized DFT calculations were performed using VASP
(ref. 32). Exchange and correlation effects were approximated with the Perdew–
Burke–Ernzerhof (PBE) exchange–correlation energy functional33 and the core
electrons were calculated using the projector augmented wavefunction. For the
WxOy cluster/Pt(111), we put a W3Ox cluster on Pt(111) slabs of 4 layers. In order
to eliminate periodic image interactions, we used a big Pt(111) slab (5 × 4) to
support one trimer. For the (WO3)n monolayer/Pt (111), we used a (4 × 2) Pt(111)
slab to support the (WO3)4 2D monolayer structure (Fig. 1a). For each calculation,
the bottom two layers of the Pt slab were fixed at the bulk position while the top
two layers with the supported oxide cluster and adsorbates were fully relaxed.
We applied a Monkhorst–Pack mesh34. The k-points were (3 × 2 × 1) and (2 ×
4 × 1) for the 5 × 4 slabs and the monolayer supercells, respectively. Transition
state searches were carried out using the climbing-image VTST NEB method35.
For geometry optimization calculations, the cut-off energy was set to 520 eV and
the energy and force convergence criteria were set to 10−5 eV and 0.01 eV Å−1,
respectively. In the NEB calculations, the force criterion was set to 0.025 eV Å−1.
We used pMuTT (ref. 36), together with the DFT-calculated reaction energies and
vibrational frequencies, to simulate a phase diagram.
Simulated phase diagrams. Thermodynamic phase diagrams were generated
using the PhaseDiagram function in pMuTT (ref. 36). DFT reaction energies and
frequencies were used to calculate Gibbs free energies. The cluster model we used
has a formula of W3OxHy·nH2O (x = 4 to 10, y = 0 to 5, n = 0 or 1) (Supplementary
Table 1). The PhaseDiagram function in pMuTT generates a graph of grids, where
each grid represents the most stable state under a certain reaction condition.
State-based MKM24 was performed using the CHEMKIN package. A batch
reactor was chosen to study the reduction kinetics of WOx/Pt using H2/H2O
mixed gas at different temperatures. Two sites were considered in this model: the
Pt(111) site and the WOx site. The Pt site is responsible for H2 activation, while the
tungsten oxides can be hydrogenated and hydrated to generate hydroxyl groups.
H-spillover from Pt to WOx sites can also reduce the oxide by forming H2O. Due to
the huge set of structures of all possible hydrogenated/hydrated tungsten oxides, we
did not calculate all activation energies in the reaction network using DFT. Instead,
we estimated them using Brønsted–Evans–Polanyi relationships developed herein
from a smaller subset of DFT calculations (Supplementary Fig. 25). To keep the
chemical potential of H2 and H2O constant during the reaction, we chose a small
surface area to volume ratio of 10−20 cm−1 for the batch reactor. For the calculations
starting with fully oxidized W3O9 clusters, the reaction time for every condition in
MKM was set to 36,000 s. For the re-oxidation of W3O5 clusters, the MKM reaction
times were set to 106 s. The Brønsted acid density heatmaps were generated by
calculating the OH group coverages at each reaction condition at long times. The
heatmaps were smoothed by interpolation.
XANES simulation. XANES simulation of W atoms was carried out using
FDMNES (refs. 37,38), where the convoluted L1 edge spectra of all W atoms in
each supercell were calculated using the muffin-tin approximation with a radius
of 4.0 Å. To account for broadening effects (for example, core–hole lifetime), a
Lorentzian energy-dependent convolution was applied to the calculated spectrum.
The energy range was set to −20 to 80 eV relative to the Fermi level. The simulated
spectra were aligned with the experimental spectra of W standards (WO3 and
WO2) using the first inflection point.
Catalyst synthesis. The 4 wt% Pt/C catalyst was prepared by the conventional
incipient wetness impregnation method, as described previously5,12. First, 0.08 g Pt
precursor (Pt(NH3)4(NO3)2, 99.99%, Alfa Aesar) was dissolved in an 8 g water/
ethanol (mass ratio = 3:1) solution. Then, 0.96 g carbon black (225 m2 g−1, Vulcan
XC-72R) was added to the solution and it was kept at room temperature for 12 h.
After drying the catalyst in an oven at 60 °C for 12 h, the resulting powder was
reduced in a 7-mm-inner diameter quartz tube with 60 ml min−1 He and 5 ml min−1
H2 at 500 °C for 2 h using a heating ramp of 3 °C min−1.
The WOx overlayer was prepared by ALD using a home-built, static system, as
described previously5,6,39. In a typical experiment, the tungsten precursor (W(CO)6,
99%, Strem) was first evacuated and then heated to 130 °C. The W(CO)6 vapour
was then introduced to approximately 0.3 g catalyst, which was evacuated and
held at 200 °C before the deposition, for 1.5 min. Next, the catalyst chamber was
evacuated for 5 min to remove excess precursors. The catalyst was then exposed to
151
Articles
air for 6 min to oxidize the adsorbed W(CO)6 precursor. After completing all ALD
cycles, the catalyst was treated twice by rapid thermal annealing (400 °C, 1 min) to
ensure the complete removal of ligands. Samples prepared using n ALD cycles are
referred to here as PtWOx/C-nR.
10%PtWOx/C-15R was prepared by the same procedure as PtWOx/C-nR using
0.2 g Pt precursor and 0.9 g carbon black. WOx/C-1R was prepared by the same
procedure as PtWOx/C-nR except that the carbon black was impregnated with
water/ethanol solution without the addition of Pt precursor. The carbon black
samples, used as the reference for reaction and characterization, were prepared
similarly by impregnation without salts, reduction and finally rapid thermal
annealing. The actual Pt and W loadings were determined using inductively
coupled plasma–atomic emission spectrometry (PerkinElmer, Optima 5300V) by
Galbraith Laboratories.
Kinetics experiments. Kinetics experiments were performed using a fixed-bed
reactor at near ambient pressure. The reactor was a quartz tube placed in the
middle of a resistance furnace. For a typical experiment, the catalyst bed was
fixed between two inert quartz beds. The catalyst bed was placed in the constant
temperature zone of the furnace to minimize the axial temperature gradient. The
temperature of the catalyst bed was measured using a thermocouple covered by a
quartz sleeve. Three mass flow controllers were used to feed a mixture of nitrogen,
hydrogen and helium at specified ratios. The typical flow rate of the gas mixture
was 100 ml min−1 at room temperature.
The alcohol flow rate was adjusted by controlling the temperature of the
bubbler and the flow rates of the carrier gas (nitrogen, 99.999% purity) and the
balance gas (helium, 99.999% purity). All the pipelines containing alcohol and
water were heated to above 150 °C to avoid condensation. The reactor outlet was
connected to a gas chromatograph (Agilent, 7890A) equipped with a Plot-Q
column and a flame ionization detector. For a typical experiment, the catalyst was
pre-reduced with a 100 ml min−1 H2–He mixture (10% H2) at 250 °C for 1 h with a
ramping rate of 10 °C min−1. The reactor was then purged with He for 1 h before
the reaction.
In situ XAS. XAS measurements were performed on beamline 5BM-D at the
Advanced Photon Source, Argonne National Laboratory. Spectra collected in
transmission mode were recorded using ionization chambers. Samples were
pressed into pellets and loaded into a three-well sample holder. The sample holder
was then placed in a tubular reactor with two Kapton windows at the ends for
in situ reduction studies. After purging the reactor with a flowing hydrogen–
helium mixture (10% hydrogen) and acquiring the spectra of the fresh samples,
the reactor was heated to the desired reduction temperature. The XAS spectra of
the reduced samples were collected in situ at the desired reduction temperatures
under a flowing hydrogen–helium mixture after 1 h reduction. X-ray absorption
near edge structure (XANES) spectra were recorded at the W L1 edge, W L3 edge
and Pt L2 edge and extended X-ray absorption fine structure (EXAFS) spectra were
recorded at the W L3 edge. The XAS spectra of PtWOx/C-1R were also collected
before and after reduction at 250 °C in fluorescence mode. The measurements
were performed using a Linkam cell and a Vortex-ME4 silicon drift detector. The
time-dependent XANES spectra for PtWOx/C-1R at W L1 and W L3 edges were
recorded on the Quick X-ray Absorption and Scattering (QAS) beamline at the
National Synchrotron Light Source (NSLS) II, Brookhaven National Laboratory
(BNL). Fluorescence spectra were measured using a passivated implanted planar
silicon (PIPS) detector. The spectra of reference foils were measured using
ionization chambers. XAS data processing was performed using the Athena
(0.9.26) program following standard procedures40. The first maximum of the first
derivative of the raw XAS data was selected to be the edge energy.
In situ XPS. The near-ambient pressure X-ray photoelectron spectroscopy
(NAP-XPS) experiments were performed at the Center for Functional
Nanomaterials (CFN), Brookhaven National Laboratory (BNL). Catalyst samples
were scanned with a laboratory-based spectrometer (Phoibos NAP 150, SPECS
Surface Nano Analysis). The instrument was equipped with a monochromated
Al Kα (1,486.6 eV) photon source and a differentially pumped hemispherical
analyser (SPECS Phoibos). Each powder sample was pressed to a cleaned thin Cu
foil and transferred into the main chamber. Pt 4f, W 4f, O 1s and C 1s spectra were
collected during all procedures. Each sample was heated to 250 °C under reduced
pressure for 1 h to remove adsorbed surface species. During in situ reduction and
the following oxidation studies, hydrogen or oxygen were introduced into the main
chamber through a variable leak valve. The pressure of the main chamber was
maintained at 0.5 mbar. XPS data analysis was performed using CasaXPS version
2.3.23. The peak model parameters are shown in Supplementary Table 4.
Brønsted acid site quantification. The Brønsted acid site density of PtWOx/C-1R
was measured under different gaseous environments using reactive gas
chromatography27, equipped with H2 or He to simulate reducing or annealing
treatment environments, respectively. Initially, 20 mg of catalyst was sandwiched
between two plugs of deactivated glass wool (Restek, 24324) and placed within a
quartz inlet liner (Agilent, 5062-3587). The catalyst-containing liner was placed
in the inlet of a gas chromatograph (Agilent, 7890B) equipped with an HP-PlotQ
152
Nature Catalysis
column (Agilent, 19091P-Q04) and a flame ionization detector. Unless otherwise
noted, the catalyst was pre-treated in situ at 400 °C under a flow of H2, pure He
or 10% H2O/He mixture to reduce, anneal or steam the catalyst, respectively. The
pre-treated catalyst was then saturated at 140 °C with 10 consecutive 1 µl pulses of
tert-butylamine, delivered at a rate of 1 µl min−1 using an automated liquid sampler.
The saturated catalyst was then purged in 1,000 sccm of He to remove any weakly
bound tert-butylamine. The catalyst temperature was then ramped up to 400 °C at
10 °C min−1 and held for 30 min, with the chromatography column held at 30 °C to
trap desorbing butenes resulting from the Hofmann elimination of tert-butylamine
adsorbed on Brønsted acid sites. The temperature of the chromatography column
was then ramped to 270 °C eluting any trapped isobutene which was then
quantified and taken as a direct measure of the Brønsted acid site density.
Data availability
The data to support the findings of this study are provided in the Supplementary
Information and paper or from the corresponding author upon request. The atomic
coordinates of the DFT-optimized WxOy/Pt(111) models and data that support the
plots in this paper are available on Mendeley Data41.
Code availability
The code to convert ab initio data to microkinetic model inputs is available on
Mendeley Data41. No specialized, home written software was used for this work.
Received: 23 March 2021; Accepted: 13 January 2022;
Published online: 21 February 2022
References
1. Védrine, J. C. Heterogeneous catalysis on metal oxides. Catalysts 7, 341 (2017).
2. Tomishige, K., Nakagawa, Y. & Tamura, M. Selective hydrogenolysis and
hydrogenation using metal catalysts directly modified with metal oxide
species. Green Chem. 19, 2876–2924 (2017).
3. Chia, M. et al. Selective hydrogenolysis of polyols and cyclic ethers over
bifunctional surface sites on rhodium–rhenium catalysts. J. Am. Chem. Soc.
133, 12675–12689 (2011).
4. Daniel, O. M. et al. X-ray absorption spectroscopy of bimetallic Pt–Re
catalysts for hydrogenolysis of glycerol to propanediols. ChemCatChem 2,
1107–1114 (2010).
5. Wang, C. et al. Mechanistic study of the direct hydrodeoxygenation of
m-cresol over WOx-decorated Pt/C catalysts. ACS Catal. 8, 7749–7759 (2018).
6. Wang, C. et al. A study of tetrahydrofurfuryl alcohol to 1,5-pentanediol over
Pt–WOx/C. Catal. Letters 148, 1047–1054 (2018).
7. Yang, F. et al. Enhancement of m-cresol hydrodeoxygenation selectivity
on Ni catalysts by surface decoration of MoOx species. ACS Catal. 9,
7791–7800 (2019).
8. Zhou, W. et al. Insight into the nature of Brönsted acidity of Pt-(WOx)n-H
model catalysts in glycerol hydrogenolysis. J. Catal. 388, 154–163 (2020).
9. Stephens, K. J. et al. A mechanistic study of polyol hydrodeoxygenation over
a bifunctional Pt-WOx/TiO2 catalyst. ACS Catal. 10, 12996–13007 (2020).
10. Lei, N. et al. Effective hydrogenolysis of glycerol to 1,3-propanediol over
metal-acid concerted Pt/WOx/Al2O3 catalysts. ChemCatChem 11,
3903–3912 (2019).
11. Fu, J. et al. C–O bond activation using ultra-low loading noble metal catalysts
on moderately reducible oxides. Nat. Catal. 3, 446–453 (2020).
12. Luo, J. et al. Mechanisms for high selectivity in the hydrodeoxygenation
of 5-hydroxymethylfurfural over PtCo nanocrystals. ACS Catal. 6,
4095–4104 (2016).
13. Falcone, D. D. et al. Evidence for the bifunctional nature of Pt–Re catalysts
for selective glycerol hydrogenolysis. ACS Catal. 5, 5679–5695 (2015).
14. Rasmussen, M. J. & Medlin, J. W. Role of tungsten modifiers in bimetallic
catalysts for enhanced hydrodeoxygenation activity and selectivity. Catal. Sci.
Technol. 10, 414–423 (2020).
15. Murugappan, K. et al. Operando NAP-XPS unveils differences in MoO3 and
Mo2C during hydrodeoxygenation. Nat. Catal. 1, 960–967 (2018).
16. Deo, S., Medlin, W., Nikolla, E. & Janik, M. J. Reaction paths for
hydrodeoxygenation of furfuryl alcohol at TiO2/Pd interfaces. J. Catal. 377,
28–40 (2019).
17. Li, Z. et al. Growth of ordered ultrathin tungsten oxide films on Pt(111).
J. Phys. Chem. C. 115, 5773–5783 (2011).
18. Horsley, J. A., Wachs, I. E., Brown, J. M., Via, G. H. & Hardcastle, F. D.
Structure of surface tungsten oxide species in the WO3/Al2O3 supported oxide
system from X-ray absorption near-edge spectroscopy and Raman
spectroscopy. J. Phys. Chem. 91, 4014–4020 (1987).
19. Yamazoe, S., Hitomi, Y., Shishido, T. & Tanaka, T. XAFS study of tungsten
L1- and L3-edges: structural analysis of WO3 species loaded on TiO2 as a
catalyst for photo-oxidation of NH3. J. Phys. Chem. C. 112, 6869–6879 (2008).
20. Baertsch, C. D., Soled, S. L. & Iglesia, E. Isotopic and chemical titration of
acid sites in tungsten oxide domains supported on zirconia. J. Phys. Chem. B
105, 1320–1330 (2001).
Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal
Nature Catalysis
21. He, J. et al. Synthesis of 1,6-hexanediol from cellulose derived
tetrahydrofuran-dimethanol with Pt-WOx/TiO2 catalysts. ACS Catal. 8,
1427–1439 (2018).
22. Iglesia, E. et al. Selective isomerization of alkanes on supported tungsten
oxide acids. Stud. Surf. Sci. Catal. 101, 533–543 (1996).
23. Fouad, N. E., Attyia, K. M. E. & Zaki, M. I. Thermogravimetry of WO3
reduction in hydrogen: kinetic characterization of autocatalytic effects.
Powder Technol. 74, 31–37 (1993).
24. Alexopoulos, K., Wang, Y. & Vlachos, D. G. First-principles kinetic
and spectroscopic insights into single-atom catalysis. ACS Catal. 9,
5002–5010 (2019).
25. Regalbuto, J. R., Fleisch, T. H. & Wolf, E. E. An integrated study of Pt/WO3/
SiO2 catalysts for the NO–CO reaction: i. catalyst characterization by XRD,
chemisorption, and XPS. J. Catal. 107, 114–128 (1987).
26. Kammert, J. D., Brezicki, G., Miyake, N., Stavitski, E. & Davis, R. J. Reduction
of propanoic acid over Pd-promoted supported WOx catalysts. ChemCatChem
12, 314–325 (2020).
27. Abdelrahman, O. A. et al. Simple quantification of zeolite acid site
density by reactive gas chromatography. Catal. Sci. Technol. 7,
3831–3841 (2017).
28. Bernholc, J., Horsley, J. A., Murrell, L. L., Sherman, L. G. & Soled, S. Brønsted
acid sites in transition metal oxide catalysts: modeling of structure, acid
strengths, and support effects. J. Phys. Chem. 91, 1526–1530 (1987).
29. Liu, L. et al. Selective hydrogenolysis of glycerol to 1,3-propanediol over
rhenium-oxide-modified iridium nanoparticles coating rutile titania support.
ACS Catal. 9, 10913–10930 (2019).
30. Liu, S., Amada, Y., Tamura, M., Nakagawa, Y. & Tomishige, K. One-pot
selective conversion of furfural into 1,5-pentanediol over a Pd-added
Ir-ReOx/SiO2 bifunctional catalyst. Green Chem. 16, 617–626 (2014).
31. Shi, D. et al. WOx supported on ɣ-Al2O3 with different morphologies as
model catalysts for alkanol dehydration. J. Catal. 363, 1–8 (2018).
32. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio
total-energy calculations using a plane-wave basis set. Phys. Rev. B 54,
11169–11186 (1996).
33. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation
made simple. Phys. Rev. Lett. 77, 3865–3868 (1996).
34. Pack, J. D. & Monkhorst, H. J. “Special points for Brillouin-zone
integrations”—a reply. Phys. Rev. B 16, 1748–1749 (1977).
35. Henkelman, G., Uberuaga, B. P. & Jónsson, H. A climbing image nudged
elastic band method for finding saddle points and minimum energy paths.
J. Chem. Phys. 113, 9901–9904 (2000).
36. Lym, J., Wittreich, G. R. & Vlachos, D. G. A python multiscale
thermochemistry toolbox (pMuTT) for thermochemical and kinetic
parameter estimation. Comput. Phys. Commun. 247, 106864 (2020).
37. Joly, Y. X-ray absorption near-edge structure calculations beyond the
muffin-tin approximation. Phys. Rev. B 63, 125120 (2001).
38. Bunǎu, O. & Joly, Y. Self-consistent aspects of X-ray absorption calculations.
J. Phys. Condens. Matter 21, 345501 (2009).
39. Onn, T. M. et al. Smart Pd catalyst with improved thermal stability supported
on high-surface-area LaFeO3 prepared by atomic layer deposition. J. Am.
Chem. Soc. 140, 4841–4848 (2018).
40. Ravel, B. & Newville, M. ATHENA, ARTEMIS, HEPHAESTUS: data analysis
for X-ray absorption spectroscopy using IFEFFIT. J. Synchrotron Radiat. 12,
537–541 (2005).
41. Fu, J. et al. Dataset for modulating the dynamics of Brønsted acid sites on
PtWOx inverse catalyst. Mendeley Data https://doi.org/10.17632/hp45n9ykd6.1
(2021).
Nature Catalysis | VOL 5 | February 2022 | 144–153 | www.nature.com/natcatal
Articles
Acknowledgements
This work was financially supported primarily by the Catalysis Center for Energy
Innovation, an Energy Frontier Research Center funded by the US Department of
Energy, Office of Science, Office of Basic Energy Sciences under award number
DESC0001004. Portions of this work were performed at the DuPont–Northwestern–Dow
Collaborative Access Team (DND-CAT) located at Sector 5 of the Advanced Photon
Source. DND-CAT is supported by Northwestern University, E.I. DuPont de Nemours
and Co. and The Dow Chemical Company. This research used resources of the Advanced
Photon Source, a US Department of Energy, Office of Science user facility operated
for the Department of Energy, Office of Science by Argonne National Laboratory
under contract no. DE-AC02-06CH11357. The research was carried out in part at the
7-BM (QAS) beamline of the National Synchrotron Light Source II at Brookhaven
National Laboratory, which is supported by the Synchrotron Catalysis Consortium,
US Department of Energy under grant no. DE-SC0012335 and at the Center for
Functional Nanomaterials, Brookhaven National Laboratory, which is supported by
the US Department of Energy, Office of Basic Energy Sciences under contract no.
DE-SC0012704. Some of the XPS measurements were carried out using the Thermo
Scientific K-Alpha+ XPS System at the University of Delaware surface analysis facility
supported by NSF (no. 1428149). Y.L. acknowledges financial support from the Chinese
Academy of Science Youth Innovation Promotion Association (no. 2018220) and
LiaoNing Revitalization Talents Program (no. XLYC1907053), and X.L. acknowledges
financial support from the National Natural Science Foundation of China (nos.
21872163, 22072090, 21991153 and 21991150) for the annular dark-field (ADF)-STEM
analysis. The authors acknowledge H. Matsumoto for assistance with STEM data
acquisition from a Hitachi HF5000 microscope. We acknowledge W. Wu for assisting
with the XPS measurements, Q. Ma, S. Ehrlich, N. S. Marinkovic and L. Ma for helping
with the XAS measurements and S. Prodinger for valuable discussions.
Author contributions
J.F. performed experimental kinetics, XAS and XPS analyses. S.L. carried out the DFT
calculations, the XANES simulation and the microkinetic modelling. W.Z. performed
XPS and XAS studies. R.H. and C.W. synthesized the catalysts. A.L. conducted reactive
gas chromatography. K.A. provided guidance for XANES simulation. K.A. and S.L.
assisted with XAS measurements. S.L., Y.W. and J.A.B. conducted the XPS measurements.
K.Y. performed the TPR measurements. Y.L. and X.L. carried out the ADF-STEM
analysis. A.I.F. provided guidance for the XAS data analysis. O.A.A. provided guidance
for the acid site density measurements. R.J.G. provided insights into the experimental
work. S.C. guided the DFT calculations. D.G.V. directed the project and provided
guidance for the experimental and theoretical work. The manuscript was written by J.F.,
S.L. and D.G.V. with input from all authors.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s41929-022-00745-y.
Correspondence and requests for materials should be addressed to Dionisios G. Vlachos.
Peer review information Nature Catalysis thanks Gianfranco Pacchioni and the other,
anonymous, reviewer(s) for their contribution to the peer review of this work.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2022
153