Professional Documents
Culture Documents
https://doi.org/10.1038/s41929-022-00745-y
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 activity to Brønsted acid sites on partially reduced WOx (ref.14).
petrochemical, fine chemical, pharmaceutical and Therefore, it is crucial to understand the dynamics and coupling
biomass upgrade reactions1–10 owing to cooperative of acid and redox sites under various reaction environments. Aside
catalysis among multiple active sites. For example, synthesizing from the ambiguity on the active sites of these materials, the active
selectively high-value C3–C6 terminal diols and aromatics over sites are taken to be stationary.
noble metal catalysts from biomass-derived polyols and furanics In this Article, we investigate Brønsted acid site dynamics on the
by hydrodeoxygenation is challenging. The former molecules core–shell metal/metal oxide PtWOx/C inverse catalyst as a case
require cleaving secondary C–O bonds instead of accessible pri- study. The catalyst is synthesized by atomic layer deposition (ALD).
mary hydroxyls2,3,6,10; the latter entail activating the primary C–O Using density functional theory (DFT) calculations, microkinetic
bonds without reducing the aromatic groups5,7,11,12 that interact modelling (MKM) and experimental kinetics with probe molecules,
strongly with extended noble metal surfaces. Surprisingly, deco- we elucidate the structure of the WOx overlayer, the catalyst phase
rating the surface of supported noble metals (for example, Pt, Ir, behaviour and the dynamics of the acid sites under various work-
Rh) with reducible oxides (WOx, MoOx, ReOx) enables the selec- ing conditions. Unlike the metal and the oxide that are not acidic
tive production of many chemicals, such as terminal diols, and also under our conditions, we demonstrate Brønsted acid catalysis by the
C–O bond scission in m-cresol while preserving aromaticity2–9 and synergy of the elements and identify water dissociation, hydrogen
the ring-opening of tetrahydrofurfuryl alcohol5,6. These multifunc- spillover and oxide reduction as crucial contributors to the dynam-
tional catalysts consist of metal sites, redox sites and Brønsted and ics and density of Brønsted acid sites, leading to profound changes
Lewis acid sites. in the dehydration rate of probe reactions. While the acidity of these
It has been suggested that the Brønsted acid sites activate the sec- materials can be strong compared to microporous aluminosilicate
ondary C–O bond and stabilize the cationic intermediates towards zeolites, their site density is modulated by the chemistry, the oper-
the formation of diols from biomass-derived substrates such as ating conditions, the catalyst pretreatment and the longitudinal
glycerol and tetrahydrofurfuryl alcohol2,3,6,8,13. Establishing struc- position in a reactor. The fundamental insights provide methods
ture–activity relationships is complicated by the intimate coupling for modulating the sites in the selective C–O bond activation pav-
of Brønsted acidity and catalyst reducibility. Specifically, in situ ing numerous routes to catalyst design. Using the dehydration of
Brønsted acid formation is often accompanied by partial reduction tert-butanol as a case study, we demonstrate that periodic pulsing
of the oxide, leading to redox centres (vacancy formation)14. Such of hydrogen allows the PtWOx/C catalyst to achieve alternate states
redox centres have been proposed to activate the primary hydroxyl with about one order of magnitude higher activity than under con-
groups in aromatic compounds5,15,16; others have attributed the 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
d e
WO3 monolayer W L1 WO3 W L1
WO3 trimer WOx /C-1R (fresh)
1.5 1.5
WO2 trimer PtWOx /C-1R (fresh)
PtWOx /C-1R (reduced)
Normalized µ(E )
Normalized µ(E )
1.0 1.0
0.5 0.5
Simulation Experiment
0 0
12,080 12,120 12,160 12,080 12,120 12,160
Energy (eV) 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.
a c
Protonation
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
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
0.06
PH (bar)
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.
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 diagram (Fig. 2c). In the absence of hydrogen, W3O9 remains
the oxidation state of WOx. Figure 2b compares the electronic oxidized (the black region in Fig. 2c) and does not possess either
energies of reduction of W3O9 and W3O6, chosen as representative Brønsted acid or redox sites. With increasing hydrogen pressure,
examples. Reduction of W3O6 is not as thermodynamically driven the catalyst is first partially reduced to W3O7H1 over a broad tem-
as that of W3O9. Desorption of H2O (the last step of the cycle) is perature range, then protonated to W3O7H3 at low temperatures or
energetically costly. reduced to W3O5 at high temperatures with a concomitant decrease
Figure 2b depicts the energy profiles for the first protonation. in Brønsted acid density (Fig. 2d) and an increase in redox centre
We conducted comprehensive DFT calculations for subsequent density (Fig. 2e). The degree of protonation and reduction is consis-
protonations on WOx from W3O4 to W3O9 (Supplementary Table 1 tent with the DFT energy profiles (Fig. 2b).
and Supplementary Fig. 6 present all reactions and models, respec- To unravel the catalyst dynamics, we built a discrete state-based
tively). The enthalpy change for the protonation of W3O9 to W3O9H1 microkinetic model24 to simulate the catalyst state when it is
is −0.08 eV; for the protonation of W3O9H2 to W3O9H3 it is 0.43 eV exposed to different conditions. When the catalyst is not trapped
(reactions 64 and 67, respectively, Supplementary Table 1). This trend in a metastable state, the long-time kinetic behaviour (Fig. 3a) coin-
holds for reduced clusters as well. For example, protonation of W3O5 cides with the thermodynamic behaviour, as happens here (Fig. 2c).
by H spillover becomes endothermic (reaction 5, Supplementary High H2O partial pressure leads to more Brønsted acid sites at high
Table 1); H prefers binding to Pt rather than W3O5. As a result, and temperatures because equilibrium favours the dissociative adsorp-
depending on conditions, not all terminal oxygens are protonated. tion of H2O (Supplementary Fig. 7).
The protonation slows down as it proceeds, that is, it is self-inhibited. Figure 3a shows the predicted predominant catalyst state
In summary, H spillover first protonates the terminal oxygens as a function of time when the catalyst is exposed to different
of the catalyst and eventually eliminates them in the form of water, environments starting from the fully oxidized state, W3O9.
creating redox centres, whereas water hydroxylates and re-oxidizes At lower temperatures, Brønsted acid sites form quickly by
the catalyst. Interestingly, the protonation and reduction are hydrogen spillover and mild reduction occurs below 140 °C.
self-inhibited; WOx does not become either fully reduced or fully Further catalyst reduction occurs slowly at 150–200 °C and accel-
protonated. Instead, it dynamically responds to the reaction condi- erates above 200 °C, due to the higher energy barriers for the
tions, controlling site identity and density, as discussed below. The reduction of W3O7H1 to W3O5. This analysis has several impli-
protonation, reduction and oxidation kinetics depend on the degree cations. First, the in situ formation of Brønsted acid sites is
of catalyst protonation and oxidation. rapid. Second, a gradual increase in redox site density at the
expense of Brønsted acidity is anticipated at 140–250 °C. Third,
Phase diagram and dynamic behaviour. To examine the active above 250 °C, a swift reduction of WO3 leads to a significant loss of
sites under working conditions, we calculated the ab initio phase Brønsted acidity.
125 °C 5 to the WOx overlayer with a stable, partially reduced WOx overlayer.
Oxidation state
0.6 Experimental results are qualitatively consistent with model predic-
tions (Fig. 2c,e).
0.4 In situ XPS was also conducted over WOx/C-1R, PtWOx/C-1R
150 °C 4 125 °C and 10%PtWOx/C-15R. Due to the high WOx loading of
0.2 150 °C 10%PtWOx/C-15R, its W 4f signal was much stronger than that of
175 °C 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 3
0 5 10 15 20 0 5 10 15 20
WOx/C-1R undergoes minimal changes (Fig. 5b), whereas W(V) of
Time (s) 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-
Fig. 4 | Predicted kinetic effect of H2 starting from a fully oxidized W3O9 blet contribution of the 10%PtWOx/C-15R increases significantly.
state. a, Brønsted acid site density (number of OHs per W atom). b, The spectrum of PtWOx/C-1R resembles that of 10%PtWOx/C-15R
Oxidation state. Initially, the catalyst is exposed to 0.1 bar H2 and 0.1 bar (Supplementary Fig. 13). Noble-metal-promoted reduction of WOx
H2O. H2 is turned off at 10 s. The oxidation state of W is calculated from the to partially reduced 5+ and 4+ states above 400 °C was reported
number of W=O, W–O–W and W–OH bonds. 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-
Expectedly, the kinetic behaviour varies with the initial catalyst tance of Pt in modulating the oxidation state of the WOx overlayer
state. Figure 3b depicts the dynamics starting from a reduced state, by facilitating partial reduction only.
W3O5. Unlike W3O9, which suffers rapid reduction under high H2
partial pressure, W3O5 undergoes reduction at a slower rate. First, Exposing the dynamics of acidity using alcohol dehydration. We
the W3O5 domain is replaced with W3O7H1 at 120–250 °C; below have used the dehydration of 2-propanol to correlate the catalyst state
120 °C the catalyst is oxidized to W3O7H3 after 1,000 s. Up to this with activity. All samples were pre-reduced at 250 °C for 1 h to remove
point, the kinetic catalyst state agrees with the thermodynamic one any PtOx. Over the metallic sites of Pt/C, 2-propanol undergoes
except when no H2 is present. It takes 106 s to overcome the kinetic almost exclusively dehydrogenation to produce acetone (Fig. 6a,b),
barriers and fully oxidize W3O8Hx to W3O9. A slightly reduced cata- consistent with previous work10. Pt and C show negligible dehydra-
lyst would not become fully oxidized at relevant timescales. tion activity. On the other hand, WOx dehydrates 2-propanol to
To better understand the catalyst dynamics upon exposure to propene (Fig. 6b). PtWOx/C-1R exposes multifunctional behaviour:
H2 and H2O, we plot the time-dependent Brønsted acid site density WOx Brønsted acid sites dehydrate 2-propanol; Pt dehydrogenates
and oxidation state at three temperatures (Fig. 4). Upon exposure to 2-propanol to acetone and H2 and subsequently hydrogenates pro-
H2, the Brønsted acid density increases rapidly at short times owing pene—with the produced H2—to propane. The substantially lower
to fast H spillover from Pt. Consistent with the phase diagram, the dehydrogenation activity compared to Pt reveals a large fraction
maximum Brønsted acid density is anticorrelated with temperature. of Pt sites covered with WOx, in agreement with energy disper-
Interestingly, after reaching a maximum, the Brønsted acid site den- sive X-ray spectroscopy (EDX) results (Supplementary Fig. 15)6.
sity decreases due to the partial reduction of WOx (Fig. 4b). When Increasing the number of ALD cycles of W on Pt/C increases the
H2 is turned off, the Brønsted acid site density decreases sharply at dehydration activity at the expense of metal-catalysed 2-propanol
first and then gradually at lower temperatures (<125 °C). By con- dehydrogenation and propene hydrogenation.
trast, at 175 °C, the Brønsted acid site density first increases, owing To eliminate the complication from the in situ-formed hydro-
to the re-oxidation of the redox centres by H2O and eventually gen (from 2-propanol dehydrogenation) causing online catalyst
decreases via backward H spillover to Pt and H2 desorption. Overall, reduction, we investigated tert-butanol dehydration (Fig. 6a). We
the complex dynamic catalyst behaviour is found to be controlled compared the PtWOx/C-1R catalysts, reduced at 250 and 400 °C,
by rapidly forming Brønsted acid sites, slower catalyst reduction by with the untreated (fresh) catalyst. Figure 6c depicts a monotonic
H2 and even slower re-oxidation by H2O. As the amounts of H2 and decrease in dehydration rate with increasing reduction temperature
H2O change along the reactor, as well as with time and temperature, by nearly eight times due to the reduction of W(VI) to W(V) and
the catalyst protonation and oxidation state will change dynamically. W(IV) (XANES data; Fig. 5a) and the lowering of the Brønsted acid
Clearly, a static picture of acidity does not hold for these materials. site density, consistent with the predicted phase behaviour (Fig. 2c).
To disentangle the interplay of redox and Brønsted acid sites,
In situ catalyst characterization under reducing conditions. To we annealed the catalyst at 400 °C in a mixture of N2 and He. The
assess the computational predictions, we investigated the stabil- annealed catalyst is less active than the fresh one (Fig. 6d), due to the
ity and the electronic state of PtWOx/C-1R at various reduction loss of OHs by dehydration, but more active than the reduced cata-
temperatures and hydrogen partial pressures. CO chemisorption lyst. The surface OH groups are partially regenerated by steaming the
reveals that roughly 32% of the Pt surface is covered by WOx clus- annealed catalyst in 0.1 bar steam at 400 °C (grey square in Fig. 6d).
ters (Supplementary Note 1) and most of those clusters remain on The initial dehydration activities of the pretreated catalysts (Fig. 6d)
Pt after reduction and annealing treatments at 400 °C. The W L3 are in satisfactory agreement with the Brønsted acid site densities
edge energy of PtWOx/C-1R is 10,209.8 eV (Fig. 5a), close to that of measured by reactive gas chromatography with tert-butylamine
WO3 (10,209.4 eV). Mild reduction at 250 °C with 0.1 bar of H2 leads (Supplementary Fig. 16). For PtWOx/C-1R, this acid quantification
to a white line intensity decrease and a slight edge energy reduction technique is preferred over Fourier transform infrared spectroscopy
(0.2 eV), suggesting a W(VI) oxidation state (Fig. 5a). Reduction of adsorbed pyridine because of the low Brønsted acid site density
at 400 °C further reduces WOx to a state between 4+ and 5+, esti- of PtWOx/C-1R and the strong absorption of infrared by the carbon
mated based on the W L3 edge energy (Supplementary Fig. 10). support. The adsorbed pyridine cannot be observed with Fourier
By contrast, WOx/C-1R is not reduced (Supplementary Fig. 11). transform infrared spectroscopy even after a tenfold dilution with
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
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)
gcatalyst )
5 5
−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.
a
Hydrogen ON
Periodic hydrogen pulse
Hydrogen OFF
b c
3.0 1.5
No H2 cofeed No H2 cofeed
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