CatalysisScience &Technology

PAPER

Cite this: DOI: 10.1039/c5cy00493d

The influence of catalyst amount and Pd loadingon the H2O2 synthesis from hydrogen and

rcA

me

by

acid

con

tive

ase

con

g

2 c

s

ically favored. This adversely afthe yield of H2O2.

Therefore, the selectivity towfor a viable industrial applicatiothe reaction conditions, trying treactions.313 The hydrogenaappears the major route of idesign is the main tool to impH2O2.

10,19 Researchers investiga

U

d

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

.

View Article OnlineView JournalDipartimento di Scienze Chimiche, University of Padova, via Marzolo 8, 35131,

Padova, ItalydDepartment of Chemistry, Ume University, SE-90187 Ume, SwedeneMicroelectronics and Materials Physics Laboratories, University of Oulu,

FI-90014 Oulu, FinlandThis journal is The Royal Society of Chemistry 2015

Electronic supplementary information (ESI) available: See DOI: 10.1039/c5cy00493d Scheme 1 Reactions involved in the direcaDipartimento di Ingegneria Industriale,

35131, Padova, ItalybDepartment of Chemical Engineering, b

Biskopsgatan 8, bo-Turku, 20500, Finlancthe industry. Its direct productithesis, DS) has been attractingof the potentially reduced costsAlthough the DS is the simpleinvolves multiple paths, Schemallel reactions leading to water,n from H2 and O2 (direct syn-considerable interest becauseand flexibility of its process.1,2

t route to H2O2, the reactione 1, with consecutive and par-which is by far thermodynam-fects the selectivity and limits

ards H2O2 is the first issuen. Many Authors investigatedo suppress the undesired sidetion of hydrogen peroxidets consumption.1418 Catalystrove the selective synthesis ofted the catalyst properties to

cially when combined to each other, but corrosion and metalleaching could ensue from their use. In addition, halide ionscan be detrimental in down-stream applications. The oxida-tion state of palladium has been recognized to affect theselectivity. However, there is not yet consensus on its actualactive form during the H2O2 DS. According to Choudharyet al.20,21 PdO is the active form, whereas Fierro et al.22 con-cluded that PdII) ions are most selective towards H2O2 whena Pd catalyst supported on a strongly acidic resin was used.Burato et al.23 adopted the same catalyst formulation asFierro's one, but trasformed it into Pd(0) upon pre-reductionof the metal; they suggest that the metallic nanoparticles arethe most active for peroxide formation. More recently,Edwards et al.19 reported that the Pd(0) is mainly responsiblefor the hydrogenation reaction, whereas PdII) is active in pro-ducing the peroxide. Melada et al.24 investigated Pd

niversity of Padova, via Marzolo 9,

o Akademi University,

. E-mail: bpierdom@abo.fimentally friendly and efficient oxidizing agents available too

Br ) are known to be effective selectivity enhancers, espe-Received 2nd April 2015,Accepted 3rd May 2015

DOI: 10.1039/c5cy00493d

www.rsc.org/catalysis

oxygen

Nicola Gemo,ab Stefano StePaolo Canu,a Marco Zecca,c

and Jyri-Pekka Mikkolabd

Palladium catalysts with an active

macroporous resin were prepared

out in the absence of promoters (

varied by changing A) the catalyst

cases, smaller amounts of the ac

with option B) better than A). In c

to decrease at lower catalyst Pd

HL2/Pd molar ratio, the metal loadin

tivity, which reached 57% at 60% H

Introduction

Hydrogen peroxide is considered one of the most environ-hele,bc Pierdomenico Biasi,*bd Paolo Centomo,c

ndrey Shchukarev,d Krisztin Kords,e Tapio Olavi Salmib

tal content from 0.3 to 5.0 wt.% and supported on a strongly acidic,

ion-exchange/reduction method. H2O2 direct synthesis was carried

s and halides). The total Pd amount in the reacting environment was

centration in the slurry and B) the Pd content of the catalyst. In both

metal enhance the selectivity towards H2O2, at any H2 conversion,

A), the PdII)/Pd0) molar ratio (XPS) in the spent catalysts was found

tent. With these catalysts and this experimental set-up the dynamic

and the metal particle size were the key factors controlling the selec-

onversion, and 80% at lower conversion.

limit the undesired side reactions.1,13 Pd has proven to bethe best active metal, also on standard supports such as car-bon, SiO2 or Al2O3. Mineral acids and halides (either Cl

or 1,13Catal. Sci. Technol.

t synthesis of H2O2.

supported on a modified zirconia, exposing the reduced cata-

and spent catalysts. A strongly acidic resin (LewatitK2621)24,25,29,30 was used to synthesize catalysts with differ-

Fig. 1 TEM images of the X-Pd/K2621 catalysts (a: fresh, left side; b:spent, right side) with different Pd loadings: 1) 0.3-Pd/K2621; 2) 0.5-Pd/K262; 3) 1.0-Pd/K2621; 4) 2.5-Pd/K2621; 5) 5.0-Pd/K2621.

Fig. 2 Particle size distribution of the Pd/K2621 catalysts withdifferent Pd loading, before (dark) and after (red) the reaction.

Catalysis Science & TechnologyPaper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article Onlineent Pd loadings, ranging from 0.3 to 5.0 wt.%. These catalystsalready demonstrated good selectivity in the H2O2 DS.

2224

The correlation between the metal nanoparticles size, theH2/Pd ratio, the oxidation state of the catalyst and its activitytowards the hydrogen peroxide direct synthesis is discussed.The size distribution and oxidation state of Pd nanoparticleswere provided by TEM and XPS.

The ultimate focus of this work is on the reaction mecha-nism, still uncertain, where the role of the Pd/H2 ratio on theselectivity appears a key factor. Here, the aim of the work isto understand if the reactions involved are structure sensi-tive, trying to get more insights in the understanding of thischallenging reaction mechanism.

Results and discussionCatalyst characterization method

TEM images, Fig. 1, illustrate the morphological properties ofthe Pd nanoparticles in the X-Pd/K2621 catalysts (X = 0.3, 0.5,1.0, 2.5 and 5.0 represents the Pd wt.%), before and after use.Spherical nanoparticles uniformly distributed in the supportare observed for all the catalysts. The particle size distribu-tions determined from image analysis are shown in Fig. 2.

Nanoparticles diameter ranged from 2 to 12 nm. The aver-age nanoparticles diameters are summarized in Table 1.Interestingly, the nanoparticles size of the fresh catalystshifts to smaller values as the amount of palladiumincreased. This could be explained by the features of the acidgroups of the resin. In the catalysts with the higher Pd con-tent, ion-exchange of the PdII) ions involved the sulfonicgroups located both on the pores walls and in the layerjust beneath them, which is to some extent swollen andaccessible in the aqueous environment.27,28 Hence, in thelyst to mildly oxidizing conditions, and they concluded thatreduced Pd nanoparticles with a PdO-rich surface favored theH2O2 direct synthesis. More recently, Lunsford et al.

25

observed no activity over the oxidized form of a Pd/SiO2 cata-lyst, but 60% selectivity was obtained with the reduced formof this catalyst. In this context, subtle changes in the activenanoparticles surface during the reaction should be tackled.For instance, the oxidation state could be affected by thereaction environment and the metal nanoparticle size, inparticular if it changes during the reaction. This problem isstill a matter of debates in the H2O2 community. Only veryrecently the influence of reaction conditions on the catalyststate has attracted some attention. In a very recent study,Arrigo et al.26 reported that variations of the reaction condi-tions affect the performance and features of the catalyst.

In this study, we report on the performance of palladiumcatalysts supported on a strongly acidic resin22 (LewatitK2621) and used with addition of no promoter.

Catalyst characterizations were performed on the freshCatal. Sci. Technol. This journal is The Royal Society of Chemistry 2015

subsequent reduction step, a sensible fraction of the metalnanoparticles was formed within the polymer framework,efficiently limiting their growth. On the contrary, at low metalcontent, the PdII) ions are exchanged only by the sulfonicgroups on the pore walls. As a consequence, the polymerframework was less efficient in controlling the nanoparticlesize. All the spent catalysts show a particle size distributionbroader and shifted towards larger values, Fig. 2 and Table 1,

change from the least to the most dispersed catalyst (ESI,Fig. S1).

The XPS spectra for the fresh and spent 5.0-Pd/K2621catalyst are reported in Fig. 3 (other catalysts' spectra are similar).

XPS can sample Pd down to approx. 10 nm below themetal surface, as can be calculated by established correla-tion.30 Since the metal nanoparticles in our catalysts haddiameters in the range of 36.5 nm (Table 1), we can assumethat their XPS analysis is representative of most of their vol-ume. The spectra before and after the reaction (Fig. 4) com-

Table 1 Pd average nanoparticles size and oxide content of fresh and spent catalysts

Cat. Nanoparticle size Pd dispersiona (%) PdII)/Pd0)

Pd (%) Fresh (nm) Spent (nm) Difference (%) Fresh Spent Fresh Spent

0.3 4.7 / / 27.3 / / /0.5 5.3 5.5 +4 24.5 23.7 / 0.801.0 4.5 6.4 +42 28.4 20.6 2.67 1.432.5 3.6 5.2 +44 34.6 25.0 2.75 3.305.0 3.1 4.3 +39 39.3 29.6 4.36 5.80

a Estimated (see ESI, Section 1). /: data unavailable.

Catalysis Science & Technology Paper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article Onlinesuggesting some aggregation during the reaction. A similarinvestigation was reported by Ouyang et al.,29 with slightlydifferent results due to the different supports used. Probablythe different particle sizes obtained in the present work (withdifferent Pd loadings) are to be ascribed to the pores distri-bution inside the resin, its ability to be swollen in liquidphase and the sulfonic groups. In our work, as expected, the cat-alysts with the highest Pd content were most affected, showingapprox. 40% of nanoparticles size increase, whereas sinteringwas negligible in the 0.5 wt.% catalyst (the lowest Pd content).

The dispersion of the metal phase was estimated as thepercent ratio between surface and total Pd atoms from theaverage nanoparticle size, assuming that a) the metal nano-particles were spherical, b) Pd had the fcc structure typical ofthe bulk and c) Pd atomic radius is 137 pm (see ESI, Section 1,for further details). The estimated dispersion values(Table 1) were in the range 2540%, with a less than two-foldThis journal is The Royal Society of Chemistry 2015

Fig. 3 XPS spectra of the 5.0 Pd/K2621 catalyst: a) fresh; b) used.pare well, although a decrease in the intensity of the Pd 3dcore level was observed, possibly due to the effect of Pd nano-particle size on the binding energy (BE), as commentedbelow. Signals from carbon, oxygen, sulfur and nitrogen werefound in addition to the palladium one. Sulfur is present inthe sulfonic groups of the polymeric support catalysts andnitrogen could stem from the nitrate ions of the metal pre-cursor used in the ion-exchange step. In fact the position ofthe nitrogen peak (ca. 400 eV) suggests the presence in thesamples of ammonium ions, possibly obtained by reductionof NO3

during the reduction step.31

Fig. 4 shows the high-resolution XPS spectra of the Pd 3dcore level. The binding energy of Pd 3d5/2 photoelectron lineCatal. Sci. Technol.

Fig. 4 XPS signals of Pd 3d core level in the X-Pd/K2621 fresh (left)and spent (right) catalysts (X is the Pd wt.%, reported in each panel).

in metallic Pd and in PdII) fall at 335.1 and 336.3 eV, respec-tively.32 In the fresh catalyst two peaks for the Pd 3d5/2 sig-nals were observed at 336.0 eV and 337.8 eV, indicating thatPd is present in two different oxidation states; both bindingenergies are higher than reported in the literature. However,the positive shift of 1 eV in the BE has already been reportedby Penner et al.33 for small Pd(0) and PdII) nanoparticles (210 nm). Both the BE shift and the nanoparticle size observedfor our catalysts compare well with the values reported byPenner et al.,33 hence the two peaks can be attributed toPd(0) and PdII), respectively. A third peak observed around344.8 eV corresponds to energy loss features. For each sam-ple, the signals of the Pd(3d) core level were deconvolutedinto two curves, the integration of which allowed the estima-tion of the PdII)/Pd0) atomic ratio in the catalysts. Estimated

Activity in the H2O2 direct synthesis

product, with H2O2 selectivity values never exceeding 30%. Asexpected, the larger the [Pd] in the reactor, the higher thereaction rates. But the highest selectivity and final H2O2 con-centration (78 mM) values were achieved with the smallestconcentration of catalyst (hence of Pd) in the reaction mix-ture. In these cases, the selectivity initially increases,reaching a maximum after ca. 10 minutes and then decreasessteadily. With higher [Pd], the initial increase of selectivity istoo fast, appearing to decrease from the very beginning ofthe reaction. The drop in selectivity is larger with higher cata-lyst concentration. Further comments follow in the commondiscussion.

Changing the Pd loading in the catalyst

H2

HL2/(mo

534229179

1015394

Catalysis Science & TechnologyPaper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article OnlineA summary of the main results of the catalytic tests isreported in Table 2. The experiments with different concen-tration of the same catalyst in the slurry (1 to 5) will bepresented first, and separately from those where the sameamount of different catalysts was used (3, 6 to 9). A commondiscussion will follow.

Changing the catalyst concentration in the slurry

The evolution of H2O2 and H2O concentrations and of H2O2selectivity varying the amounts of the 1.0-Pd/K2621 (Table 2,runs 1 to 5) are reported in Fig. 5. H2O was always the main

Table 2 Summary of the experimental conditions and the main results at X

Catalyst

[Pd] (M)# ID mass (g) Pd (mol)

1 1.0-Pd/K2621 0.075 7 16.72 1.0-Pd/K2621 0.11 10 23.83 1.0-Pd/K2621 0.15 14 33.34 1.0-Pd/K2621 0.3 28 66.75 1.0-Pd/K2621 0.5 47 111.96 0.3-Pd/K2621 0.15 4 9.57 0.5-Pd/K2621 0.15 7 16.78 2.5-Pd/K2621 0.15 35 83.39 5.0-Pd/K2621 0.15 71 169.0values are reported in Table 1. In the fresh catalysts with Pdcontent 1.0 wt.% this ratio was higher than 1, showing thatPd was mostly present in the oxidized form.

The data of Table 1 show that in the fresh catalyst thehigher the Pd amount, the higher the dispersion and the oxi-dation degree. The oxidation degree still increases with thePd amount also in the spent catalyst, but the PdII)/Pd0) ratiochanges during the reaction and at least in one case (1.0-Pd/K2621) a partial reduction of the metal is observed. With thehighest Pd amounts in the catalysts (2.5-Pd/K2621 and 5.0-Pd/K2621) further oxidation of the metal is evident at the endof the reaction.Catal. Sci. Technol.The evolution of H2O2 and H2O concentrations and of H2O2selectivity varying the Pd loading within the same amount ofcatalyst in the slurry (Table 2, runs 3 and 6 to 9) are reportedin Fig. 6. Again, in these experiments the highest and lowestselectivity values were obtained with the lowest and thehighest Pd concentration in the reaction environment, respec-tively. Note that using the strategy of different Pd loadingsboth the H2O2 concentration and the selectivity were higherthan simply decreasing the catalyst amount (Fig. 5 and 6).

The selectivity of the 0.3-Pd/K2621 and 0.5-Pd/K2621 cata-lysts topped 80% after ca. 10 min and dropped to a finalvalue of approx. 40%, which is higher than the final value inany of the experiments of the first set. These two catalystsyield a final concentration of H2O2 up to 14.6 mM (withoutany promotion by acids and/or halides). Generally speaking,this is a quite low concentration, but it is due to the low over-all H2 amount introduced in the reactor. Moreover, the use ofthe same palladium concentration with different catalystsleads to much better results when the metal load in the sup-port is lower (Table 2, runs 1 and 7).

Comparison with other catalysts

A comparison with literature productivity data is not trivial,because reaction conditions might differ and H2 conversion(XH2) results not always are reported. Unfortunately, differentresearch groups normally operate at different experimental

= 0.6

Pdl mol1)

Selectivity(%)

H2O2 productivity103 (mol h1)

H2O productivity103 (mol h1)

0 24 2.33 7.203 24 4.71 14.819 23 12.76 41.823 16 10.22 54.802 12 16.17 12.841 45 5.89 7.190 57 12.39 9.309 22 16.22 58.837 10 16.63 148.01This journal is The Royal Society of Chemistry 2015

Catalysis Science & Technology Paper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article OnlineFig. 5 H2O2 selectivity and H2O2 and H2O concentration withdifferent concentration of 1.0 Pd/K2621 catalyst in the slurry: * 0.075This journal is The Royal Society of Chemistry 2015

conditions,1,13,34 so that an ideal comparison is almostimpossible. However, a comparison based on the specificH2O2 productivity, i.e. the ratio between the peroxide produc-tion rate and the Pd amount, can give interesting informa-tion. Note that the specific productivity is often referred asturnover frequency, TOF, although the ratio is only anapproximation of the microscopic TOF, and its definitionquestionable in a multisteps reaction. Literature data arereported in Table 3.

In our conditions the specific H2O2 productivity spans from234 (5.0-Pd/K2621) to 1770 (0.5-Pd/K2621) molH2O2/molPd h)at XH2 = 0.6. These results compare well with the highestones reported in the literature, if conveniently scaled withthe total pressure, that boosts the reaction rates. Notethat we achieved such performances without promoters(mineral acids and/or halide ions) in the reaction environ-ment and at fairly low pressure. The selectivity and theproductivity are comparable to those achieved using selectiv-ity enhancers, indicating the effectiveness of the catalystsynthesized.

Mass transfer limitations

The experiments were performed varying the amount of theactive metal, in two different ways. The reaction rates variedaccordingly, as clearly shown in Fig. 5 and 6, being quite fast

g; 0.110 g; 0.150 g; 0.300 g; 0.500 g.Fig. 6 H2O2 selectivity and H2O2 and H2O concentration with thesame amount of different catalysts: * 0.3 Pd/K2621; 0.5 Pd/K2621;Catal. Sci. Technol.

in some cases, so that concerns of mass transfer limitationsmay arise. Hence, H2 consumption rate was related to theactive metal amount: would the mass transfer be limiting,increasing the amount of catalyst should not bring anyimprovement in the conversion rate. However, a linear rela-tion was observed between H2 consumption rate and catalystamount (ESI, Section 2), clearly proving that the reactionsare occurring under the kinetic regime, as already demon-strated in comparable conditions.14 The H2 mass transferlimitation was also experimentally excluded in our previouswork,41 where the presence of H2 in the bulk of the liquidphase was experimentally observed in conditions comparableto those of this work. Would the reactor be limited by masstransfer phenomena, the presence of H2 in the bulk of theliquid phase should not have been observed. Moreover, in astirred slurry reactor, values of mass transfer coefficientabout 10 min1 have been measured.6 Hence, the H2 transferrate would be approximately 70 mmol min1 at our experi-mental conditions. The maximum H2 consumption rate thatwe measured was 3 mmol min1 (ESI, Fig. S3), once againconcluding that H2 mass transfer limitation can be excluded.

Disproportionation kinetics

The analysis of the H2 consumption rate helps to clarify therole of each reaction of Scheme 1 during the experiments.

1.0 Pd/K2621; 2.5 Pd/K2621; 5.0 Pd/K2621.

Table 3 Comparison between this work and literature results

T[C

2

24

6

Pd/SBA15 =7.7 : 15.4 : 61.5 :2

1

Catalysis Science & TechnologyPaper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article OnlineThe highest H2O2 concentration obtained (Fig. 5 and 6)always corresponds to the complete H2 consumption. Hence,

15.412 0.5% Pd0.5%

Au/TiO2Continuous CH3OH :

H2O =1.94

H2 : O2 : CO2 = 4 : 4 :92

40 2.12% Pd/ZrO2 Continuous CH3OH H2 :O2 : CO2 = 2 :18 : 80Ref. Catalyst ReactorLiquidphase

Gas composition[%]

Thiswork

0.3 5.0%Pd/K2621

Batch CH3OH H2 :O2 : CO2 = 4 :20 : 76

35 2.5% Pd2.5%Au/carbon

Batch CH3OH :H2O =1.93

H2 : O2 : CO2 = 3.6 :7.1 : 89.3

2.5% Pd2.5%Au/silica

36 1.8% Pd2.5%Au/HZSM-5

Batch CH3OH :H2O =1.93

H2 : O2 : CO2 = 3.6 :7.1 : 89.3

37 1.22%Pd0.92%Au/S-ZrO2

Semibatch CH3OH H2 :O2 = 4 : 96

1.3% Pd0.2%Pt/S-ZrO3

24 2.64% Pd/ZS Semibatch CH3OH H2 :O2 = 4 : 9638 1.6% Pd/SiO2 Semibatch CH3OH H2 :O2 :N2 = 3.6 :

46.4 : 50HSO3-functionalized)

3 5% Pd/C Semibatch H2O H2 : O2 : CO2 = 4 :16 : 80

39 4.4% Semibatch CH3OH H2 :O2 : CO2 : N2Catal. Sci. Technol.

according to the stoichiometry of the reactions, the subse-quent slow H2O2 depletion (if any) is due only to the dispro-portionation reaction. This is clearly much slower than allthe other reactions, as already demonstrated.14 A quantitativeestimation of the first-order disproportionation kinetic con-stants is given in the ESI, Section 3.

Activity and selectivity

In a batch reactor the production rate and the selectivity varyover time. This is a consequence of variations in the reactionenvironment during the reaction, i.e. the reagents and prod-uct concentrations (Scheme 1). In this work, the productionrates and the selectivity values were compared at a fixed H2conversion (60%, see ESI, Section 2 for motivations) ratherthan at a fixed reaction time. Though the latter convention isfrequently used, Fig. 5 and 6 clearly show how misleading itcould be, if one compares selectivities at a fixed time, e.g. 1h. Hence, catalysts have been compared at the same H2 con-version, high enough to ensure appreciable rates of the con-secutive undesired reactions, if any. The production rates of

H2O2 ( R602 2H O )and H2O ( R602

H O ) at 60% H2 conversion were cal-

culated as the slopes of the time profiles of H2O2 and H2Oconcentrations (Fig. 5 and 6) up to XH2 = 0.6 (achieved at verydifferent reaction times). Selectivity values were compared aswell. These are cumulative values, from the beginning up to60% H2 conversion, hence they can differ from the actualinstant values at that conversion. However, since the H2O2

]P[bar] Additives

ProductivitymolH2O2/molPd h)]

Selectivity[%]

H2 conv.[%]

2 19.5 / 1770 234 57 10 60

2 37 / 304 80 /

298 80 /

2 37 / 467 / /

0 1 H2SO4 (0.03 M) 90 62 64

124 65 64

0 1 H2SO4 (0.03 M) 58.5 40 99.50 95 HBr 1857 83 >90

0 80 H3PO4 (0.004 M),NaBr (0.03 M)

5632 74 29

0 6.5 H2SO4 (0.03 M) 282 40 /

2 10 / 26 23 18

0 10 / 4 45 15This journal is The Royal Society of Chemistry 2015

and especially H2O concentration profiles were quite linearup to XH2 = 0.6, cumulative and instant productivities andselectivities should match reasonably.

Values are reported in Table 2 and in Fig. 7 as a function

of the concentration of Pd in the reaction mixture. R602H O

increases linearly with the amount of palladium in both sets

of experiments, whereas R602 2H O does it only raising the total

amount of the same catalyst (Fig. 7). When the same catalystis used in different amounts, the metal nanoparticles are thesame, having always the same initial size and initial PdII)/P-0) ratio. This implies that the metal surface has always thesame features and the only changing parameter is the totalsurface area available, directly proportional to the total con-centration of palladium in the reaction environment. We can

conclude that in runs 1 to 5 (Table 2) both R602 2H O and R602

H O

were directly proportional to the metal surface area, asexpected for a reaction occurring under the kinetic regime.Although the rates of formation of both products vary linearly

with the palladium concentration, R602H O increases more rap-

idly, depressing the selectivity. Therefore it is clear that themore palladium is used the more water is relativelyproduced.

When the total concentration of palladium is changed byusing equal amounts of catalysts with different metal loads,

PdII)/Pd0) ratio change from one experiment to another and

Catalysis Science & Technology Paper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article Onlineonly R602H O linearly depends on the palladium concentration,

but R602 2H O does not. In particular when the metal load in the

catalyst is 0.5% or more, only a very slight increase of R602 2H O

is observed and the ratio between its highest and lowest

Fig. 7 Production rates of H2O2, R602 2H O (squares, above), and of water,

R602H O (triangles, below), with a) different amounts of 1.0-Pd/K2621

(open symbols) and b) with different X-Pd/K2621 catalyst (X = 0.3, 0.5,1.0. 2.5, 5.0) (solid symbols).This journal is The Royal Society of Chemistry 2015

values is less than 3. By contrast, there is a 20-fold increase

in R602H O from the catalyst with the lowest water productivity

to the most productive one. As the consequence, the decreaseof the selectivity towards H2O2 from the most to the leastselective catalyst is 5-fold as opposed to the 2-fold decreaseobserved changing the quantity of the same catalyst (first setof experiments). When using different metal loads, the totalmetal surface varies both with the catalyst palladium contentand the size of the nanoparticles. However, the surfaceamount of palladium is directly proportional to the totalamount of the metal even taking into account its differentdispersion in the different catalysts (ESI, Fig. S2). Hence thepalladium concentration can be taken as the only key param-eter and it can be again concluded that the more palladiumis used the lower is the selectivity.

The data above pinpoint that hydrogen peroxide produc-tion is structure sensitive while water production is not.Although final conclusions about the structure sensitivityrequire more detailed characterizations along the reactioncourse, being the measured rates apparent values arisingfrom different reactions, this findings confirms previousstudies related to the structure sensitivity.4251 In any case, itis clear that when different catalysts are used the nanoparti-cle size, the dispersion of the metal phase and the initialclusion that small metal nanoparticles can be more easily oxi-dized: during the reaction relatively small nanoparticlesundergo a net oxidation, but the relatively large ones ratherundergo a net reduction (when Pd nanoparticles change theirsize during the reaction) in spite of the abundance of oxygenin the reaction environment, especially in the late stage ofthe reaction. The relative role of PdII) and Pd(0) in the DS isa most debated one. Our results suggest that a catalyst highlyoxidized is detrimental for the selectivity towards H2O2. It istherefore useful to operate under conditions where palladiumreduction is favoured, as discussed later.

As a final remark, one of the most probable explanation ofthe simultaneous increase of the Pd nanoparticles size andthe increase in the PdII)/Pd0) molar ratio can be ascribed tothe dissolution and re-deposition of Pd from small nano-clusters to bigger ones. However, the sintering effect can notbe excluded, since it is a common behavior of nanoparticlesin heterogeneous catalysis.26 Both of the effects reportedabove (the Pd dissolution and re-deposition and the sinteringeffect) can result in an increase of the nanocluster size. TEMimages before and after reaction suggested that these nano-cluster changes can be explained by both the mechanismsreported above. The increase/decrease in PdII)/Pd0) molarratio (interconnected with Pd nanoparticles size variationduring the reaction) can also be explained by the Pd dissolu-tion and re-deposition on the nanocluster.52 The dissolutionof Pd can be enhanced by H2O2, as already proved.

53 AllThese facts strongly support the idea that Pd is dissolved inthe reaction medium during the H2O2 reaction and then re-deposited on the catalyst, raising the formation of biggercluster size.54 Moreover, sintering effect is detrimental forH2O2 direct synthesis,

26 and in the present work we are show-ing that the catalyst that is not increasing its nanocluster sizeis the one that is best performing (Table 1 and Fig. 6).

HL2/Pd ratio

The importance of controlling the oxidation state of palla-dium in the catalyst prompted us to investigate the ratiobetween the H2 in the liquid phase and the active metalamount. This parameter has already been proved to be rele-vant.9,14,41 In our experiments, both reducing (H2) and oxidiz-ing (O2) agents were present at the same time in the reactionenvironment and, of course, both can affect the Pd surface.However, the ratio between the concentration of dissolvedmust be taken into account. In particular, we observed thatthe higher the Pd load in the catalyst, the smaller the metalnanoparticle size (i.e. the higher the dispersion). Moreover, itseems that the catalysts with the smallest palladium nano-particles are more rapidly oxidized (column 7, Table 2). Thisis even more clearly observed in the spent catalysts, wherethe PdII)/Pd0) ratio increases upon decreasing the nanopar-ticle size. However, in the spent 1.0-Pd/K2621 the PdII)/Pd0)ratio is lower than in the fresh catalyst. 2.5-Pd/K2621 and 5.0-Pd/K2621 behave the opposite. This further supports the con-Catal. Sci. Technol.

reagents (H and O ) and available Pd varies. O was always

cluster size and metal oxidation state.6267 For these reasonsthe HL2/Pd initial value is an important parameter that affectsthe sorption phenomena.

Summarizing, it was demonstrated that the different ini-tial HL2/Pd molar ratio, along with the Pd nanoparticle size,can determine different oxidation state of the catalyst, even-tually controlling the relative rates of H2O and H2O2 produc-tion, i.e. the instantaneous selectivity of the catalyst.

Conclusions

The results reported in this work show that in the direct syn-thesis of hydrogen peroxide the HL2/Pd ratio is a key factors incontrolling the rates of H2O and H2O2 production over mono-metallic Pd catalysts supported on the commercial ion-exchange resin K2621 in the absence of promoters (halidesand acids). This work successfully demonstrated that a lowPdII)/Pd0) molar ratio (i.e. a more reduced catalyst) has apositive effect on the production of H2O2. A high H

L2/Pd ratio,

helping to reduce the catalyst, and a relatively low dispersionof the metal phase, which makes it more resistant to oxida-tion, are fundamental conditions to obtain an active catalystfor the direct synthesis. In fact, the 0.5-Pd/K2621, which is theleast oxidized catalyst and has the largest initial nanoparticles,was the most selective when operating at HL2/Pd = 530 (the

Fig. 8 Selectivity at 60% H2 conversion (a: different amounts of 1.0Pd/K2621; , different X-Pd/K2621 catalyst; X = 0.3, 0.5, 1.0. 2.5, 5.0)and PdII)/Pd0) molar ratio of spent catalysts (b) as a function of theHL2/Pd molar ratio.

Catalysis Science & TechnologyPaper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article Online2 2 2

present in large excess, so that O2/H2 molar ratio55 in the liq-

uid phase was always higher than 15. In spite of this, no cata-lyst was ever fully oxidized at the end of the reaction andeven some reduction was observed. This suggests that the keyfactor controlling the selectivity is rather the molar ratiobetween the H2 dissolved in the liquid phase (H

L2) and the

quantity of Pd. We correlate the total amount of palladiumwith the H2 solubilised in the liquid phase and not only thenanocluster surface because of the experimental evidencethat the H2O2 selectivity may be impaired by the formation ofpalladium -hydrides.56 The -hydrides can be formed alsobeneath the metal surface so that the composition of theinterior of the nanoparticles can be relevant to the selectivity,too. Moreover, in this study the amount of the surface palla-dium is directly proportional to its total amount (ESI, Fig.S4). For these reasons the value of HL2/Pd is a proper parame-ter to compare the catalyst activity. In a batch reactor thisratio steadily decreases as the reaction proceeds, so that itsinitial value will be used hereafter. The initial H2 concentra-tion in the liquid phase was calculated as the H2 concentra-tion in equilibrium with the initial gas phase composition.55

We have already reported that the initial HL2/Pd molar ratioplays an important role in the H2O2 direct synthesis.

13,41

Here we have found that in both sets of catalytic runs, higherHL2/Pd values favour the production of H2O2 (Fig. 8a). This isin agreement with our previous findings,9 where again ahigher selectivity was achieved by decreasing the catalystamount at constant H2 concentration. Not surprisingly, theinitial HL2/Pd ratio is correlated also with the PdII)/Pd0) ratioin the spent catalysts (Fig. 8b). In particular, when the initialvalue of HL2/Pd is relatively high the final PdII)/Pd0) ratio isrelatively low. The extent of metal oxidation during the reac-tion decreases also with the nanoparticle size (see above) andboth factors certainly play a role. The final PdII)/Pd0) ratiois not simply a matter of the initial HL2/Pd ratio; this issuggested by the value of PdII)/Pd0) ratio in spent 1.0-Pd/K2621, which is lower than in the same fresh catalyst,although at the end of the reaction there is no more hydro-gen. The higher resistance to oxidation of relatively largemetal nanoparticles is also expected, because a relatively lowdispersion implies the prevalence of low-index faces on thePd surface, that are expected to be less reactive.33

Although we cannot clearly separate the effect of theseparameters (i.e. oxidation state, nanoparticle size, HL2/Pd ini-tial value), we conclude that the combination of a relativelyhigh HL2/Pd initial value (i.e. low catalyst amount) with rela-tively large metal nanoparticles is recommended to improvethe selectivity towards H2O2. These conditions contribute to alow oxidation degree of the active metal, which appears bene-ficial to the selectivity. The positive effect of Pd reductionwas also reported by Biasi et al.,41 Burato et al.,23 Meladaet al.24 and Liu et al.25 As already reported in theliterature,5761 there exists a competition in the sorption phe-nomena between oxygen and hydrogen on the catalyst sur-face. The sorption extent and rate depend on the metalCatal. Sci. Technol. This journal is The Royal Society of Chemistry 2015

second highest value in this investigation), with 57% selectiv-

introduced into the microscope.

Catalytic tests

The catalysts were tested in a slurry, within an autoclaveoperated batchwise, following a common catalyst testingpractice.19,39,6971 In a batch reactor the gas and the liquidcomposition and possibly the catalyst state change over time.This is the same as the variations along the flow direction in

Table 4 Analytical data on Pd catalysts with different metal loadings

Catalysis Science & Technology Paper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article Onlineity at 60% conversion of H2 (up to 80% at lower conversion)and an H2O2 productivity of 12.4 mol h

1 (corresponding to aspecific peroxide productivity of 1313 molH2O2molPd

1 h1).These are interesting results using a monometallic Pd cata-lyst in the absence of promoters. Further investigations arerequired to explain why less oxidized Pd is more selectiveand how its oxidation can be even more effectively limited, inorder to achieve high selectivity and H2O2 concentration alsoat high H2 conversions. Most remarkably, structure sensitivityin H2O2 production and structure insensitivity in the H2Oproduction were noticed in the direct synthesis reaction net-work. This discovery will open new insights in the reactionmechanism of the direct synthesis.

Experimental section

Materials

Lewatit K2621 (sulfonated polystyrenedivinylbenzene macro-reticular ion-exchange resin; exchange capacity = 1.92mmol g1) was provided by Lanxess. Pd(NO3)2 (Alfa Aesar)was used as received. Tetrahydrofuran (THF, Sigma-Aldrich)used freshly distilled. HPLC grade (99.99%) methanol forHPLC is from J.T. Baker. H2 and O2 as reagents and CO2 asan inert gas (AGA-Linde 5.0 purity) were used in the directsynthesis experiments.

Catalyst preparation

Pd supported on K2621 materials were prepared by ionexchange method. The reduction of the precursors to themetals takes place inside the polymer framework, which isable to control the dispersion, the size and, therefore, the cat-alytic properties of the metal nanoparticles. Lewatit K2621 isa macroreticular, sulfonated polystyrenedivinylbenzene (S-PSDVB) resin. It possesses permanent meso- and macroporesboth in the dry and in the swollen states. The permanentmeso- and macropores of macroreticular resins make the dif-fusion of reagents and products less dependent on the swell-ing and generally faster. Thus, Lewatit K2621, with its macro-porous structure and its ability to swell in polar solventsshould be suited to overcome internal mass transfer limita-tions. Moreover, its acidic nature (exchange capacity is1.92 mmol g1) appears to provide the same functionality ofadded acids, avoiding the use of a corrosive reaction environ-ment.22,23 Overall, S-PSDVB resins have several featuresmakingthem attractive supports for the direct synthesis catalysts.68

The resin K2621 was washed with water (300 ml for 10 gof material) and rinsed with methanol (100 ml for 10 g mate-rial). 2.0 g of K2621 were suspended in 10 ml of distilledwater and left standing for 2 hours. An aqueous solution ofPdNO3)2 was added to the suspension. The amount ofPdNO3)2 was varied to obtain 0.3, 0.5, 1, 2.5 and 5 wt.%nominal Pd loadings, respectively (Table 4). After adding themetal solution, the suspension reacted overnight underThis journal is The Royal Society of Chemistry 2015mechanical stirring (swirling plate). The material was filteredand washed with deionized water (5 10 ml). The motherliquors were analyzed for the unreacted metal by means ofinductively coupled plasma mass spectrometry (ICP-MS). Theresidual amount of metal after the ion-exchange was alwaysless than 0.1 mol% of the initial amount, confirming that theactual metal loadings were essentially equal to the nominalvalues. The as synthesized materials were suspended in THFand reduced under H2 flux, at room temperature, for 5 hours.After recovery by vacuum filtration, the materials werewashed on filter paper with THF (3 10 ml) and dried at110 C overnight.

Characterization method

All XPS spectra were recorded with Kratos Axis Ultra electronspectrometer equipped with a delay line detector. A mono-chromated Al K source operated at 150 W, hybrid lens sys-tem with magnetic lens, proving an analysis area of 0.3 0.7mm2, and charge neutralizer were used for the measure-ments. The binding energy (BE) scale was referenced to theC1s line of aliphatic carbon, set at 285.0 eV. Processing of thespectra was accomplished with the Kratos software. An Al KX-rays was used as excitation source with photon energy of1486.6 eV. Depth of analysis for our samples is 10 nm and isnot dependent on excitation source in conventional XPS.

Active metal content in the reaction medium was accessedby a PerkinElmer Sciex ICP Mass Spectrometer Elan 6100DRC Plus. The timing parameters were: sweeps/reading: 11,readings/replicate: 1, number of replicates: 7, dwell time:50.0 ms, scan mode: peak hopping. The calibration solutionswere prepared from commercial single element solution forPd, diluted into standard serial from 1 ppb to 100 ppb.

TEM analyses were carried out with an energy filteredtransmission electron microscopy (EFTEM, LEO 912 OMEGA,LaB6 filament, 120 kV). Samples were prepared bysuspending a few milligrams of the powder in high purityisopropyl alcohol (or ethanol). After ultrasonic bath (30 s) asmall droplet (5 l) of the suspension was transferred onto aholey-carbon film coated Cu grids, which was eventually

CatalystK2621/H+

(g)PdNO3)2(g)

Measured Pd loading(wt.%)

0.3-Pd/K2621 10.6705 0.0875 0.300.5-Pd/K2621 1.1702 0.0174 0.521.0-Pd/K2621 5.2103 0.1375 1.032.5-Pd/K2621 1.0256 0.0668 2.565.0-Pd/K2621 1.0189 0.1314 5.07Catal. Sci. Technol.

continuous flow reactors.12 Reactors where the gas is contin- 3 T. M. Rueda, J. G. Serna and M. J. C. Alonso, J. Supercrit.

Catalysis Science & TechnologyPaper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article Onlineconstant H2 and O2 concentration in the slurry, but still theproducts (which can be reagents for side reactions, such aH2O2) change concentration over time and the catalyst may varyits state. Thus, the batch reactor leads to a more complicated,but feasible kinetic analysis14 while allowing for a rapid com-parison of catalysts, even if differences in the reaction environ-ment at the same elapsed time may occur. These are expectedto be fairly small when comparing similar catalysts.

The catalysts were analyzed before and after the reaction,to correlate the reaction conditions to the possible variationof the catalyst properties. Hydrogen peroxide direct synthesiswas performed in a 600 ml autoclave, described elsewhere.41

Briefly, after the introduction of the catalyst, CO2 and O2were fed to the reactor (partial pressures at 23 C were 20.7and 7.2 bar, respectively), followed by 420 ml of methanol. Atthis stage the internal pressure was 28 bars. The temperaturedecreased to 2 C and stirring (1000 rpm) started and keptfor 30 min, allowing the gasliquid equilibrium to be reached(lowering the pressure down to 19.5 bar). Then H2 (0.0183moles) was introduced in the reactor as the limiting reagentfrom a reservoir of known volume and the reaction wasstarted. The measured pressure drop in the H2 reservoirallowed for a precise and reproducible determination of theamount of H2 introduced in the reactor. Several liquid sampleswere withdrawn through a GC-valve during each experiment.H2O2 and H2O concentrations were determined by iodometricand Karl-Fischer titrations, respectively. TheH2O2 selectivity wascalculated as follows: H2O2 selectivity = (moles H2O2 produced)/(moles H2O + H2O2 produced) 100. Experiments wereperformed varying the total amount of Pd in the reaction systemin two different ways. A first set of experiments was carriedout using different amounts (0.0750.500 g) of a single catalyst(1.0-Pd/K2621) in the same liquid volume, and a second setwas carried out using always the same fixed amount (0.150 g)of different catalysts (Pd content 0.35.0 wt.%). All tests last 3 h.

Acknowledgements

Lanxess is kindly acknowledged for providing the K2621resin. NG gratefully acknowledges the Cariparo Foundation inItaly and the Technology Development center of Finland(TEKES) for financial support. NG and SS gratefully acknowl-edge the Johan Gadolin Scholarship. PB gratefully acknowl-edges the Otto A. Malm Foundation and the Kempe Founda-tions for financial support. Anne-Riikka Rautio is greatfullyacknowledged for some of the TEM analyses. This work ispart of the activities at the bo Akademi Process ChemistryCentre (PCC). Also, the Bio4Energy program is acknowledged.

References

1 C. Samanta, Appl. Catal., A, 2008, 350, 133149.2 J. K. Edwards, S. J. Freakley, R. J. Lewis, J. C. Pritchard and

G. J. Hutchings, Catal. Today, 2015, 248, 39.uously bubbled through a well-mixed slurry provide approx.Catal. Sci. Technol.28 B. Corain, M. Zecca, P. Canton and P. Centomo, Synthesisand catalytic activity of metal nanoclusters inside functionalresins: an endeavour lasting 15 years, 2010.2010, 333, 109113.and R. Schlgl, ChemSusChem, 2014, 7, 179194.27 K. Jebek, L. Hankov and L. Holub, J. Mol. Catal. A: Chem.,Netherlands, 2009.26 R. Arrigo, M. E. Schuster, S. Abate, S. Wrabetz, K. Amakawa,

D. Teschner, M. Freni, G. Centi, S. Perathoner, M. Hvecker25 Q. Liu, K. Gath, J. Bauer, R. Schaak and J. Lunsford, Theactive phase in the direct synthesis of H2O2 from H2 and O2over Pd/SiO2 catalyst in a H2SO4/ethanol system, SpringerJ. Catal., 2006, 239, 422430.and B. Corain, Adv. Synth. Catal., 2006, 348, 255259.24 S. Melada, R. Rioda, F. Menegazzo, F. Pinna and G. Strukul,22 G. Blanco-Brieva, E. Cano-Serrano, J. Campos-Martin andJ. L. G. Fierro, Chem. Commun., 2004, 11841185.

23 C. Burato, P. Centomo, M. Rizzoli, A. Biffis, S. Campestrini2002, 83, 235239.Lett., 2002, 84, 8187.21 V. Choudhary, A. Gaikwad and S. Sansare, Catal. Lett.,A. F. Carley, C. J. Kiely and G. J. Hutchings, J. Catal.,2012, 292, 227238.

20 V. R. Choudhary, S. D. Sansare and A. G. Gaikwad, Catal.19 J. K. Edwards, J. Pritchard, M. Piccinini, G. Shaw, Q. He,Res., 2008, 47, 81198125.18 T. Deguchi and M. Iwamoto, J. Catal., 2011, 280, 239246.17 Y. Voloshin, J. Manganaro and A. Lawal, Ind. Eng. Chem.Fluids, 2013, 74, 8088.16 Y. Voloshin and A. Lawal, Appl. Catal., A, 2009, 353, 916.15 I. Huerta, J. Garca-Serna and M. J. Cocero, J. Supercrit.and T. O. Salmi, Green Chem., 2014, 16, 2320.14 N. Gemo, P. Biasi, P. Canu and T. O. Salmi, Chem. Eng. J.,

2012, 207208, 539551.Moulijn and G. J. Hutchings, ACS Catal., 2013, 3, 487501.13 J. Garcia-Serna, T. Moreno, P. Biasi, M. J. Cocero, J. Mikkola11 P. Biasi, F. Menegazzo, F. Pinna, K. Ernen, P. Canu andT. O. Salmi, Ind. Eng. Chem. Res., 2010, 49, 1062710632.

12 S. J. Freakley, M. Piccinini, J. K. Edwards, E. N. Ntainjua, J. A.Hutchings, Acc. Chem. Res., 2014, 47, 845854.Chem., 2013, 15, 25022513.10 J. K. Edwards, S. J. Freakley, A. F. Carley, C. J. Kiely and G. J.8 S. Abate, M. Freni, R. Arrigo, M. E. Schuster, S. Perathonerand G. Centi, ChemCatChem, 2013, 5, 18991905.

9 P. Biasi, J. Garcia-Serna, A. Bittante and T. Salmi, Green7 T. Inoue, Y. Kikutani, S. Hamakawa, K. Mawatari, F.Mizukami and T. Kitamori, Chem. Eng. J., 2010, 160,909914.P. Canu, Chem. Eng. J., 2011, 176177, 172177.6 T. Deguchi and M. Iwamoto, Ind. Eng. Chem. Res., 2011, 50,

43514358.4 P. Biasi, S. Zancanella, F. Pinna, P. Canu and T. O. Salmi,Chem. Eng. Trans., 2011, 24, 4954.

5 P. Biasi, F. Menegazzo, F. Pinna, K. Ernen, T. O. Salmi andFluids, 2012, 61, 119125.This journal is The Royal Society of Chemistry 2015

29 L. Ouyang, P. Tian, G. Da, X. Xu, C. Ao, T. Chen, R. Si, J. Xuand Y. Han, J. Catal., 2015, 321, 7080.

30 B. D. Ratner and D. G. Castner, Electron Spectroscopy forChemical Analysis, in Surface analysis the principaltechniques, ed. J. C. Vickerman and I. S. Gilmore, John Wiley& Sons, Ltd, 2009, pp. 47112.

31 C. R. Lederhos, J. M. Badano, N. Carrara, F. Coloma-Pascual,M. C. Almansa, D. Liprandi and M. Quiroga, Metal andprecursor effect during 1-heptyne selective hydrogenation usingan activated carbon as support, 2013.

32 J. F. Moulder, J. Chastain, R. C. King, in Handbook of x-rayphotoelectron spectroscopy: a reference book of standard spec-tra for identification and interpretation of XPS data, Physical

49 D. W. Blakely and G. A. Somorjai, J. Catal., 1976, 42,181196.

50 G. A. Somorjai, Acc. Chem. Res., 1976, 9, 248256.51 J. T. Yates, J. Vac. Sci. Technol., A, 1995, 13, 13591367.52 W. M. Lemke, R. B. Kaner and P. L. Diaconescu, Inorg.

Chem. Front., 2015, 2, 3541.53 Sariolan, Platinum Met. Rev., 2013, 57, 289.54 F. Zhao, M. Shirai, Y. Ikushima and M. Arai, J. Mol. Catal. A:

Chem., 2002, 180, 211219.55 N. Gemo, P. Biasi, T. O. Salmi and P. Canu, J. Chem.

Thermodyn., 2012, 54, 19.56 F. Menegazzo, M. Manzoli, M. Signoretto, F. Pinna and G.

Catalysis Science & Technology Paper

Publ

ished

on

05 M

ay 2

015.

Dow

nloa

ded

by F

AC

DE

QUIM

ICA

on 13

/05/20

15 21

:01:32

. View Article Online95105.48 B. Lang, R. W. Joyner and G. A. Somorjai, Surf. Sci., 1972, 30,

454474.47 M. Boudart and F. Rumpf, React. Kinet. Catal. Lett., 1987, 35,46 M. Boudart, A. Aldag, J. E. Benson, N. A. Dougharty and C.Girvin Harkins, J. Catal., 1966, 6, 9299.44 R. A. Van Santen, Acc. Chem. Res., 2009, 42, 5766.45 H. S. Taylor, Proceedings of the Royal Society of London A:

Mathematical, Philos. Trans. R. Soc., A, 1925, 108, 105111.43 M. Che and C. O. Bennett, Adv. Catal., 1989, 36, 55172.89038912.42 G. C. Bond, Chem. Soc. Rev., 1991, 20, 441475.Ind. Eng. Chem. Res., 2012, 51, 88838890.41 P. Biasi, N. Gemo, J. R. Hernndez Carucci, K. Ernen, P.

Canu and T. O. Salmi, Ind. Eng. Chem. Res., 2012, 51,40 P. Biasi, P. Canu, F. Menegazzo, F. Pinna and T. O. Salmi,39 S. Abate, P. Lanzafame, S. Perathoner and G. Centi, Catal.Today, 2011, 169, 167174.Cruciani and G. Strukul, Appl. Catal., A, 2009, 358, 129135.38 G. Blanco-Brieva, F. E. De, J. Campos-Martin and J. L. G.

Fierro, Green Chem., 2010, 12, 11631166.37 G. Bernardotto, F. Menegazzo, F. Pinna, M. Signoretto, G.36 G. Li, J. Edwards, A. F. Carley and G. J. Hutchings, Catal.Today, 2007, 122, 361364.Angew. Chem., Int. Ed., 2006, 45, 69626984.35 J. K. Edwards, A. Thomas, A. F. Carley, A. A. Herzing, C. J.

Kiely and G. J. Hutchings, Green Chem., 2008, 10, 388394.34 J. M. Campos-Martin, G. Blanco-Brieva and J. L. G. Fierro,33 S. Penner, P. Bera, S. Pedersen, L. T. Ngo, J. J. W. Harris andC. T. Campbell, J. Phys. Chem. B, 2006, 110, 2457724584.Electronics, Eden Prairie, Minn., 1995.This journal is The Royal Society of Chemistry 2015Canu and B. Corain, Appl. Catal., A, 2009, 358, 224231.122130.71 C. Burato, S. Campestrini, Y. Han, P. Canton, P. Centomo, P.Cruciani, F. Pinna and G. Strukul, J. Catal., 2009, 268,2006, 18, 26892695.70 F. Menegazzo, M. Signoretto, M. Manzoli, F. Boccuzzi, G.T. Salmi and M. Zecca, Appl. Catal., A, 2013, 468, 160174.69 B. E. Solsona, J. K. Edwards, P. Landon, A. F. Carley, A.

Herzing, C. J. Kiely and G. J. Hutchings, Chem. Mater.,67 C. Ntzenadel, A. Zttel, D. Chartouni, G. Schmid and L.Schlapbach, Eur. Phys. J. D, 2000, 8, 245250.

68 S. Sterchele, P. Biasi, P. Centomo, P. Canton, S. Campestrini,2006, 110, 2178321787.65 M. Shao, J. Power Sources, 2011, 196, 24332444.66 S. Huang, C. Huang, B. Chang and C. Yeh, J. Phys. Chem. B,4854.64 S. Kim, D. Lee and K. Lee, J. Mol. Catal. A: Chem., 2014, 383

384, 6469.62 M. Shao, T. Yu, J. H. Odell, M. Jin and Y. Xia, Chem.Commun., 2011, 47, 65666568.

63 S. Kim, D. Lee and K. Lee, J. Mol. Catal. A: Chem., 2014, 391,Hydrogen Energy, 2013, 38, 952965.61 P. Dibandjo, C. Zlotea, R. Gadiou, C. Matei Ghimbeu, F.Cuevas, M. Latroche, E. Leroy and C. Vix-Guterl, Int. J.26282634.60 A. Baldi, T. C. Narayan, A. L. Koh and J. A. Dionne, Nat.

Mater., 2014, 13, 11431148.1983, 45, 431434.58 S. J. Teichner, Appl. Catal., 1990, 62, 110.59 J. F. Lynch and T. B. Flanagan, J. Phys. Chem., 1973, 77,Strukul, Catal. Today, 2015, 248, 1827.57 D. Sderberg and I. Lundstrm, Solid State Commun.,Catal. Sci. Technol.