Influence of Metal Precursors and Reduction Protocols on the Chloride-Free Preparation of Catalysts for the Direct Synthesis of Hydrogen Peroxide without Selectivity Enhancers

Article abstract of DOI:10.1002/cctc.201600021

Different metal precursors and reducing agents were applied in the preparation of 1wt % Pd catalysts supported on commercial ion-exchange resin (Lewatit K2621) and used in the direct synthesis of H₂O₂. The catalysts were characterize

Full text of DOI:10.1002/cctc.201600021

DOI: 10.1002/cctc.201600021
Full Papers
Influence of Metal Precursors and Reduction Protocols on
the Chloride-Free Preparation of Catalysts for the Direct
Synthesis of Hydrogen Peroxide without Selectivity
Enhancers
[
a, b]
[b, c]
[a]
[d]
Stefano Sterchele,
Pierdomenico Biasi,
Paolo Centomo, Andrey Shchukarev,
[
c, d]
[d]
[b, c]
[b]
Krisztiµn Kordµs,
Anne-Riikka Rautio, Jyri-Pekka Mikkola,
Tapio Salmi,
[e]
[a]
Patrizia Canton, and Marco Zecca*
th
th
In memory of Professor Benedetto Corain (July 8 1941–September 24 2014), a catalyst for our research activities and ideas.
Different metal precursors and reducing agents were applied
in the preparation of 1 wt% Pd catalysts supported on com-
mercial ion-exchange resin (Lewatit K2621) and used in the
direct synthesis of H O . The catalysts were characterized by
both the choice of the Pd precursor and the reduction condi-
tions had a strong influence on the size and size distribution
of the resulting supported nanostructured metal nanoparticles
and, consequently, on the catalytic performance. The best
combination of metal precursor and reduction agent was
2
2
using TEM and their performance was evaluated in the direct
synthesis of H O (in a batch and semi-batch reactor) to investi-
[Pd(NH ) ]SO4 reduced with hydrogen. This catalyst had the
2
2
3 4
gate the relationship between the catalyst preparation meth-
ods, morphology, and catalytic performance. As expected,
largest average size of the Pd nanoparticles and the broadest
size distribution.
Introduction
Hydrogen peroxide is a versatile and green commodity chemi-
cal used, for example, as a bleaching agent in the pulp and
demand is increasing, and for some applications a process al-
ternative to the auto oxidation process is desirable. It has
a number of shortcomings: the capital expenditure (CAPEX) is
high, it is affordable only for large-scale operation, and leads
to laborious wastewater treatment. Consequently, the catalytic
[1]
paper industry and in wastewater treatment. Its worldwide
[
a] Dr. S. Sterchele, Dr. P. Centomo, Prof. M. Zecca
Dipartimento di Scienze Chimiche
Università degli Studi di Padova
direct synthesis of H O2 (CDS) could be a viable route for
2
small-scale, on-site, and on-demand production of dilute solu-
[
1]
via Marzolo 8, I35131 Padova (Italy)
E-mail: marco.zecca@unipd.it
tions, which are often required.
The fundamental aspects of the CDS process have been re-
[
b] Dr. S. Sterchele, Dr. P. Biasi, Prof. J.-P. Mikkola, Prof. T. Salmi
Department of Chemical Engineering
[1–5]
viewed recently,
but many issues that concern the reaction
mechanism still remain unclear because of the complexity of
Laboratory of Industrial Chemistry and Reaction Engineering
Johan Gadolin Process Chemistry Centre
[
6–14]
the three-phase system
and the reaction network
Š
bo Akademi University
(Scheme 1). Metallic Pd is the most active catalyst for this reac-
Biskopsgatan 8, FI-20500 Šbo-Turku (Finland)
[1–3,12,15]
tion
and is applied generally in a supported form. In
[
c] Dr. P. Biasi, Dr. K. Kordµs, Prof. J.-P. Mikkola
Department of Chemistry
Chemical-Biochemical Centre (KBC), Technical Chemistry
Umeå University
most studies, the catalysts utilized in the CDS were prepared
typically by the impregnation of inorganic supports or carbon.
In general, the impregnated solids are typically subjected to
a subsequent reduction/calcination step at medium to high
temperature (473–773 K) often under a hydrogen flow. Al-
though this is a simple way to produce supported noble-metal
catalysts, both in academia and industry, catalysts for the CDS
SE-90187 Umeå (Sweden)
[
d] Dr. A. Shchukarev, Dr. K. Kordµs, Dr. A.-R. Rautio
Faculty of Technology, Microelectronics and Materials Physics Laboratories
EMPART Research Group of Infotech Oulu
University of Oulu
FI-90014 Oulu (Finland)
[
e] Dr. P. Canton
Department of Molecular Sciences and Nanosystems
Università Ca’ Foscari di Venezia
via Torino 155/b, 30170 Venezia-Mestre (Italy)
Supporting Information and the ORCID identification number(s) for the
author(s) of this article can be found under http://dx.doi.org/10.1002/
cctc.201600021.
2 2
Scheme 1. Reaction network of H O direct synthesis (side-reactions shaded).
ChemCatChem 2016, 8, 1564 – 1574
1564
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
often give rise to a moderate activity or modest selectivity to-
ty of the catalysts. To avoid the unintentional presence of
these promoters that could hinder the real catalytic behavior
of the pristine catalyst, we used halide-free conditions in both
the preparation of the catalysts and the catalytic runs. K2621
ion-exchanged with [Pd(NH ) ]SO was employed to screen the
[
2]
wards H O .
2
2
If crosslinked ion-exchange resins such as sulfonated poly-
styrene/divinylbenzene (S-PS/DVB) are applied as the catalyst
support, a different approach is possible. The ion-exchange
ability and hydrophilicity of S-PS/DVB resins such as Lewatit
K2621 make the introduction of cationic metal precursors into
the support straightforward. The procedure entails a simple
3
4
4
different reducing agents, which included hydrogen flowing in
a suspension of the precatalyst (either at room temperature
and atmospheric pressure or 333 K and 0.5 MPa); an aqueous
[
15–22]
[21]
ion-exchange step,
in which
a
À
precursor such as
solution of formaldehyde (37 wt%); a solution of methanol
2+
À
[15]
[
Pd(NH ) ] in water or [Pd(OAc) ] (AcO =CH COO ) in a (pref-
in water (50% v/v); and an aqueous solution of sodium bor-
3
4
2
3
+
[30]
erably polar) organic solvent replaces the counterions (H or
ohydride.
The different conditions applied to prepare the
+
Na ) of the sulfonic groups of the resin. A remarkable advant-
two sets of catalysts (different precursor and same reducing
agent; same precursor and different reducing agents) were ex-
pected to lead to different morphologies of the supported
nanostructured phase, which was investigated by using TEM.
After the screening of the catalysts under batch conditions,
a few selected nanocomposites were tested in semi-batch ex-
periments at 283 K and 50 bar to mimic conditions of CDS fea-
sible for industrial operation.
age of S-PS/DVB as an ion-exchanger is that the concentration
and the distribution of the sulfonic groups inside the polymer
[
19,20]
can be controlled,
and they are optimized for continuous
operation. To expand the scope of this method, the ion-ex-
change step can be replaced by a ligand-exchange reaction, in
which functional groups that are good ligands for the metal to
[23,24]
be introduced enter its coordination sphere.
However, the
commercial availability, the relatively low cost of S-PS/DVB
[
25]
resins, their good properties, and their full compatibility with
the working conditions for CDS make them particularly attrac-
tive as heterogeneous catalytic materials. Moreover, a number
of industrial processes based on these materials as either acid
Results and Discussion
The fresh metal-polymer nanocomposites were characterized
by using TEM and elemental analysis (Table 1). A commercial
1 wt% Pd/C catalyst was applied as the benchmark catalyst
and was also characterized by TEM (Table 1). The size distribu-
tion of the Pd nanoparticles in the fresh catalysts were as-
sessed by TEM, for which at least 250 particles were counted
for each sample.
[
23,24,26,27]
[26,27]
catalysts
or catalytic supports
have already been
implemented (e.g., methylisobutylketone and methyl-tert-buty-
lether processes, Bisphenol A synthesis, direct synthesis of alco-
hol from olefins and syn-gas, and the hydrogenation of differ-
ent substrates).
Although the open literature on CDS is pretty extensive, the
problem of the role of the different metal precursors and re-
duction protocols on the catalytic performance has not yet
This showed that, for all the samples, most of the individual
aggregates had a spherical shape and were well dispersed in
the polymeric support. They generally have a lognormal size
[
28]
[31]
been addressed systematically. Only recently, Sankar et al.
distribution, with moderate statistical dispersions (Figure 1
demonstrated that an excess of chloride ions during the im-
pregnation stage produced a stable bimetallic Au/Pd catalyst
with a smaller size distribution and better productivity in CDS
than that prepared using the more traditional and established
routes. The performance of the materials depended strongly
on the type of Pd and Au precursors used and, in particular,
the use of a tetrahalide complex seemed to be the key factor
that influenced the size of the nanoparticles marked-
and Table 1). The Pd aggregates in the samples 1PdK/N/H and
1PdK/A/H (Figure 2A–B and C–D, respectively) feature broad
distributions centered at 4.4 and 6.7 nm, respectively. 1PdK/S/
H has a broader size distribution with an average diameter of
12.2 nm (Figure 1). The commercial 1 wt% Pd/C used as the
benchmark showed the narrowest size distribution centered at
3.8 nm.
ly and, consequently, the catalytic behavior of the re-
sulting material.
Table 1. Analytical data for the 1 wt% Pd catalysts supported on K2621.
In this study, we examined the dependence of the
catalytic performance on the synthesis conditions
and, in particular, on the nature of the metal precur-
sor applied in the ion-exchange step as well as the
reducing agent applied in the subsequent reduction
step. We screened three different metal precursors
and five different reduction protocols. The three
metal precursors, [Pd(NH ) ]SO , Pd(NO3)2, and
Sample
Reductant Metal
precursor
Pd found
[wt%]
Fresh
d [nm]
Spent
d [nm]
[c]
[d]
[c]
[d]
SC
SC
[
a]
1
1
PdK/S/H
PdK/N/H
H
H
H2
2
[Pd(NH
Pd(NO
Pd(OAc)2
3
)
4
]SO
4
1.01
1.03
1.03
1.00
1.00
1.02
1.02
–
12.2
4.4
6.7
8.1
5.5
1.06 10.2
1.06
0.81
n.d.
n.d.
n.d.
n.d.
n.d.
1.21
[a]
a]
2
3
)
2
1.46
1.40
1.44
1.61
0.88
0.91
0.83
4.18
n.d.
[
1PdK/A/H
1PdK/S/F
[e]
HCHO
MeOH
[Pd(NH
[Pd(NH
[Pd(NH
[Pd(NH
–
3
3
3
3
)
)
)
)
4
4
4
4
]SO
]SO
]SO
]SO
4
4
4
4
n.d.
n.d.
n.d.
n.d.
4.03
1
1
1
PdK/S/M
PdK/S/B
PdK/S/HPT
NaBH
4
3.83
4.80
3.68
3
4
4
[f]
[b]
H
–
2
[
Pd(AcO) ], were introduced into the K2621 resin and
1Pd/C
2
subsequently reduced under a continuous hydrogen
flow in which the catalyst was immersed in a THF
suspension of the ion-exchanged materials. It is well
[
a] Reduction as described in Supporting Information Section 1.3.1; [b] Reduction as
described in Supporting Information Section 1.3.5; [c] Diameter estimated by lognor-
mal fit on size distribution data; [d] Skewness coefficient (SC) calculated by lognormal
parameters on size distribution data as reported in Ref. [36]; [e] From Ref. [21];
[
1,2,8,29]
known
that bromide or chloride ions, in combi-
[
f] From Ref. [22].
nation with an acid, are able to enhance the selectivi-
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1565
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
of the nanoparticles is strictly comparable with the pore diam-
[
32–39]
eters in the polymer mass.
Apparently, the polymer frame-
work of K2621 did not swell in THF: no gel phase was detected
by inverse size-exclusion chromatography (ISEC; Table S2). This
II
suggests that the reduction of Pd occurred on the surface of
the mesopores of the resin, which had diameters of 13 and
4
3 nm. These sizes are too large to give the effective control of
the nanoparticle size observed with swollen gel-type
[
34–35]
[37,39]
resins
or hyper-cross-linked resins.
Hence, the final
size of the metal nanoparticles is expected to be controlled
[
40]
only by the kinetics of their formation. In this case, the ratio
between their growth rate and the reduction-nucleation rate
should increase as the nanoparticles size and size distribution
Figure 1. Nanoparticle size and size distribution of the fresh catalysts in the
first set (different metal precursors, reduction with hydrogen) and of bench-
mark catalyst. 1Pd/C: commercial catalyst 1% Pd on carbon (benchmark);
[40]
become larger and broader. This ratio should, therefore, in-
crease in the following order of the metal precursor:
1
PdK/N/H: 1% Pd on K2621 from Pd(NO
3
)
2
precursor reduced with hydro-
]SO precursor reduced with
] precursor reduced
Pd(NO ) <[Pd(OAc) ]<[Pd(NH ) ]SO . Although it is clear that
3
2
2
3 4
4
gen; 1PdK/S/H: 1% Pd on K2621 from [Pd(NH
hydrogen; 1PdK/A/H: 1% Pd on K2621 from [Pd(OAc)
with hydrogen.
3
)
4
4
different morphologies of the nanostructured Pd phase ensue
from different precursors, it is not trivial to understand how
2
[
41]
II
this occurs. The speciation of Pd to be reduced is expected
to affect the rate of reduction directly. For instance, the
2+
II
[
Pd(NH ) ] ion, which is a stable complex of Pd , is also likely
3 4
more difficult to reduce, and the relevant 1PdK/S/H catalyst
contained clusters with a much larger size and size distribution
of metal nanoparticles than that of 1PdK/N/H and 1PdK/A/H. In
the synthesis of Pt nanoparticles, the final size and morpholo-
gy were affected by the presence of methylamine, a good
ligand for the metal to be reduced, which stabilized the most
[
41]
defective facets of the growing nanoparticles. It cannot be
ruled out that NH played a similar role when [Pd(NH ) ]SO
3
3 4
4
was used as the precursor in the preparation of our catalysts.
In line with the arguments above, the use of different reduc-
ing agents should lead to metal nanoparticles of different sizes
and morphologies. To gather information on the reducing
agent, a second set of catalysts was prepared upon the ion-ex-
change of K2621 with [Pd(NH ) ]SO and its subsequent reduc-
3
4
4
tion with aqueous solutions of either methanol (1:1, v/v; reflux
[
8]
[21]
temperature), formaldehyde (37%; reflux temperature), or
[
30]
sodium borohydride (ꢀ1m; room temperature). These treat-
ments produced 1PdK/S/M, 1PdK/S/F, and 1PdK/S/B, respec-
tively. It is well known that noble metals can be produced
upon reduction of their compounds with alcohols and alde-
[
42]
hydes under mild conditions, which can be applied for the
[
16,34,43]
preparation of Pd catalysts supported on resins.
Howev-
Figure 2. TEM images of A, B) fresh 1PdK/N/H material (both the scale bars
are 50 nm); C, D) fresh 1PdK/A/H material (the scale bars are 100 and 50 nm,
respectively); E, F) fresh 1PdK/S/H material (both the scale bars are 100 nm).
er, it is difficult to compare published data because different
[
30,44]
precursors and resins have been used so far.
For compari-
son, a second batch of 1PdK/S/H was produced upon reduc-
tion with H and in THF at a high temperature (333 K) and
2
[
22]
pressure (0.5 MPa). It will be hereafter referred to as 1PdK/S/
HPT. The size distributions of the metal nanoparticles in this
second set are reported in Figure 3.
In general, a poor relationship was observed between the
average size of Pd nanoparticles in the hydrogen-reduced cata-
lysts 1PdK/S/H, 1PdK/N/H, and 1PdK/A/H and the pore size in
K2621 swollen in THF, which is the solvent employed in the re-
duction stage. This is different from what is usually observed
with gel-type resins and hyper-cross-linked resins, in which the
reduction of the metal precursors and the formation of the
nanoparticles takes place inside the swollen polymer frame-
work or in permanent, very small mesopores so that the size
In this second set of catalysts the relationship between the
average values of the nanoparticle diameter and of the resin
mesopores was poor. 1PdK/S/B, 1PdK/S/F, and 1PdK/S/M were
prepared from suspensions of ion-exchanged K2621 in an
aqueous environment. ISEC analysis of K2621 in a sodium sul-
fate solution in water showed clearly that a gel phase was
formed in the resin (Table S2). This suggests that some of the
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1566
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 3. Nanoparticle size and size distribution of the fresh catalysts of the
second set (same metal precursor, different reduction agents). 1PdK/S/F: 1%
Pd on K2621 from [Pd(NH
PdK/S/B: 1% Pd on K2621 from [Pd(NH
sodium borohydride; 1PdK/S/M: 1% Pd on K2621 from [Pd(NH
cursor reduced with methanol; 1PdK/S/HPT: 1% Pd on K2621 from
Pd(NH ]SO precursor reduced with hydrogen; 1PdK/S/H: 1% Pd on K2621
from [Pd(NH ]SO precursor reduced with hydrogen at 333 K and 0.5 MPa.
3
)
4
]SO
4
precursor reduced with formaldehyde;
]SO precursor reduced with
]SO pre-
1
3
)
4
4
3
)
4
4
[
3
)
4
4
3
)
4
4
nanoparticles were formed in the swollen polymer framework,
although this did not lead to the full control of the metal
nanoparticle morphology. This is in line with previous findings
II
on the reduction of Pd loaded onto K2621 from [Pd(OAc) ]
2
[16]
and reduced with aqueous methanol or ethanol.
In the set of catalysts prepared from [Pd(NH ) ]SO , the
3
4
4
smallest nanoparticles were achieved in 1PdK/S/B (NaBH re-
4
duction), and the largest ones were observed in 1PdK/S/F
[22]
(
formaldehyde reduction ). In the latter case, the assessment
of the nanoparticle size was difficult, mainly because of the rel-
[22]
atively small number of nanoparticles detected. The size dis-
tribution of nanoparticles of 1PdK2621S/HPT (obtained from
[
Pd(NH ) ]SO and reduction with H /THF in an autoclave at
3 4 4 2
0
.5 MPa and 333 K) has a lognormal shape (Figure 3). The
nanoparticle size was much smaller and distributed more nar-
rowly in comparison with 1PdK/S/H (and generally smaller and
less dispersed in comparison with all the other catalysts too).
At a relatively high temperature and pressure, the reduction
rate was expected to be comparatively high and this could
Figure 4. TEM images of A, B) fresh 1PdK/S/F material (the scale bars are 50
and 10 nm, respectively); C, D) fresh 1PdK/S/B material (both the scale bars
are 100 nm); E, F) fresh 1PdK/S/M material (the scale bars are 50 and 5 nm,
respectively); G, H) fresh 1PdK/S/HPT material (both the scale bars are
100 nm).
[
40]
again allow for the comparatively small nanoparticle size.
Moreover, the total number of counted Pd aggregates was re-
markably high for 1PdK/S/H , which indicates that this reduc-
PT
tion protocol generated a large number of nanoparticles in the
outer portion of the polymer matrix (Figure 4).
The concentration profiles of the products (Figures S4 and
S6) illustrate that there are two reaction stages for almost all
the catalysts. In the first stage, the concentration of the prod-
ucts increased relatively quickly and linearly (at least in the first
1
0–30 min after the reaction onset). The slopes of these initial
Catalytic performance under batch conditions
linear portions of the kinetic plots represent the initial rates of
H O and H O production, respectively. Their values were used
The catalytic performance of our catalysts depended on the
metal precursor used in the preparation of materials and is il-
lustrated in Figures S3–S6. The concentrations of products
2
2
2
to calculate the initial selectivity as described in the Experi-
mental Section. The two catalysts obtained with hydrogen as
(
H O and water), as well as the H conversion and H O selec-
the reducing agent (1PdK/S/H and 1PdK/S/H ) showed a long
2
2
2
2
2
PT
tivity data are presented as a function of time. The commercial
catalyst, 1 wt% Pd/C, is included as the benchmark.
induction time in the production of water. The changes in the
product concentrations were much smaller, if any, in the
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1567
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
second stage (Figures S4 and S6): the amount of H O either
2
2
decreased slowly or not at all and the amount of water in-
creased slowly. H was always consumed completely in the first
2
stage so that the consumption of H O in the second stage
2
2
can be attributed only to its dismutation [Scheme 1, Eq. (4)].
Notably, the H mass balance had an error of 3% only, hence
2
the induction time is real and must depend on features of the
catalysts.
From the initial rates of cumulative production (H O+H O )
2
2
2
and of H O production, it is possible to calculate the initial
2
2
productivities (Table 2).
The analysis of the initial rates of formation of the individual
products also allowed us to estimate the initial selectivity
values towards H O (Table 2). This method, however, is not
2
2
feasible if there is an induction period in the formation of one
of the products, which was observed in the production of
water over 1PdK/S/HPT and 1PdK/S/H. In these cases, the cata-
lysts seemed to undergo a transformation from a poorly active
form into a much more active one for the production of water.
To circumvent this problem, the catalytic performance can
Figure 5. t
1
=
2
1
=
2
[min] and selectivity at t [%] for 1Pd/C, 1PdK/N/H, 1PdK/A/H,
and 1PdK/S/H. 1Pd/C: commercial catalyst 1% Pd on carbon (benchmark);
PdK/N/H: 1% Pd on K2621 from Pd(NO precursor reduced with hydro-
gen; 1PdK/S/H: 1% Pd on K2621 from [Pd(NH ]SO precursor reduced with
] precursor reduced
1
3 2
)
3
)
4
4
hydrogen; 1PdK/A/H: 1% Pd on K2621 from [Pd(OAc)
with hydrogen.
2
be also evaluated in terms of t = (the time required to convert
1
2
5
0% of H ) and selectivity at t
1
(S ; Table 2). This approach is
1
2
=
2
=
2
[
18,45]
applicable to all the catalysts and will be used hereafter.
In the set of the catalysts reduced with hydrogen, at relative-
ly low pressures and temperatures, the activity decreased gen-
In line with previous discoveries,
the best results are ach-
ieved with relatively large nanoparticles and a poor per-
formance is correlated to the small nanoparticles. The selectivi-
ty towards H O in CDS is expected to depend on the topology
erally (increasing t
Figure 5). However, the selectivity increased with the increas-
ing nanoparticle size. As the result, although t = was around
1
=2) as the Pd nanoparticles increased in size
2
2
(
of the metal surface, which dictates the number and the
[
46]
1
nature of the catalytic sites. The largest nanoparticles, which
have a flattened surface that may be more favorable to ach-
2
eight times higher for 1PdK/S/H than for 1Pd/C, its S
1
in-
=
2
[
54,47]
creased almost fourfold. In general, all the polymer-supported
catalysts were more selective than 1 wt% Pd/C and the two
best ones gave selectivities comparable to those achieved with
ieve high selectivity in CDS,
were found in 1PdK/S/H. In
general the size was modulated by the nature of the Pd pre-
cursor, likely because of the different kinetics of the nanoparti-
[1–5,25]
[40]
selectivity enhancers in other systems.
cle formation.
If the same reducing conditions with different precursors
were applied in the catalyst preparation, the catalytic per-
formance was apparently controlled by the particle size. The
selectivity increased more or less linearly with the increasing
nanoparticle size and the activity decreased steadily (Figure 5).
Interestingly, the benchmark catalyst (1Pd/C) fits the correla-
tion, which suggests that in this set of catalysts the interaction
of the metal with the support does not play a significant role.
The catalysts obtained upon the reduction of the precursors
with hydrogen were analyzed by using X-ray photoelectron
II
spectroscopy (XPS) to evaluate the atomic ratio between Pd
0
0
II
0
and Pd . Indeed, the Pd /Pd +Pd value depends on the reduc-
[
18,45]
tion procedure and can affect the catalytic performance.
0
The proportion of Pd is generally around 25% and not much
different in 1Pd/C, 1PdK/N/H, and 1PdK/A/H (Table S3). In
1PdK/S/H it is around 37%, which could contribute to some
Table 2. Results of the catalytic runs under batch conditions over the Pd catalysts supported on K2621.
[f]
Catalyst
Pd
amount
Initial
Initial
cumulative
productivity
Initial H
productivity
2
O
2
Initial
selectivity
t
1
=
2
Selectivity
[
a]
[d]
[e]
[e]
cumulative
at t
=
2
1
[
b]
[c]
rate
1
1
1
1
1
1
1
1
Pd/C
1.41
1.43
1.43
1.42
1.42
1.45
1.44
1.45
235
6801
3792
4032
1160
2460
2343
1388
n.d.
1401
877
1605
760
977
792
313
n.d.
21
23
40
66
40
34
23
n.d.
6
18
23
40
64
41
34
24
50
PdK/N/H
PdK/A/H
PdK/S/H
PdK/S/M
PdK/S/F
PdK/S/B
135.0
143.3
40.9
83.8
83.6
48.3
29.3
10
10
33
19
17
29
38
[g]
[h]
PdK/S/HPT
5
À3 À1
PdÀ1 À1
PdÀ1 À1
; [d] molH2 O2 mol h ; [e] %; [f] Time [min] for 50% H
[
[
a] mol·10 ; [b] mmolH2 O2 þH2
O
dm
h
; [c] molH2 O2 þH2
O
mol
h
2
conversion; [g] From Ref. [11];
h] From Ref. [12].
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1568
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
extent to the better selectivity of this catalyst, in line with the
0
[18]
beneficial effect of Pd for the direct synthesis.
However, the effect of the oxidation state, if any, here is
minor, probably because the differences from one catalyst to
another are not so great. Hence, the most important morpho-
logical parameter was apparently the nanoparticle size when
hydrogen was the reducing agent.
Some changes of the nanoparticle size distributions were
observed in the spent catalysts compared to the pristine mate-
rials (Figure S5). In the case of the 1Pd/C sample, a slightly
broader distribution with an average size shifted to 4.0 nm
(
from 3.8 nm) was observed, which indicates moderate sinter-
[48]
[49]
ing as reported by Abate et al. and Chinta and Lunsford.
In the case of 1PdK/N/H and 1PdK/S/H, no changes in the aver-
age nanoparticle size were observed, although the size distri-
bution was somewhat narrower than that in the fresh catalyst
for 1PdK/N/H.
Figure 6. t
1
1
=
2
1
=
2
[min] and selectivity at t [%] for 1Pd/C, 1PdK/S/B, 1PdK/S/M,
PdK/S/F, 1PdK/S/H, and 1PdK/S/HPT. 1Pd/C: commercial catalyst 1% Pd on
carbon (benchmark); 1PdK/S/F: 1% Pd on K2621 from [Pd(NH ]SO precur-
sor reduced with formaldehyde; 1PdK/S/B: 1% Pd on K2621 from
Pd(NH ]SO precursor reduced with sodium borohydride; 1PdK/S/M: 1%
Pd on K2621 from [Pd(NH ]SO precursor reduced with methanol; 1PdK/S/
H: 1% Pd on K2621 from [Pd(NH ]SO precursor reduced with hydrogen;
PdK/S/HPT: 1% Pd on K2621 from [Pd(NH ]SO precursor reduced with hy-
drogen at 333 K and 0.5 MPa.
3
)
4
4
The use of different reducing agents with [Pd(NH ) ]SO as
3
4
4
the metal precursor led to catalysts with Pd nanoparticles of
different sizes. The smallest nanoparticles were obtained with
[
3
)
4
4
3
)
4
4
3
)
4
4
NaBH , which is the most reactive of the reducing agents used
4
1
3
)
4
4
and led to a fast reduction at room temperature. Methanol
and formaldehyde are more sluggish and their water solutions
were used at reflux temperature and for relatively long time.
II
Moreover, to achieve with hydrogen a faster reduction of Pd ,
it was used at 333 K and 0.5 MPa. The performance of this set
of catalysts (Figures S3 and S4) was then evaluated under ex-
perimental conditions similar to those employed for the first
set. The experimental data were also treated in the same way
showed that the Pd{111} face was more selective in CDS than
[
47]
the Pd{100} face. Moreover, it was observed that different
[
52]
reduction agents affect the morphology of Pd nanoclusters.
In this case, gas adsorption techniques, such as CO chemisorp-
tion monitored by diffuse reflectance infrared Fourier trans-
(
Table 2) using t =2
Apparently, the catalysts obtained by using reducing agents
other than hydrogen, are less active (higher t =2) and selective.
1
and S =
1
.
2
[
50]
form spectroscopy could be helpful in principle, but for the
reasons illustrated above, they are not applicable for polymer-
supported catalysts.
1
At variance from the previous set of catalysts, in this one no
relation between the nanoparticle size and the catalytic per-
formance was found (Figure 6). Differently from the catalysts
prepared by the reduction of the Pd precursor with hydrogen,
this second set was prepared upon reduction in an aqueous
environment. Under these conditions, K2621 formed a swollen
gel phase and the Pd nanoparticles, or at least some of them,
were likely formed inside this phase. One of the possible rea-
sons for the apparent lack of any relationship between the cat-
alytic performance and the nanoparticle size could be a differ-
ent balance between the number of nanoparticles inside the
gel phase and those on the surface of the mesopores. These
two kinds of particles are expected to interact differently with
the support and, therefore, to show different catalytic behav-
iors. Gas adsorption techniques could be helpful to assess their
balance. For instance, these techniques were employed recent-
The catalyst obtained by reduction with hydrogen at a rela-
tively high pressure and temperature, 1PdK/S/H , was included
PT
in this second set of catalysts (the reduction conditions are
anyway different from those used for the previous set). The in-
crease in the H pressure and in the reduction temperature led
2
to a remarkable decrease in the nanoparticle size and to
a much narrower size distribution (Figure 3). In spite of this,
1PdK/S/HPT is not only appreciably less selective than 1PdK/S/H
but also slightly less active. This was the result of a shorter in-
duction time in the production of water for 1PdK/S/HPT and
a slower initial production of H O (Figure S4). In line with the
2
2
above reasoning, the worse catalytic performance in compari-
son with 1PdK/S/H could be the result of a different balance of
the exposed crystallographic faces of Pd. In addition, the
II
[18]
amount of Pd , which is detrimental for the reaction,
is
[
50]
ly for the characterization of the Pd surface of CDS catalysts.
higher in 1PdK/S/HPT than in 1PdK/S/H.
However, if flexible materials, such as ion-exchange resins, are
used as the catalyst support, the metal surface can be inacces-
In any case, it is clear that any catalyst supported by K2621
is much better than Pd/C. This could arise from the acidity of
[51]
[25]
sible to the gas to be adsorbed, and therefore, in our case
these methods do not seem feasible. Another possible explan-
ation for the lack of any relationship between the catalytic per-
formance and the nanoparticle size could be a different bal-
ance between the areas of the exposed crystallographic faces
if different reducing agents are employed. Recently, Kim et al.
the sulfonic groups contained in the resin.
However, our
[
16]
data do not support this hypothesis. In the first place, the
presence of CO as the diluent of the reacting gases made the
2
reaction medium acidic and this could have hindered the influ-
ence of the acidic groups of the resin. Moreover, even if this
was not true, 1PdK/N/H was practically as selective as 1Pd/C,
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1569
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
although it contained sulfonic groups in the acidic form (in the
tion rates obtained under both batch and semi-batch condi-
tions were relatively similar, although the operational tempera-
ture was not the same and the H O profile changed after ꢀ
absence of bases during the reduction step, the oxidation of
+
H produces H , which should replace the metal ions that are
2
2
2
reduced).
90 min, as a consequence of the side-reactions.
These results are of extremely high value as it was demon-
In conclusion, the catalysts supported by K2621 prepared
using [Pd(NH ) ]SO and H2 as the reducing agent gave
[53]
strated that the size of the Pd nanoparticles is not the only
parameter that affects the direct synthesis.
3
4
4
a better performance in the CDS than the benchmark catalyst
because they showed the best balance between the rates of
H O and H O formation and the rates of the side-reactions. In
2
2
2
Catalytic results under semi-batch conditions
this way it was possible to achieve a final H O concentration
2
2
The best catalysts found in the batch experiments were tested
by using a semi-batch reactor under conditions similar to
of 0.5 wt% without the use of any selectivity enhancers, such
as acids or halides. These promising results confirm the hy-
pothesis that the reaction conditions play an important role in
[12,49]
those reported by Lunsford et al.
The results were ob-
tained under a total pressure up to 50 bar and without any se-
lectivity enhancers. The catalytic results are reported in Table 3
and Figures S6 and S7.
the direct synthesis of H O and that the catalyst has to be op-
2 2
[
54]
timized to the reaction conditions used.
Conclusions
Table 3. Results of the catalytic runs under semi-batch conditions over the Pd cata-
lysts supported on K2621.
The aim of the present study was to understand how
different conditions for the preparation of new, ad-
vanced polymer-supported monometallic Pd catalysts
can affect the direct synthesis of H O . For this pur-
PdÀ1 À1
[a]
[a]
Sample
H
2
O
2
Rates [molH2 O2 mol
hydrogenation
h
]
C
S
2 2
H O
[wt%]
formation
dismutation
[%]
[%]
2
2
pose, we prepared two sets of catalysts, one using
different metal precursors and the same reduction
procedure and the other using the same metal pre-
cursor but different reduction procedures. The cata-
lysts thus obtained were characterized by TEM to
assess the morphological differences that arose from
the different preparation conditions and to disclose
1
1
1
Pd/C
PdK/S/H
PdK/S/HPT
185
594
499
10754
269
298
254
197
180
66
43
21
3
25
36
0.1
0.5
0.4
[
a] Conversion of H
2
and selectivity towards H
2
O
2
after 5 h.
The production rates of H O and H O were more or less
the existence of a relationship, if any, with the catalytic per-
formance (activity, selectivity).
2
2
2
constant up to 5 h (Figure S9), but the selectivity changed
during the catalytic tests. As expected from the results ob-
tained under batch conditions, the benchmark catalyst, 1PdC,
When different precursors were reduced with the same
agent (hydrogen flowing through a THF suspension of the ma-
terial to be reduced) were used, the activity decreased and the
selectivity increased with the increasing size of the metal nano-
particles; in particular, the selectivity correlated very well with
was the most active with 66% of H conversion, but gave the
2
poorest selectivity. The 1PdK/S/H sample gave a moderate ini-
tial selectivity (ꢀ50%), but it decreased almost linearly with
time to 25%, and the conversion was almost constant (40–
II
0
the nanoparticle size. The Pd /Pd ratio played only a minor
role, if any. Apparently, the morphological changes, which con-
trolled the nature and the relative number of the catalytic sites
on the surface, depended only on the nanoparticle size. This
indicates that the number of relatively poorly active and highly
selective sites would increase at the expense of exceedingly
highly active and poorly selective sites with the increasing
nanoparticle size. The simple change of their relative numbers,
and not of their intrinsic nature, is enough to explain this
result. Therefore, the morphology of the metal surface apart
from its curvature could be even the same in all these catalysts
and the simple flattening of the active surface is desirable as
far as the selectivity is concerned.
4
3%). This catalyst gave the highest final concentration of
H O (ꢀ0.5 wt%). However, in this case the least active and
2
2
most selective catalyst was 1PdK/S/H . It had an initial selectiv-
PT
ity towards H O2 of around 70%, which decreased to 35%
2
after 5 h, but in spite of a final conversion as low as 20%, the
final concentration of H O was ꢀ0.4 wt% (almost as high as
2
2
with 1PdK/S/H).
In these tests it was possible to estimate the H O formation
2
2
rate (Table 3). In addition, the estimated rates of H O hydroge-
2
2
nation and dismutation were obtained by experiments using
a methanol solution of 2 wt% H O (Table 3). Compared to the
2
2
catalysts supported on K2621, Pd/C had the lowest rate of
H O formation and the highest rate for its dismutation and
This is not the case if different reduction conditions (reduc-
ing agent and reaction medium or pressure and temperature)
were applied for a single Pd precursor ([Pd(NH ) ]SO ). In this
2
2
hydrogenation. The hydrogenation of the product is particular-
ly high (two orders of magnitude larger than all the others),
which was apparently responsible for its very poor selectivity.
However, the rates of the side-reactions for the resin-support-
ed catalysts were all close to each other so that the main dif-
ference in the catalytic performance arose from the relative
rates of H O and H O formation. In particular the H O forma-
3
4
4
case, there must be differences that depend on the reduction
conditions. Our results demonstrate clearly that in the direct
synthesis of H O , the control of the dimensions of the nano-
2
2
clusters is not always enough. Morphological features different
from the curvature of the surface can also be important. An in-
2
2
2
2
2
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1570
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
spection of the literature suggests that the balance of the
areas of the exposed crystal faces could play an important
role. Moreover, with flexible supports such as ion-exchange
resins (such as K2621), conditions that favor the swelling of the
polymer framework during the preparation of the catalyst
could lead to different results in comparison to conditions
under which swelling is negligible. In this context, the balance
between the number of active nanoparticles formed in the
swellable domains of the support and the nanoparticles
formed on the surface of permanent pores is expected to
change and to lead to different catalytic performances. Howev-
er, further work is needed to provide a stronger experimental
support to these hypotheses.
Typically, K2621 (2.0 g) was suspended in distilled water (10 mL)
and left to stand for 2 h. An aqueous solutions prepared with Pd
precursor (0.188 mmol) was added (Table 1). The suspension was
left to react overnight under mechanical stirring, and then the
product was recovered by filtration and washed carefully with dis-
tilled water (3”10 mL) on the filter (water was always replaced
with THF if [Pd(OAc) ] was used). The mother liquor (filtrate and
2
water from washing) was analyzed for the unreacted metal by ICP-
MS. In the case of THF, the mother liquor was evaporated by using
a rotavapor, and the dry residue was dissolved in a few milliliters
of aqua regia. This solution was then analyzed for the metal by
ICP-MS. The amount of unreacted metal after the ion-exchange
step was less than 0.1 wt% of the respective precursor amounts.
The mass balance of the metal showed that their uptakes were
always complete and the experimental metal loading was equal to
the nominal value for each sample.
The best combination of metal precursor and reduction con-
ditions was [Pd(NH ) ]SO for the ion-exchange step and hy-
3
4
4
drogen flowing through a THF suspension of Pd-exchanged
K2621 at room temperature and atmospheric pressure. This
gave a catalyst that featured the largest average size of the Pd
nanoparticles and the broadest distribution thereof. This cata-
lyst showed an initial rate of H O production of 760 and
Reduction protocols
Different reduction protocols were performed after the metalation
of the K2621 with the [Pd(NH ) ]SO precursor.
3
4
4
2
2
À1 À1
5
94 molH2O2 molPd
h
under batch and semi-batch condi-
Reduction with hydrogen
tions, respectively. It was, therefore, the most selective catalyst:
the average initial selectivity was 66% (considering both ex-
periments) and the instant value topped 70% shortly after
The ion-exchanged resin (generally a beige-colored solid) was sus-
pended in THF (50 mL) in a vessel, which was then flushed with H2,
directly bubbled into the suspension at RT and ambient pressure,
for 5 h. The resulting solid was collected by filtration and washed
carefully with THF. Finally, the solid (dark gray color) precipitate
was dried in an oven at 383 K overnight and then ground with
pestle and mortar. Data on the catalysts obtained in this way
5
0% conversion under batch-wise operation. This catalyst, as
well as most of the catalysts supported on K2621, outper-
formed the benchmark 1 wt% Pd/C catalyst under the reaction
conditions employed in this work. Moreover, no selectivity en-
hancers in the form of acids or halide ions were present.
(
1PdK/A/H, 1PdK/N/H, 1PdK/S/H, where A stands for acetate, N for
nitrate, and S for sulfate) are collected in Table 1.
Experimental Section
Reduction with formaldehyde
Catalyst Synthesis
The details are reported in Ref. [21]. Briefly, the beige material, al-
ready swollen in distilled water, was suspended in 37% aqueous
formaldehyde solution (50 mL) and left to react under reflux for
Unless otherwise stated, all the reagents and the materials were
used as received. A batch of Lewatit K2621 (sulfonated polystyr-
ene-divinylbenzene macroreticular, ion-exchange resin (SPS); ex-
change capacity=1.92 mmolg ) was kindly provided by Lanxess;
it was used after washing carefully with water and methanol.
3
h. The black product was recovered by vacuum filtration and
À1
washed carefully on the filter with distilled water (3”10 mL). The
solid was dried in an oven at 383 K overnight.
[
[
Pd(NH ) ]SO and Pd(NO ) were purchased from Alfa Aesar, and
3 4 4 3 2
Pd(OAc) ] was purchased from Sigma Aldrich; sodium thiosulfate
Data on the catalysts obtained in this way (1PdK/S/F, where S
stands for sulfate and F for formaldehyde) are collected in Table 1.
2
pentahydrate (99.5%), potassium iodide, starch, concentrated sul-
furic acid, 37 wt% aqueous formaldehyde, sodium borohydride,
and methanol were purchased from Sigma–Aldrich; THF was sup-
plied by Sigma–Aldrich and used after distillation. HPLC-grade
methanol (99.99%) was obtained from J.T. Baker; H , O , and CO
2
Reduction with methanol
2
2
The details are reported in Ref. [15]. The (beige) solid, already swol-
len in distilled water, was suspended in 1:1 (v:v) methanol/water
solution (20 mL) and left to react under reflux conditions for 3 h.
The black product was recovered by vacuum filtration and washed
carefully over a filter with distilled water (3”10 mL). The solid (dark
gray-black), was dried in an oven at 383 K overnight and then
ground with a pestle and mortar. Data on the catalysts obtained in
this way (1PdK/S/M, S stands for sulfate and M for methanol) are
collected in Table 1.
(
99.999% mol/mol purity) and were provided by AGA gas (Linde
group). Methanol for Karl Fischer titration, Hydranal composite 2,
and ammonium molybdate tetrahydrate were purchased from
Fluka. The reference catalyst was a standard, reduced 1 wt% Pd/C
(
Alfa Aesar), which was used as received.
Pd ion exchange
All the samples were prepared with a Pd loading of 1 wt%. Al-
though [Pd(NH ) ]SO and Pd(NO ) are soluble in water, [Pd(OAc) ]
is soluble only in less polar solvents, and therefore, two protocols
of metalation were employed, the former in water, the latter in
THF.
3
4
4
3 2
2
Reduction with sodium borohydride
The beige material, already swollen in distilled water, was suspend-
ed in water (10 mL). A solution of NaBH (0.58 g) in water (15 mL)
4
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1571
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
was added to the beige suspension under moderate stirring. The
color of the solid immediately turned to dark gray. After 2 h, the
solid became black, and the product was recovered by vacuum fil-
tration and washed carefully on the filter with distilled water (5”
pressure and temperature (275 K), the stirrer was switched off, and
H was fed up to the desired amount. The number of moles of hy-
2
drogen introduced into the reactor (nH2,0) was calculated from the
[7]
pressure drop of hydrogen in the precylinder. The final composi-
tion of the gases was 2.5% H , 25.5% O , and 72% CO . When the
1
0 mL). The solid was dried in an oven at 383 K overnight and
2
2
2
then ground with a pestle and mortar. Data on the catalysts ob-
tained in this way (1PdK/S/B, where S stands for sulfate and B for
borohydride) are collected in Table 1.
delivery of the desired amount of H was complete, the stirrer and
2
the pump (in recirculation mode) were switched on again. This
was taken as start of the reaction. Unless otherwise stated, the re-
action was stopped after 3 h. Portions of the liquid phase were
withdrawn periodically from a six-way valve placed in the recircula-
Reduction in autoclave with 0.5 MPa of hydrogen at 333 K
tion line of the reactor for H O and water analysis; their volume
2 2
(
Vtaken) was small enough to ensure that the amount of the catalyst
The details are reported in Ref. [22]. Briefly, the beige solid recov-
ered from the ion-exchange reaction was suspended in THF
per unit volume of the liquid phase was practically constant
throughout the whole test. The concentration of the products, in
the samples ([H O ], [H O]’ [mm]) were determined by iodometric
and Karl Fischer titrations. The initial concentration of water, [H O] ,
was also determined by Karl Fischer titration of the solvent and
used to calculate the concentration of water effectively produced
(
50 mL) in a glass vessel inserted into an autoclave. After flushing
2
2
2
three times with H , the autoclave was filled with H and the re-
2
2
2
0
duction was performed at 0.5 MPa H at 333 K for 5 h. After the au-
2
toclave was cooled and vented, the catalyst was recovered by
vacuum filtration and washed carefully on the filter with distilled
water (5”10 mL). The solid was dried in an oven at 383 K over-
night and then ground with a pestle and mortar.
by the reaction, [H O] [Eq. (5)]:
2
0
½
H OŠ ¼ ½H OŠ À½H OŠ
ð5Þ
2
2
2
0
Data on the catalysts obtained in this way (1PdK/S/H , where S
PT
stands for sulfate and HPT for hydrogen at high temperature and
high pressure) are collected in Table 1.
These data were used to monitor the progress of the reaction and
its selectivity. Finally, the cumulative yield (C) was calculated
[
Eq. (6)]:
Characterization
C ½%Š ¼ 100 Á f½H
2
O
2
Š
t
þ½H
2
OŠ
t
g Á V
L
=nH2,0
ð6Þ
ICP-MS measurements were performed by using a Perkin Elmer
Sciex ICP Mass Spectrometer 6100 DRC Plus. The analysis was per-
formed in the quantitative standard mode. Consequently, the sam-
ples were prepared upon diluting the mother liquor from the met-
alation experiments (aqueous solutions of the unreacted metal pre-
cursor) to a known fixed volume.
The selectivity towards H O2 at time t (SH2O2,t) was calculated as
2
[
Eq. (7)]:
SH2O2,t ½%Š ¼ 100 Á ½H O Š =f½H O Š þ½H OŠ g
ð7Þ
2
2
t
2
2
t
2
t
The initial productivities (rates of H O and H O production per
2
2
2
Karl Fisher titrations were performed by using a Titrino GP 736
from Metrohm.
mol of metal) were calculated in a similar way as described in
Ref. [21]. The initial (up to 10–30 min from the onset of the reac-
tion, which depends on the activity of the catalyst) sections of the
kinetic plots were linear and their slopes were provided by linear
interpolation. The division of the slope values by Pd moles yielded
the apparent rates of water and H O formation for each catalyst
TEM analyses were performed by using an energy-filtered transmis-
sion electron microscope (EFTEM, LEO 912 OMEGA, LaB6 filament,
1
20 kV). Samples were prepared by suspending a few milligrams of
the powder in high-purity isopropyl alcohol (or ethanol). After soni-
cation (30 s), a small droplet (5 mL) of the suspension was trans-
ferred onto a holey-carbon film coated Cu grid, which was intro-
duced into the microscope.
2
2
i
i
À3 À1
(rH2O, r
[mmolproduct/s dm h ]. Their sum is the initial cumula-
). The initial selectivity towards H O was calculated
H2O2
i
tive rate (r
cumul.
2
2
as [Eq. (8)]:
XPS spectra were recorded by using a Kratos Axis Ultra electron
spectrometer equipped with a delay line detector. A monochro-
mated AlKa source operated at 150 W, hybrid lens system with
i
i
Initial H O2 selectivity ½%Š ¼ 100 Á r
=r
cumul:
ð8Þ
2
H2O2
2
magnetic lens, an analysis area of 0.3”0.7 mm , and charge neu-
The initial rates were transformed into the respective initial pro-
À3 À1
tralizer were used for the measurements. The binding energy (BE)
scale was referenced to the C1s line of aliphatic carbon, set at
BE=285.0 eV. The spectra were processed with the Kratos soft-
ware.
ductivity (cumulative and of H
lows [Eqs. (9) and (10)]:
2
O
2
; mmolproduct/s mmolPd h ) as fol-
i
=f10À3 Á molPd
Initial cumulative productivity ¼ r cumul: Á V
g
ð9Þ
L
i
À3
Initial H O2 productivity ¼ r
Á V =f10 Á molPdg
ð10Þ
2
H2O2
L
Batch experiments
in which V is the volume of the liquid phase.
[6,18,22]
L
Catalytic tests were performed as reported previously.
Briefly,
the catalyst (0.15 g) was loaded into the reactor (600 mL). The reac-
tor was then closed, CO (1.84 MPa) and O (0.6 MPa) were fed di-
rectly from cylinders at 296 K. After stabilization of the gas pres-
2
2
Semi-batch experiments
sure, 420 mL (V ) of methanol was fed by using a HP pump at
The semi-batch experiments were performed in a 300 mL stainless-
steel reactor (Parr) in the absence of any selectivity enhancers.
L
À1
[13]
4
mLmin . Then the reactor was cooled to 275 K. The equilibrium
between the liquid and the gas phase was reached upon starting
the stirrer, which operated at 1000 rpm. After the stabilization of
The catalyst (70 mg) was loaded into the injection system of the
reactor with 2–3 mL of methanol and left to stand overnight.
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1572
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Methanol (180 mL; HPLC grade) was placed in the reactor, and the
injection system was connected (t0 is the injection of catalyst
slurry). H , O , and N were bubbled through a stainless-steel frit
(ŠA). Financial support from the Academy of Finland is gratefully
acknowledged. P.B. gratefully acknowledges the Kempe Founda-
tions (Kempe Stiftelserna) for support. In Sweden, also the Bio4-
Energy program and Wallenberg Wood Science Center under the
auspices of the Knut and Alice Wallenberg Foundation are ac-
knowledged. We are indebted to Dr. K. Je ˇr µbek (Institute of
Chemical Process Fundamentals, Academy of Sciences of the
Czech Republic) for providing the ISEC data of K2621.
2
2
2
(
(
2 mm) into the reaction medium under vigorous stirring
1000 rpm) and cooled to 275 K with a O /H /N 20:4:76 gas mix-
2
2
À1
2
ture and a total flow rate of 300 stdmLmin , in which the molar
ratio of H to O in the feed stream was fixed at 0.2. The methanol
2
2
was presaturated with the gas mixture before the introduction of
catalyst slurry through the injector (t ). The catalytic run went on
0
for 5 h.
Portions of the liquid phase were withdrawn periodically from
a six-way valve placed in the recirculation line of the reactor for
H O and water analysis; their volume (Vtaken) was small enough to
Keywords: hydrogen peroxide · nanoparticles · palladium ·
reduction · supported catalysts
2
2
ensure that the amount of the catalyst per unit volume of the
liquid phase was practically constant during the whole test cycle.
The concentration of H O and water was determined by iodomet-
[
[
1] J. M. Campos-Martin, G. Blanck-Brieva, J. L. G. Fierro, Angew. Chem. Int.
Ed. 2006, 45, 6962–6984; Angew. Chem. 2006, 118, 7116–7139.
2] C. Samanta, Appl. Catal. A 2008, 350, 133–149.
2
2
ric titration and the Karl Fisher method, respectively.
The gas mixture that flowed out the autoclave was analyzed for
[3] J. García-Serna, T. Moreno Rueda, P. Biasi, M. J. Cocero, J. P. Mikkola, T.
unreacted H by using GC (Agilent 6890GC) equipped with a ther-
Salmi, Green Chem. 2014, 16, 2320–2343.
2
[
[
[
4] J. K. Edwards, S. J. Freakley, R. J. Lewis, J. C. Pritchard, G. J. Hutchings,
Catal. Today 2015, 248, 3–9.
5] R. Dittmeyer, J. D. Grunwaldt, A. Pashkova, Catal. Today 2015, 248, 149–
mal conductivity detector (TCD) and a capillary column (HP
MOLSIV, 30 m, ID 0.53 mm, film 25 mm). The conversion of H and
2
selectivity to H O were calculated according to Equations (11) and
2
2
159.
(
12):
6] P. Biasi, N. Gemo, J. R. Hernandez Carucci, K. Eranen, P. Canu, T. O. Salmi,
Ind. Eng. Chem. Res. 2012, 51, 8903–8912.
[7] N. M. Wilson, D. W. Flaherty, J. Am. Chem. Soc. 2016, 138, 574–586.
[8] P. Biasi, J. Garcia-Serna, A. Bittante, T. Salmi, Green Chem. 2013, 15,
moles of H2 unreacted
H2 conversion ð%Þ ¼ 100 Á moles of H2 supplied
ð11Þ
ð12Þ
2502–2513.
moles of H O formed
2
2
[
9] I. Huerta, P. Biasi, J. Garcia-Serna, Cocero, M. J. Cocero, J. P. Mikkola, T.
Salmi, Catal. Today 2015, 248, 91–100.
H O selectivity ð%Þ ¼ 100 Á
2
2
moles of hydrogen reacted
[
[
[
10] P. Biasi, F. Menegazzo, F, Pinna, K. Eranen, P. Canu, T. O. Salmi, Ind. Eng.
Chem. Res. 2010, 49, 10627–10632.
in which the moles of H reacted is the sum of the moles of H O
2
2
2
11] M. Piccinini, E. N. Ntainjua, J. K. Edwards, A. F. Carley, J. A. Moulijn, G. J.
Hutchings, Phys. Chem. Chem. Phys. 2010, 12, 2488–2492.
12] a) Q. Liu, J. H. Lunsford, Appl. Catal. A 2006, 314, 94–100; b) N. Gemo, P.
Biasi, P. Canu, F. Menegazzo, F. Pinna, A. Samikannu, K. Kordàs, T. O.
Salmi, J.-P. Mikkola, Top. Catal. 2013, 56, 540–549.
13] a) A. Bernardini, N. Gemo, P. Biasi, P. Canu, J. P. Mikkola, T. Salmi, R.
Lanza, Catal. Today 2015, 256, 294–301; b) P. Biasi, S. Sterchele, F. Biz-
zotto, S. Lindholm, P. Ek, J. Bobacka, J.-P. Mikkola, T. Salmi, Catal. Today
2015, 246, 207–215.
and H O formed.
2
To estimate the activity of catalysts, in terms of the side-reactions
such as hydrogenation or dismutation of H O , the same equip-
2
2
ment was used in additional experiments. H O hydrogenation ex-
2
2
[
periments were performed in a similar way as the H O synthesis
2
2
experiments by simply replacing O2 with N2 and introducing
a 30 wt% H O commercial solution to give a reaction mixture that
2
2
contained 2 wt% H O . Dismutation experiments were performed
[14] a) T. Salmi, N. Gemo, P. Biasi, J. Garcia-Serna, Catal. Today 2015, 248,
108–114; b) N. Gemo, P. Biasi, T. O. Salmi, P. Canu, J. Chem. Thermodyn.
2
2
in the same way with a gas flow composed only of N . No con-
2
2
2
012, 54, 1–9; c) N. Gemo, P. Biasi, P. Canu, T. O. Salmi, Chem. Eng. J.
012, 207–208, 539–551.
sumption of H O was detected up to 10 h.
2
2
The H O productivity, hydrogenation rate, and dismutation rate
were estimated as follows [Eqs. (13)–(15)]:
[15] C. Burato, P. Centomo, M. Rizzoli, A. Biffis, S. Campestrini, B. Corain, Adv.
Synth. Catal. 2006, 348, 255–259.
2
2
[
[
[
16] C. Burato, S. Campestrini, Y.-F. Han, P. Canton, P. Centomo, B. Corain,
Appl. Catal. A 2009, 358, 224–231.
17] P. Biasi, F. Menegazzo, P. Canu, F. Pinna, T. O. Salmi, Ind. Eng. Chem. Res.
2013, 52, 15472–15480.
18] N. Gemo, S. Sterchele, P. Biasi, P. Centomo, P. Canu, M. Zecca, A. Shchu-
karev, K. Kordas, T. Salmi, J. P. Mikkola, Catal. Sci. Technol. 2015, 5, 3545–
ds
Á V=f10À3 Á molPdg
H O productivity ¼ r
ð13Þ
ð14Þ
ð15Þ
2
2
H2O2
h
Á V=f10À3 Á molPdg
H O hydrogenation rate ¼ r
2
2
H2O2
d
Á V=f10À3 Á molPdg
H O dismutation rate ¼ r
2
2
H2O2
3555.
[
19] A. Guyot in Synthesis and Separations using Functional Polymer (Eds.:
X
H2O2
À3 À1
in which r
[mmolH2O2 dm h ] is the rate of reaction that pro-
D. C. Sherrington, P. Hodge), Wiley, New York, 1988.
duces (ds) or consumes (h and d) H O in the medium, respectively,
and V is the volume of the liquid phase.
[20] R. L. Albright, I. J. Jakovac, Catalysis by Functionalized Porous Organic
Polymers, Bulletin IE-287, Rohm&Haas Company, Philadelphia, 1985,
pp. 4–6.
2
2
[21] S. Sterchele, P. Biasi, P. Centomo, P. Canton, S. Campestrini, T. Salmi, M.
Zecca, Appl. Catal. A 2013, 468, 160–174.
Acknowledgements
[22] S. Sterchele, P. Biasi, P. Centomo, P. Canton, S. Campestrini, A. Shchukar-
ev, A. R. Rautio, J. P. Mikkola, T. Salmi, M. Zecca, Catal. Today 2015, 248,
4
0–47.
Lanxess is acknowledged for kindly providing the K2621 resin.
This work is part of the activities of Johan Gadolin Process
Chemistry Centre (PCC), financed by the Šbo Akademi University
[
[
23] B. Corain, M. Basato, M. Zecca, G. Braca, A. M. Galletti, S. Lora, G. Palma,
E. Guglielminotti, J. Mol. Catal. 1992, 73, 23–41.
24] W. Keim, P. Mastrorilli, C. Nobile, N. Ravasio, B. Corain, M. Zecca, J. Mol.
Catal. 1993, 81, 167–184.
(
ŠA). S.S. is grateful to the Johan Gadolin scholarship 2013–14
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1573
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[
25] G. Blanco-Brieva, E. Cano-Serrano, J. M. Campos-Martin, J. L. G. Fierro,
Chem. Commun. 2004, 1184–1185.
[41] X. Huang, Z. Zhao, J. Fan, Y. Tan, N. Zheng, J. Am. Chem. Soc. 2011, 133,
4718–4721.
[
[
26] R. Wagner, P. M. Lange, Erdoel Kohle Erdgas 1989, 105, 414.
27] K. Weissermel, H. P. Arpe, Industrial Organic Chemistry, Vol. 11, 3rd ed.,
Wiley-VCH, Weinheim, 1997, p. 281.
28] M. Sankar, Q. He, M. Morad, J. Pritchard, S. J. Freakley, J. K. Edwards,
S. H. Taylor, D. J. Morgan, A. F. Carley, D. W. Knight, C. J. Kiely, G. J.
Hutchings, ACS Nano 2012, 6, 6600–6613.
[42] D. J. Lockwood, Nanotechnology in Catalysis (Eds.: B. Zhou, S. Hermans,
G. A. Somorjai), Kluwer Academic, New York, 2004.
[43] P. Centomo, M. Zecca, M. Kralik, D. Gasparovicova, K. Je rˇ àbek, P. Canton,
B. Corain, J. Mol. Catal. A 2009, 300, 48–58.
[44] M. Krµlik, M. Zecca, P. Bianchin, A. A. D’Archivio, L. Galantini, B. Corain, J.
Mol. Catal. A 1998, 130, 85–93.
[
[
[
[
[
[
29] E. N. Ntainjua, M. Piccinini, J. C. Pritchard, J. K. Edwards, A. F. Carley, J. A.
Moulijn, G. J. Hutchings, ChemSusChem 2009, 2, 575–580.
30] M. Kralik, M. Hronec, V. Jorik, S. Lora, G. Palma, M. Zecca, A. Biffis, B.
Corain, J. Mol. Catal. A 1995, 101, 143–152.
31] J. Aitchison, J. A. C. Brown, The LogNormal Distribution, Cambridge Uni-
versity Press, 1957.
[45] L. Ouyang, P.-f. Tian, G.-j. Da, X.-C. Xu, C. Ao, T.-y. Chen, R. Si, J. Xu, Y. F.
Han, J. Catal. 2015, 321, 70–80.
[46] F. Menegazzo, M. Signoretto, G. Frison, F. Pinna, G. Strukul, M. Manzoli,
F. Boccuzzi, J. Catal. 2012, 290, 143–150.
[47] S. Kim, D. W. Lee, K. Y. Lee, J. Mol. Catal. A 2014, 391, 48–54.
[48] S. Abate, S. Perathoner, G. Centi, Catal. Today 2012, 179, 170–177.
[49] S. Chinta, J. H. Lunsford, J. Catal. 2004, 225, 249–255.
[50] S. Abate, K. Barbera, G. Centi, G. Giorgianni, S. Perathoner, J. Energy
Chem. 2016, 25, 297–305.
[51] A. Biffis, H. Landes, K. Je rˇ µbek, B. Corain, J. Mol. Catal. A 2000, 151,
283–288.
[52] S. C. Kim, S. C. Jung, Y. K. H. G. Park, S. G. Ahn Seo, J. Nanosci. Nanotech-
nol. 2013, 13, 1966–1969.
32] B. Corain, K. Je rˇ àbek, P. Centomo, P. Canton, Angew. Chem. Int. Ed. 2004,
4
3, 959–962; Angew. Chem. 2004, 116, 977–980.
33] P. Centomo, P. Canton, D.; Canova, M.; Zecca, J. Nanosci. Nanotechnol.
013, 13, 6872–6879.
2
[
[
34] P. Centomo, M. Zecca, B. Corain, J. Cluster Sci. 2007, 18, 947–962.
35] P. Centomo, P. Canton, M. Ferroni, M. Zecca, New J. Chem. 2010, 34,
2
956–2961.
[
[
[
[
[
36] P. Centomo, K. Je rˇ àbek, D. Canova, A. Zoleo, A. L. Maniero, A. Sassi, P.
Canton, B. Corain, M. Zecca, Chem. Eur. J. 2012, 18, 6632–6643.
37] K. Je rˇ àbek, M. Zecca, P. Centomo, F. Marchionda, L. Peruzzo, P. Canton,
E. Negro, V. Di Noto, B. Corain, Chem. Eur. J. 2013, 19, 9381–9387.
38] R. F. Ziolo, E. P. Giannelis, B. A. Weinstein, M. P. O’Horo, B. N. Ganguly, V.
Mehrotra, M. W. Russel, D. R. Huffman, Science 1992, 257, 219–223.
39] E. Sulman, V. Matveeva, V. Doluda, L. Nikoshvili, A. Bykov, G. Demidenko,
L. Bronstein, Stud. Surf. Sci. Catal. 2010, 175, 153–160.
[53] S. Cheong, J. D. Watt, R. D. Tilley, Nanoscale 2010, 2, 2045–2053.
[54] N. Gemo, T. Salmi, P. Biasi, React. Chem. Eng. 2016, DOI: 10.1039/
C5RE00073D.
Received: January 8, 2016
40] H. Bçnnemann, K. S. Nagabhushana in Metal nanoclusters in catalysis
and materials science: the issue of size control (Eds.: B. Corain, G.
Schmid, N. Toshima), Elsevier, Amsterdam, 2008, p. 21.
Revised: February 19, 2016
Published online on April 9, 2016
ChemCatChem 2016, 8, 1564 – 1574
www.chemcatchem.org
1574
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim