Research Collection

Journal Article

Hemicellulose arabinogalactan hydrolytic hydrogenation overRu-modified H-USY zeolites

Author(s): Murzin, Dmitry Y.; Kusema, Bright; Murzina, Elena V.; Aho, Atte; Tokarev, Anton; Boymirzaev, Azamat S.;Wärnå, Johan; Dapsens, Pierre Y.; Mondelli, Cecilia; Pérez-Ramírez, Javier; Salmi, Tapio

Publication Date: 2015-10

Permanent Link: https://doi.org/10.3929/ethz-a-010792434

Originally published in: Journal of Catalysis 330, http://doi.org/10.1016/j.jcat.2015.06.022

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

1

Hemicellulose arabinogalactan hydrolytic hydrogenation over Ru-modified H-USY

zeolites

Dmitry Yu. Murzin1*, Bright Kusema2, Elena V. Murzina1, Atte Aho1,

Anton Tokarev1, Azamat S. Boymirzaev3, Johan Wärnå1,4, Pierre Y. Dapsens2, Cecilia

Mondelli2, Javier Pérez-Ramírez2, Tapio Salmi1

1Laboratory of Industrial Chemistry and Reaction Engineering, Process Chemistry

Centre, Department of Chemical Engineering, Åbo Akademi University, FI-20500

Åbo/Turku, Finland, E-mail: dmurzin@abo.fi

2Institute for Chemical and Bioengineering, Department of Chemistry and Applied

Biosciences, ETH Zurich, Vladimir-Prelog-Weg 1, CH-8093 Zurich, Switzerland

3Namangan Institute of Engineering and Technology, Department of Chemical

Technology, Namangan, 160115, Uzbekistan

4University of Umeå, Umeä, Sweden

ABSTRACT

The hydrolytic hydrogenation of hemicellulose arabinogalactan was investigated in the

presence of protonic and Ru (1-5 wt.%)-modified USY zeolites (Si/Al ratio = 15 and

30). The use of the purely acidic materials was effective in depolymerizing the

macromolecule into free sugars. While the latter partly dehydrated into 5-

hydroxymethylfurfural and furfural, the generation of high molecular-weight

compounds (aggregates of sugars and humins) was not favored, in contrast to previous

evidences over beta zeolites. Application of the bifunctional Ru/USY catalyst,

comprising well-dispersed metallic nanoparticles on the aluminosilicate support,

resulted in the formation of galactitol and arabitol, in the suppression of dehydration

side products, and further inhibition of polymerization reactions, which only yielded

low molecular-weight oligomers. Detailed analysis of the reaction pathways as well as

kinetic modelling of hydrolytic hydrogenation was performed with an advanced reaction

mechanism.

Keywords: Arabinogalactan, bifunctional catalysis, biomass upgrading, Ru/USY

zeolites, hydrolysis.

2

Introduction

In view of enabling a transition towards a biobased ecomomy, the depolymerization of

the main chemical components of lignocellulosic biomass [1,2], i.e., cellulose and

hemicellulose, to their sugars components by hydrolysis is an essential step which lies

at the very beginning of multiple value chains for the sustainable production of fuels

and chemicals. Due to the recalcitrant nature of the macromolecules, in particular the

crystalline cellulose, this process is conducted under harsh conditions. Therefore, the C6

and C5 monosaccharides formed can easily further convert into hydroxymethylfurfural

(HMF) or furfural and oligomerization reactions leading to various humins are favored.

A promising way of avoiding these undesired transformations of thermally-unstable

sugars is to hydrogenate them into the corresponding more robust polyols. this

methodology is an attractive means to boost the selectivity of the hydrolysis process and

generate valuable chemicals in one pot.

Balandin et al. [4] firstly demonstrated the hydrolytic hydrolysis of cellulose to sorbitol,

health care, food, and cosmetic additive, over a Ru/C catalyst in the presence of low-

concentrated mineral acids. A revival of the interest in this reaction came only after a

work of Fukuoka and Dhepe [5] who reported the use of an alumina-supported platinum

catalyst for the production of sorbitol and mannitol with 25% and 6% yields,

respectively. Later a range of carbon materials used as such or functionalized with

sulfonic groups [3, 6] and other solid acids including heteropolyacids, metal oxides, and

zeolites [7] have been evaluated. The latter class of catalysts has received particular

attention. Thus, the hydrolytic hydrogenation of unspecified cellulose [8],

microcrystalline cellulose in the presence of mineral acid traces [9] and birch pulp mill

cellulose with lower molecular weight and presence of xylan [10, 11] has been

performed with various metal-modified zeolites, including Ru-containing beta [8] and

USY [9].

3

In contrast to cellulose, which requires extensive pre-treatment (e.g. ball milling) to

reduce its crystallinity and thus enhance its reactivity, hemicelluloses, comprising

amorphous branched hetero-polymers with shorter chains than cellulose, have been

more readily solubilized and processed. Hemicelluloses comprise a number of

polysaccharides, e.g., xylan, glucuronoxylan, arabinoxylan, arabinogalactan,

galactoglucomannan, etc., which differ from each other in terms of nature and relative

amount of their building blocks (hexoses and pentoses) and degree of polymerization.

For instance, arabinogalactans (AG), appearing in large quantities in larch species such

as Larix sibirica, consists of -D-galactopyranose as a backbone with D-

galactopyranose and L-arabinofuranose side chains (Scheme 1). The average molar ratio

of galactose to arabinose in AG is about 6:1 and the molar mass is 20 000-100 000 g

mol−1 [12].

Scheme 1. Applications of the sugar polyols obtained by hydrolytic hydrogenation of

arabinogalactan.

4

The preparation of the free sugars arabinose and galactose has been first attempted

using sulfonic acid-functionalized polymers [13]. In comparison with such catalysis

better results in terms of catalyst activity and stability were obtained with zeolites in

their protonic form in the case of xylan hydrolysis [14]. In view of this result and due to

the superior performance of Ru-containing materials for the hydrogenation of arabinose

and galactose to their corresponding polyols (arabinitol and galactitol) [15-17], which

find application as low caloric, non-carcinogenic sweeteners, flavors, dietary

supplements, and pharmaceuticals, it was natural to apply bifunctional Ru-beta zeolites

catalyst in hydrolytic hydrogenation of arabinogalactan [18, 19]. While the side

reactions to HMF and furfural were minimized, only limited yields to the corresponding

polyols have been achieved over these bifunctional catalysts. Furthermore, substantial

formation of high molecular weight compounds, that is, aggregates of sugars and

humins (polydispersed heterogeneous carbonaceous materials), occurred. Mechanistic

aspects of humins formation have recently been addressed [20-25]. Their formation

mechanism has been related to polycondensation reactions giving a network of furan

rings linked by ether or acetal bonds [20]. At the same time infrared spectroscopy

studies [23] did not support the concept of acetal bonds formation advanced by

Sumerski et al. [20]. Patil et al. [21, 22] pointed out the involvement of a 2,5-dioxo-6-

hydroxyhexenal intermediate formed by rehydration of HMF in the condensation

process. In particular, it has been highlighted that this intermediate undergoes aldol

condensation with HMF leading to humins, which cannot be formed directly from

hexoses.

In the present work, hydrolysis and hydrolytic hydrogenation of AG were alternatively

studied over ultra-stable Y (USY) and Ru/USY zeolites, respectively. Materials

featuring distinct acidity and metal loading were compared and a range of methods was

applied to investigate the formation of humins both in terms of structural properties and

5

chemical nature of the precursor(s). Based on the experimental observations, an

advanced reaction network was proposed which was used as the basis for kinetic

modelling of the hydrolytic hydrogenation.

Experimental Section

Catalyst preparation

USY-15 (CBV720, molar Si/Al = 15, protonic form) and USY-30 (CBV760, molar

Si/Al = 30, protonic form) zeolites were purchased from Zeolyst International and used

as received. The incorporation of ruthenium (nominal loading 1-5 wt.%) into the

zeolites was performed by spray deposition using RuCl3·xH2O (Sigma-Aldrich, 99.9%)

as the metal precursor. This technique, pioneered by Yara International [26] enabled the

deposition of metal (oxide) particles onto a zeolite support with a high degree of

dispersion [27]. The preparation was carried out using a benchtop Büchi Mini Spray

Dryer B 290 equipped with a two-fluid nozzle (diameter = 1.5 mm) and a spray

chamber of 10 dm3. Prior to the synthesis, deionized water was passed through the

system for 45 min in order to reach thermal equilibrium. RuCl3·xH2O was dissolved in

deionized water (5 cm3) under magnetic stirring. Thereafter, the zeolite powder (1.0 g)

was added. The obtained slurry, continuously stirred, was pumped to the nozzle of the

spray dryer at a rate of 2 cm3 min−1 together with a constant spray air flow of 0.4 m3 h−1,

resulting in the formation of fine droplets of 5-20 µm size. The aspiration rate was set at

35 m3 h−1 and the inlet and outlet temperatures of the spray chamber were kept at

493 and 373 K, respectively. The hot drying gas flowed co-currently with the sprayed

slurry and its residence time in the spray chamber was 1 s. The dried particles were

separated and collected by a cyclone, further dried at 373 K for 12 h, and calcined in

static air at 623 K (heating rate = 3 K min−1) for 2 h. The obtained solids were reduced

in hydrogen at 573 K for 2 h prior to the catalytic experiments.

6

Catalyst characterization

The Ru content in the catalysts was determined by X-ray fluorescence spectroscopy

using an Orbis Micro-XRF analyzer (EDAX) operated with a Rh source at 30 kV. X-ray

photoelectron spectroscopy (XPS) was measured using a Perkin-Elmer PHI 5400

spectrometer with a Mg K X-ray source operated at 14 kV and 200 W. The pass

energy of the analyzer was set at 17.9 eV and the energy step at 0.05 eV. Nitrogen

sorption at 77 K was performed using a Quantachrome Quadrasorb-SI analyzer on

degassed samples (10−1 mbar, 573 K, 3 h). Powder X-ray diffraction (XRD) was

conducted using a PANalytical X’Pert PRO-MPD diffractometer. Data were recorded in

the 5-70° 2θ range with an angular step size of 0.05° and a counting time of 7 s per step.

High-resolution magic angle spinning 27Al nuclear magnetic resonance (MAS NMR)

spectroscopy was carried out using a Bruker AVANCE 700 NMR spectrometer

equipped with a 2.5-mm probe head and a 2.5-mm ZrO2 rotor at 182.4 MHz. Spectra

were acquired using a spinning speed of 20 kHz, 4096 accumulations, and a recycle

delay of 1 s. The acid properties of the catalysts were evaluated by Fourier transform

infrared spectroscopy (FTIR) using pyridine (Sigma-Aldrich, ≥99.5 %, a.r.) as the probe

molecule. The measurements were performed using an ATI Mattson spectrometer

equipped with an in situ cell containing ZnSe windows. The samples were pressed into

thin self-supporting wafers (20 mg, radius = 0.65 cm). Pyridine was first adsorbed for

30 min at 373 K and then desorbed by evacuation for 20 min at different temperatures

(523, 623, and 723 K). Spectra were recorded by co-addition of 32 scans with a

resolution of 2 cm−1. The Brønsted-acid sites (BAS) and Lewis-acid sites (LAS) were

quantified based on the intensities of the bands at 1545 and 1450 cm−1, respectively, and

using the molar extinction coefficients reported in the literature [28]. Transmission

electron microscopy (TEM) imaging was undertaken with a FEI Tecnai F30 microscope

7

operated at 300 kV (field emission gun). The samples were prepared by depositing a

few droplets of zeolites suspension in methanol onto a carbon-coated copper grid,

followed by evaporation at room temperature.

Catalytic evaluation

Hydrolysis in the presence of hydrogen over protonic zeolites and hydrolytic

hydrogenation experiments over metal-modified zeolites were carried out in a 300-cm3

Parr autoclave reactor connected to a 200-cm3 pre-reactor. The autoclave was equipped

with sampling outlet featuring a 1-m filter to prevent even very fine catalyst particles

from escaping it. The temperature was measured with a thermocouple and controlled

automatically (Brooks Instrument). At the beginning of each experiment, 400 mg of AG

were dissolved in 90 cm3 of deionized water and loaded into the pre-reactor. Thereafter,

200 mg of catalyst with a particle size below 63 m (to avoid internal diffusion

limitations) were loaded into the reactor containing 10 cm3 of deionized water. The

reactor was pressurized with hydrogen and heated to 458 K, reaching a total pressure of

31 bar. Based on the vapor pressure of the solvent at this temperature, the partial

pressure of hydrogen was 20 bar. The stirring (1000 rpm, to minimize external mass

transport limitations) was then started and the reactant solution was fed from the pre-

reactor into the reactor. This was considered as the initial reaction time. Liquid samples

from the reaction mixture were periodically withdrawn for analysis. The decrease in

volume of the reaction mixture was taken into account in the calculations of reactant

and products concentrations.

Product analysis

8

The liquid samples from the reaction mixture were quantitatively analyzed without any

pretreatment in a high-performance liquid chromatography (HPLC) system using two

different columns and a refractive index (RI) detector. A Bio-Rad Aminex HPX-87C

column heated at 353 K was used to analyze AG, sugars, sugar alcohols, and furan

compounds. A diluted (1.2 mM) CaSO4 solution flowing at 0.4 cm3 min−1 was

employed as the mobile phase. An Aminex cation H+ column heated at 338 K was used

to analyze acidic compounds and other degradation products. In this case, the mobile

phase comprised a 0.005 M H2SO4 solution flowing at 0.5 cm3 min−1.The individual

components were identified by gas chromatography-mass spectrometry (GC-MS) as

discussed in detail in [19]. The carbon mass balance was calculated considering the

concentration of AG, sugars, polyols, furfurals, and low molecular-weight compounds

analyzed by HPLC.

The weight average molar mass (Mw) and number average molar mass (Mn) were

determined via gel permeation (size exclusion) chromatography (SEC) using a system

equipped with two columns in series (300 7.8 mm Ultrahydrogel liner, Waters,

Milford, USA), a multi-angle laser light scattering MALLS unit (miniDAWN, Wyatt

Technology, USA), and RI and UV detectors. A 0.1 M NaNO3 aqueous solution

flowing at 0.5 cm3 min−1 served as the eluent. The samples were filtered with a 0.45-m

syringe Acrodisc filter. The injection volume was 100 μl. The Astra software (Wyatt

Technology) was used for data analysis.

Results and Discussion

USY and Ru/USY zeolites

The FAU-type zeolites used in this study feature a bulk Si/Al ratio of 15 (USY-15) and

30 (USY-30). According to the NMR spectroscopic analysis (not shown), both zeolites

9

contain a significant amount of distorted tetrahedral Al species as well as

extraframework penta- and, especially, hexacoordinated Al centers, which are generated

upon stabilization of the pristine Y zeolite via steaming and acid washing. Both samples

feature a mesoporous surface area (Smeso) of 125-128 m2g−1 and a micropore volume

(Vmicro) of 0.29-0.31 cm3 g−1, as typically observed for these materials (Table 1). Their

acidity was evaluated via FTIR studies using pyridine as a probe molecule. Upon

adsorption and desorption of pyridine at different temperatures, the Brønsted-acid sites

(BAS) and Lewis-acid sites (LAS) could be quantified and classified based on their

relative strength. Thus, USY-15 and USY-30 possess a similar concentration of BAS

and LAS (335 and 65 μmol g−1 and 310 and 51 μmol g−1, respectively), but the former

catalyst contains a slightly higher number of medium and strong sites of either type

(Table 1).

Table 1. Characterization data of the protonic and Ru-modified zeolites.

Catalysts BASa

(μmol g−1)

LASa

(μmol g−1)

Smesob

(m2 g−1)

Vmicrob

(m3 g−1)

Ru

contentc

(wt.%)

Cryst.d

(%)

523 K 623 K 723 K 523 K 623 K 723 K

USY-15 163 126 46 41 17 7 128 0.29 - 100

Ru(1)/USY-15 233 11 2 32 3 1 117 0.29 1.4 -

Ru(2.5)/USY-15 231 7 4 42 3 1 115 0.28 2.3 82

Ru(5)/USY-15 197 0 0 35 0 0 104 0.27 4.8 -

USY-30 152 133 25 45 5 1 125 0.31 - 100

Ru(2.5)/USY-30 207 7 0 39 5 0 118 0.31 2.0 78 aDetermined by FTIR of adsorbed pyridine. bDetermined by the t-plot method. cDetermined by XRF. dDerived from XRD.

Ruthenium was deposited onto both zeolites to attain bifunctional catalysts in an amount

of 1.4, 2.3, and 4.8 wt.% for USY-15 and 2.0 wt.% for USY-30. The mesoporous

surface area and microporous volume of the aluminosilicates were substantially retained

upon metal incorporation (Table 1), likely due to the rather low loading, while XRD

analysis (not shown) indicated a slight decrease in crystallinity (Table 1). The absence

10

of reflections specific to Ru in the patterns suggested the presence of a well-dispersed,

nanostructured metal phase. The structural features of the catalysts were further

investigated by TEM (Figure 1). As expected, both protonic zeolites comprised crystals

featuring intraparticle mesoporosity. With respect to the Ru-modified samples, the

crystallinity of the zeolites appeared hardly modified and the supported metal phase was

detected in form of nanostructures with a broad size distribution. Small nanoparticles

(3-4 nm) were visualized along with much larger structures (50-60 nm). Only few of the

latter were found in Ru(1)/USY-15, whereas they were more abundant in the catalysts

with higher metal loadings. Based on these observations, the moderate decrease in

crystallinity upon metal deposition determined by XRD seems to be mainly related to

the presence of the secondary metal phase in addition to the aluminosilicate.

Figure 1a. TEM micrographs of protonic (P) and Ru-modified USY-15 zeolites.

11

Figure 1b. TEM micrographs of protonic (P) and Ru-modified USY-30 zeolites.

The total acidity of the Ru-containing samples was substantially modified compared to

the metal-free counterparts. In particular, the BAS of medium and high strength almost

vanished, while the amount of weak BAS moderately increased (Table 1). This evidence

is in line with previous studies for various zeolite supported metal catalysts, which

pointed that introduction of metals onto the zeolite support results in redistribution of

acid sites strength [33-35]. The origin of such changes was discussed in detail in the

previous work and can be attributed to interactions between the metal crystallites and

the support material as well as changes in the support properties during catalyst

preparation due to exposure of the zeolite to the metal precursor solution.

Interestingly, the change in acid properties was rather comparable regardless of the

ruthenium loading. The absence of strong acid sites is expected to suppress side

reactions such as sugar dehydration to HMF and furfural, thus favoring the selectivity to

polyols.

XPS analysis of the reduced catalysts indicated that the maximum of the Ru3d5/2 peak

is shifted to binding energy less than 280 eV when the charging (ca. 2.7 eV) is taken

into account meaning that oxidation state of ruthenium is close to 0. The peak

assignment was based on the values reported by Pedersen and Lunsford [36].

12

Non-catalytic hydrolysis and hydrolytic hydrogenation

The key role of acidity in hydrolysis is in donating a proton to the glycosidic bond

between the sugar units in the polysaccharide chain, enabling the liberation of the

monosaccharides [19]. Disadvantageously, acid species additionally catalyze sugar

dehydration to furfural and HMF under the hydrolysis conditions. This comprises a

competitive reaction to the further desired conversion of the sugars to polyols.

Acid solids are expected to influence the hydrolysis of AG in similar manner to how

homogeneous mineral acids drive the hydrolysis of (hemi)cellulose provided that the

molecules involved in the process can have access to the acid sites. It can be speculated

that the external surface acid sites will exclusively provide the Brønsted acidity required

for the hydrolysis of the macromolecule, since the latter cannot penetrate inside the

pores, but that the cleavage of short-chain depolymerization products could also occur

on acid sites situated in the channels.

However, as the pKw value of water decreases with increasing temperature, H3O+ ions

generated in situ can also homogenously contribute to the overall hydrolysis process and

to sugar dehydration.

Thus, prior to the utilization of parent and metal-modified zeolites, a non-catalytic

experiment was conducted for the title reaction at 458 K and a total pressure of 31 bar

(Figure 2) to serve as a reference.

13

0 50 100 150 200 250

0

10

20

30

40

50

60

70

80

90

100 AG nd humins

Oligomers

Galactose

Arabinose

Ethylene glycol

HMF

Furfural

Unknown

pro

du

ct d

istr

ibu

tio

n, %

time, min

Figure 2. Non-catalytic hydrolysis of AG at 458 K and 31 bar.

Evidently, even in the absence of any acid catalyst, a significant decrease in the

concentration of hemicellulose, visible formation of oligomers, arabinose, and galactose

monosaccharides as well as furans were detected.

The peculiar shape of the AG-humins curve derives from the fact that upon the course

of the reaction polymeric humins were formed while AG was consumed and the two

classes of species cannot be separated by HPLC analysis. Polymeric humins derive from

acid-catalyzed transformations of HMF [20-23] and aggregation of galactose when

arabinose is removed from the side chain of the hemicellulose [19].

It was thus interesting to analyze the molecular mass of the polymers by SEC. Since

SEC of water-soluble polymers is less straightforward than analysis of polymers in

organic media, due to interactions between the stationary phase and polar carbohydrates

[29], electrolyte solutions of sufficiently high ionic strength were used as the mobile

phase to prevent such secondary effects.

In polymer analysis, besides for the peak apex molecular weight Mp characterizing the

sample only in a single point, relevant parameters comprise the number

14

average molecular mass of polymers Mn (more sensitive to molecules of low molecular

mass) and the molecular weight average Mw (more sensitive to molecules of high

molecular mass), which are determined in the following way:

)(

)(

Mh

MMhM n

,

MMh

MMhMW

)(

)( 2

(1)

where h(M) is the slice height at a molecular weight M when the eluted peak is divided

into several equidistant volume slices. Another significant parameter is dispersity

(Mw/Mn), which gives an indication about the distribution in the polymer, approaching

unity when the polymer chain approaches a uniform chain.

The values of the molecular weight average Mw are displayed for the non-catalytic

hydrolysis experiment in Figure 2.

0 50 100 150 200 250

0

5

10

15

20

25

30

35

40

Mw

, kD

a

time, min

a)

Figure 3. (a) Molecular weight of polymers/oligomers in the reaction mixture as a

function of time. Experimental conditions for this and subsequent figures are the same

as for Figure 2.

15

0 50 100 150 200 250

0

2

4

6D

isp

ers

ity,

[-]

time, min

Figure 3. (b) Dispersity (Mw/Mn) of polymers/oligomers in the reaction mixture as a

function of time. Experimental conditions for this and subsequent figures are the same

as for Figure 2.

These data evidence a very clear disaggregation of AG with a relatively high molecular

mass (ca. 40 kDa) and a dispersity index of ca. 5 in the first 50-75 min of reaction

leading to a polymer of ca. 2.5 kDa molecular weight and ca. 1.5 dispersity. Thereafter,

there is a clear increase of the molecular weight and of the dispersity of the resulting

polymer in excellent agreement with HPLC data which display a sudden increase in the

concentration of the macromolecule and a decrease in that of the oligomers. The UV-vis

detector confirmed the presence of aromatic moieties in the newly formed compound,

which can thus be ascribed to a humin-type polymer. As already discussed in [30],

hydrolysis of AG is, moreover, associated with the removal of arabinose from the side

chain, relaxing the steric hindrance in the hemicellulose and allowing galactose

molecules to oligomerize resulting in aggregates. This process would inevitably lead to

an increase of dispersity. It should be noted that differentiation and quantification of

humins and other potential aggregates was not in the main focus of the work, therefore

it was not pursued further.

16

Hydrolysis over acidic zeolites

The results for hydrolysis of AG carried out over USY-15 and USY-30 are presented in

Figure 4 and demonstrate that, similarly to a non-catalytic experiment (Figure 4), AG

was hydrolyzed into oligomers and monomers.

a)

0 50 100 150 200 250

0

20

40

60

80

100

AG and humins

Oligomers

Galactose

Arabinose

Ethylene glycol

HMF

Furfural

Unknown

Pro

du

ct d

istr

ibu

tio

n, %

time, min

Figure 4a. Hydrolysis of AG over USY-15.

b)

0 50 100 150 200 250

0

20

40

60

80

100

AG and humins

Oligomers

Galactose

Arabinose

Ethylene glycol

HMF

Furfural

Unknown

Pro

du

ct d

istr

ibu

tio

n, %

time, min

Figure 4b. Hydrolysis of AG over USY-30.

17

Still, a higher conversion level was attained in the presence of both zeolites compared to

the blank run. Thus, application of solid acid catalysts is necessary to enhance the

efficiency of the hydrolysis.

An interesting observation besides an increase in the rates was the substantial

suppression of humins formation, since the peak of AG constantly decreased with time.

Such evidence with USY catalysts differs not only from the blank experiment but also

from the previous data on beta zeolites [19], over which humins were generated to a

large extent. At the moment it can be only speculated that the differences in the

behavior of beta compared to USY could be related to lower Lewis acidity and

moreover additional mesoporosity of the latter allowing easier release of sugars and

dehydration products thereby avoiding extensive polymerization.

At the same time, there was a clear formation of galactose aggregates (oligomers) along

with the generation of HMF and the unknown compound. The relative concentration of

side products, HMF and furfural, was not influenced by the type of acidic catalyst

(Figure 5a) and was higher than in the blank experiment, which can be explained by the

lower conversion in the latter case.

The weight ratio between galactose and arabinose was close to two (Figure 5b) in

catalytic and non-catalytic experiments. In contrast to HCl-catalyzed hydrolysis, in

which the release of sugars follows stoichiometry (6) [13], the substoichiometric

formation of galactose in our experiments indicates a preferential cleavage of the side

chain. This behavior was already observed upon AG hydrolytic hydrogenation over beta

zeolites [19] (ratio close to 1.6) and for AG hydrolysis over a heterogeneous catalyst of

an ion-exchange type bearing sulfonic groups [13].

18

a)

0 50 100 150 200 250

0

2

4

6

8

10

12

HMF - USY-30

HMF-no catalyst

Furfural-USY-30

Furfural-no catalyst

Furfural-USY-15

HMF - USY-15R

ela

tive

co

nce

ntr

atio

n, %

time, min

Figure 5. (a) Concentration of HMF as a function of time over the USY catalysts and in

the blank experiment.

b)

0 2 4 6 8 10 12 14

0

2

4

6

8

10

12

14

16

USY-15

USY-30

blank

Ara

bo

no

se

, %

Galactose, %

Figure 5. (b) Concentrations of arabinose and galactose attained over the USY catalysts

and in the blank experiment.

Hydrolytic hydrogenation over Ru-modified zeolites

The bifunctional approach of converting carbohydrate polymers through combining two

functions (redox and acidic) in the same catalyst has been practiced by several groups [3,

5, 37-39] as indicated in the Introduction. Ref. [37] postulates for the case of carbon

19

nanofibers supported nickel catalysts that a proper balance of acid and redox site on the

material is essential in order to obtain very high polyol yields starting from cellulose.

In the current work the influence of combining sugar formation with their consecutive

hydrogenation was investigated using the bifunctional catalysts, i.e., the Ru-modified

USY zeolites. The catalytic results for Ru(2.5)/USY-15 and Ru(2.5)/USY-30 are

displayed in Figure 6. A recycling experiment was performed with Ru(2.5)/USY-30

giving exactly the same concentration profile (not shown) as depicted in Figure 6c.

It should be noted, however, that one- or two recycling experiments might not be

sufficiently good basis for assessing the catalyst stability. For example, zeolites Y and

ZSM-5 were treated in liquid water at 150 and 200oC resulting in transformations of the

zeolite Y [40]. At the same time a recent work related to stability of commercial USY

zeolites in liquid water and utilization of Ru/Y materials in hydrolytic hydrogenation of

cellulose showed that these zeolites can be in principle stable in hot liquid water at least

on the short term [41]. Due to the possible amorphization of the zeolites in the long

term, catalytic behaviour and stability of the catalytic systems studied in the current

work should be further accessed in the future by performing multiple recycling

experiments or by operation in continuous mode. This was, however, outside of the

scope of the present work.

20

a)

0 50 100 150 200 250

0

20

40

60

80

100

AG and humins

Oligomers

Galactose

Arabinose

Ethylene glycol

Arabitol

Galactitol

HMF

Furfural

Unknown

Pro

du

ct d

istr

ibu

tio

n, %

time, min

Figure 6a. Concentration of all products as a function of time in the hydrolytic

hydrogenation of AG over Ru(2.5)/USY-15.

b)

0 50 100 150 200 250

0

5

10

15

20

25

Galactose

Arabinose

Ethylene glycol

HMF

Furfural

Arabitol

Galactitol

Unknown

Pro

du

ct d

istr

ibu

tio

n, %

time, min

Figure 6b. Concentration of selected products as a function of time in the hydrolytic

hydrogenation of AG over Ru(2.5)/USY-15.

21

c)

0 50 100 150 200 250

0

20

40

60

80

100

AG and humins

Oligomers

Galactose

Arabinose

Ethylene glycol

Arabitol

Galactitol

HMF

Furfural

Unknown

Pro

du

ct co

nce

ntr

atio

n, %

time, min

Figure 6c. Concentration of all products as a function of time in the hydrolytic

hydrogenation of AG over Ru(2.5)/USY-30.

d)

0 50 100 150 200 250 300

0

5

10

15

20

AG and humins

Galactose

Arabinose

Ethylene glycol

Arabitol

Galactitol

HMF

Furfural

Unknown

Pro

du

ct co

nce

ntr

atio

n, %

time, min

Figure 6d. Concentration of selected products as a function of time in the hydrolytic

hydrogenation of AG over Ru(2.5)/USY-15.

During the hydrolytic hydrogenation of AG over these catalysts, in addition to the main

products detected also for the metal free pure zeolite, the formation of sugar alcohols

was visible, confirming the hydrogenation ability of the catalyst. Comparison of the

22

rates for the parent and metal-modified USY-15 and USY-30 revealed that the initial

AG conversion is practically identical. A somewhat similar behavior has been reported

in [31] when the addition of a supported iridium catalyst to the reaction mixture already

containing zeolite ZSM-5 did not influence the conversion of cellulose. Very similar

behavior in terms of product distribution was observed for ruthenium supported on both

USY-15 and USY-30, which can be explained by similar acidity of the catalysts.

In contrast, the presence of Ru in the USY zeolite resulted in the suppression of the

dehydration reaction in analogy to previous results with Ru/beta catalysts [19].

The ratio between galactose and arabinose was initially close to unity but substantially

decreased with an increasing conversion. This, along with a clear maximum for the

galactose concentration, points to a much faster transformations of the latter compared

to arabinose. It is also interesting to note that, contrary to a non-catalytic experiment or

hydrolysis in the presence of zeolites, the formation of HMF over Ru(2.5)/USY-15 was

less prominent than the generation of furfural at the end of the experiment.

Hydrogenation of glucose and arabinose on Ru in their mixtures typically proceeds with

a similar rate [16] at least under milder conditions. Several scenarios were considered to

explain the concentration profiles. One option is to assume that the more pronounced

consumption of galactose and HMF is related to a reaction between them, which

happens even in the absence of Ru but is certainly promoted by the metal. Another

possibility is that an isomerization of an aldose (galactose and arabinose) to a ketose

(tagatose and ribulose) may occur with different rates and the unknown compound is an

isomer of the sugars. Such reaction was recently shown to take place over Lewis-acid

catalysts such as tin-promoted beta zeolites [32]. This second hypothesis was discarded

by evaluating the retention times of ketoses, which elute at much longer times than the

unknown compounds.

23

The alternative case of generating di- and tri-saccharides from galactose should be ruled

out since the retention time of such oligomers is lower than for sugars, contrary to the

observed unknown compound, which elute after sugars and prior to sugar alcohols.

Along the same line, the formation of acids such as arabinoic acid or oxidized forms of

galactose (either acids or lactones) should be excluded since such oxidized forms of

sugars have different retention times from the one for the unknown species. In addition,

other types of transformations such as Cannizaro or cross-Canizzaro reactions of

aldehydes (galactose and arabinose) to form the corresponding acids and alcohols or

dehydration to anhydrosugars are irrelevant since they typically occur in basic media.

In order to address the issue of potential side reactions identification of the unknown

compound was done by the so-called spiking technique, when the retention time of a

range of compounds was determined separately and those compounds were added to the

reaction samples. Among tested compounds several mono- and dissacharides as well as

mentioned above sugar alcohols and acids were applied (fructose, mannose, mannitol,

maltose, saccharose, lactose, arabinonic and galacturonic acids), confirming their

absence in the reaction mixtures.

Another explanation for the observed kinetic behavior of galactose can be related to its

cleavage. Since in the literature is it often reported [3] that hydrolytic hydrogenation can

result in cleavage of carbon-carbon bond, retention times for several polyols with a

shorter carbon chain were measured, confirming formation of ethylene glycol, observed

also in the literature [42, 43] for transformations of cellulose. Interestingly enough no

formation of other alcohols and polyols mentioned in the literature [42, 43] such as

methanol, glycerol, 1,2-propanediol, meso-erythritol could be confirmed. Moreover

there also was no formation of levulinic acid.

In order to address the issue of the hydrogenolysis products with carbon-carbon bond

breaking, transformations of a mixture of arabinose and galactose in the presence of

24

hydrogen was studied where the initial ratio between the sugars was the same as

obtained during the initial stages of AG hydrolysis (Figure 7).

a)

0 50 100 150 200 250

0

10

20

30

40

50

Oligomers, Humins

Galactose

Arabinose

Ethylene glycol

Unknown

Arabitol

Galactitol

HMF

Furfural

Pro

du

ct d

istr

ibu

tio

n, %

time, min

Figure 7a. Concentration of products as a function of time for hydrogenation of a

galactose and arabinose mixture (50 wt.%/50 wt.%) on Ru(2.5)/USY-15.

b)

0 20 40 60 80 100 120 140 160 180

0

10

20

30

40

50 Oligomers

Galactose

Arabinose

Ethylene glycol

Unknown

HMF

Furfural

Pro

du

ct com

po

sitio

n, %

time, min

Figure 7b. Concentration of products as a function of time for hydrogenation of a

galactose and arabinose mixture (50 wt.%/50 wt.%) in a non-catalytic experiment.

As indicated by the profiles in Figure 7a, the hydrogenation rate of both sugars was

similar and after 25 min the reaction was practically complete. Mainly arabitol and

galactitol were present in the reaction mixture, besides traces of HMF and furfural and

25

heavier products, which eluted at the same time as oligomers. These latter compounds

could be either aggregates of galactose or some sort of humins, formed from HMF and

furfural. Another observation was that, besides these heavier compounds, the product

where the carbon-carbon bond is broken (ethylene glycol) was observed in larger

quantities than the unknown compound. The latter in fact could be could be a precursor

of oligomers/humins.

Stability of sugar alcohols was further confirmed by a blank test, where the same

mixture of sugars was treated at the same temperature without any catalyst (Figure 7b).

Obviously, no sugar hydrogenation reaction occurred due to the lack of redox metals in

the system. Still, although at a lower rate, monosaccharides were transformed into

ethylene glycol, an unknown compound, oligomers, as well as HMF and furfural. In fact,

generation of the latter components was the most prominent. The product mixture

became yellowish, which is an indication of humins formation. Note that in a catalytic

experiment the resulting mixture was transparent.

The effect of the ruthenium loading on the hydrolytic hydrogenation of AG was studied

testing the different Ru-modified USY-15 catalysts. As shown in Figure 8a, the rate of

AG hydrolysis and the concentration profiles for hemicellulose and oligomers over the

various catalysts are essentially the same regardless of the metal content. This can be

related to similar acidic properties of all Ru-modified zeolites. The same is valid for the

concentration of arabinose (Figure 8b), while a very clear decrease in the concentration

of galactose was observed for higher metal loadings (8c). Release profiles are in line

with the preferential cleavage of arabinose.

A parabola like behavior of galactose could be a result of its more facile transformation

to ethylene glycol on one hand and on another that galactose might react further forming

an unknown compound, which can be a precursor for humins.

26

Formation of sugars was enhanced with a higher metal loading in the case of arabitol

(8d), while for galactitol formation such an increase was not evident (8e). Similar

formation rates of sugar alcohols could be due to the fact that a higher metal loading

does not correlate to a much higher metal surface because of a larger fraction of big

particles. In addition, it can be also argued that the much more pronounced

disappearance of galactose is related not only to hydrogenation of the latter but also to

other competitive transformations.

a)

0 50 100 150 200 250

0

20

40

60

80

100

Oligomers

1.0 % Ru

2.5% Ru

5.0% Ru

Pro

du

ct d

istr

ibu

tio

n, %

time, min

AG

Figure 8a. Concentration of AG and oligomers as a function of time for the hydrolytic

hydrogenation of AG over Ru/USY-15 catalysts.

27

b)

0 50 100 150 200 250

0

4

8

12

16

20

Arabinose

1% Ru

2.5 % Ru

5 % Ru P

rod

uct d

istr

ibu

tio

n, %

time, min

Figure 8b. Concentration of arabinose as a function of time for the hydrolytic

hydrogenation of AG over Ru/USY-15 catalysts.

c)

0 50 100 150 200 250

0

4

8

12

16

20Galactose

1% Ru

2.5 % Ru

5 % Ru

Pro

du

ct d

istr

ibu

tio

n, %

time, min

Figure 8c. Concentration of galactose as a function of time for the hydrolytic

hydrogenation of AG over Ru/USY-15 catalysts.

28

d)

0 50 100 150 200 250

0

1

2

3

4

5

6

7

8

1% Ru

2.5% Ru

5 % Ru

Ara

bito

l , %

time, min

Figure 8d. Concentration of arabitol as a function of time for the hydrolytic

hydrogenation of AG over Ru/USY-15 catalysts.

e)

0 50 100 150 200 250

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

1% Ru

2.5% Ru

5 % Ru

Ga

lactito

l, %

time, min

Figure 8e. Concentration of galactitol a function of time for the hydrolytic

hydrogenation of AG over Ru/USY-15 catalysts.

It was thus instructive to verify the concentration profiles for HMF, furfural, ethylene

glycol and the unknown compound (Figure 9). The concentration of HMF was the

highest in the case of Ru(1)/USY-15, while it was very low for higher ruthenium

29

contents (Figure 10a). This is a result of not only catalyst acidity per se and could be

related to presence of competing pathways in the reaction system involving also

metallic sites. Interestingly, the concentration of furfural was the highest for Ru(5)-

USY-15 even if this catalyst does not have strong and medium BAS and overall has the

lowest amount of BAS. The trends in generation of ethylene glycol and the unknown

compound as a function of the metal loading, were less clear (Figure 9 b, c).

a)

0 50 100 150 200 250

0,0

0,8

1,6

2,4

3,2

4,0

4,8

5,6

6,4 1% Ru HMF

1% Ru furfural

2.5 % Ru HMF

2.5 % Ru furfural

5% Ru HMF

5% Ru furfural

Co

nce

ntr

atio

n, %

Time, min

Figure 9a. Concentration of all HMF and furfural as a function of time in the hydrolytic

hydrogenation of AG over Ru/USY-15 catalysts.

b)

0 50 100 150 200 250

0

4

8

12

16

20

24

28

1% Ru

2.5% Ru

5 % Ru

Eth

yle

ne

gly

co

l

time, min

30

Figure 9b. Concentration of ethylene glycol as a function of time in the hydrolytic

hydrogenation of AG over Ru/USY-15 catalysts.

c)

0 50 100 150 200 250

0

2

4

6

8

10

1% Ru

2.5% Ru

5 % Ru

Co

ncen

tra

tio

n o

f u

nkno

wn

, %

time, min

Figure 9c. Concentration of the unknown product as a function of time in the hydrolytic

hydrogenation of AG over Ru/USY-15 catalysts.

To elucidate a role of solid acids in transformations of galactose and arabinose,

experiments were performed with these sugars over USY-15. The results in Figure 10

demonstrated the formation of some oligomers (or low molecular-mass humins) of type

II, which eluted at the same time as oligomers formed by hydrolysis of AG, denoted as

oligomers of type I. HMF was also formed along with some amounts of hydrogenolysis

product (ethylene glycol) and the unknown compound in the case of galactose, while for

arabinose only furfural was produced in addition to oligomers and lower amounts of

ethylene glycol than for the case of galactose. Generation of oligomers was more

prominent than in the case of a galactose - arabinose mixture without any catalyst. An

unusual behavior of galactose formation after 2 h of reaction can be related to its back-

formation from aggregates.

31

a)

0 50 100 150 200 250

0

20

40

60

80

100 Oligomers

Galactose

Ethylene glycol

Unknown

HMFP

rod

uct co

mp

ositio

n, %

time, min

Figure 10a. Transformation of galactose over USY-15.

b)

0 50 100 150 200 250

0

10

20

30

40

50

60

70

80

90

100

Oligomers

Arabinose

Ethylene glycol

Unknown

Furfural

Co

nce

ntr

atio

n, %

time, min

Figure 10b. Transformation of arabinose over USY-15.

Additional experiments to elucidate the potential reaction network leading to unknown

compounds were performed with USY-15. Transformation of arabinose in a mixture

with furfural (Figure 11a) clearly that furfural is generated from arabinose, which can be

seen as an increase of the furfural concentration during the reaction in comparison to the

initial concentration. Moreover formation of oligomers, ethylene glycol and the

unknown compound, the same as in hydrolytic hydrogenation, was observed. The

32

concentration of the sugar substantially declined with time. A similar profile was

noticed for a mixture of galactose and HMF with the main difference that the formation

of the ‘Unknown’ compound was not noticed.

a)

0 50 100 150 200 250

0

10

20

30

40

50

60

70

80

90

100

110

120

Co

nce

ntr

atio

n, %

time, min

Oligomers

Arabinose

Unknown

Ethylene glycol

Unknown 2

Furfural

Figure 11a. Transformation of arabinose-furfural mixture over USY-15.

b)

0 50 100 150 200 250

0

10

20

30

40

50

60

70

80

90

100

Oligomers

Galactose

Ethylene glycol

HMF

Pro

du

ct co

nce

ntr

atio

n, %

time, min

Figure 11b. Transformation of a galactose-HMF mixture over USY-15.

It is apparent from the results presented above that hexoses and pentoses (e.g. galactose

and arabinose) degrade under acidic conditions to HMF and furfural forming also some

33

intermediates along this pathway and leading to the generation of some condensation

products of different molecular weight. Several recent reports address mechanistic and

kinetic aspects of hexoses and pentoses degradation [44-46]. One possibility discussed

in the literature is that the unknown products are in fact dimers of HMF and furfural,

although HMF is not a candidate for self-aldol condensation as it lacks an -hydrogen

atom.

It is clear that the pathways for degradation of hexoses and pentoses are still under

debate and more work is needed to establish the kinetic path for the formation of

furfural and HMF as well as for the generation of the unknown compound. The latter

can be tentatively assumed for the kinetic modelling described in the subsequent section

as an intermediate on the path from sugars to their degradation products.

Kinetic modeling

In [19] a reaction network for hydrolytic hydrogenation of AG was proposed which

besides hydrolysis and hydrogenation included formation of humins. In light of the

information about the reaction pathways obtained in the current work, the scheme

should be modified (Scheme 2) to include generation of ethylene glycol and the

unknown compound from galactose and disappearance of HMF due to formation of

minor amounts of humins.

34

Scheme 2. Reaction network for the hydrolysis of AG.

Initial modelling revealed that the reaction network can be somewhat simplified and

reactions without numbers in Scheme 2 (transformations of arabinose and galactose to

aggregates) can be omitted. Moreover reactions 9 and 12 in Scheme 2 were considered

irreversible. A set of differential equations considering relevant stoichiometry and for

simplicity first-order reactions were thus written based on the reaction network in

Scheme 2.

Numerical data fitting was done for the experimental data generated with Ru(5)/ USY-

15. The backward difference method was used for minimization of the sum of residual

squares (SRS) with non-linear regression analysis using the Simplex and Levenberg-

Marquardt optimization algorithms implemented in the parameter estimation software

ModEst [47]. The sum of squares was minimized starting with Simplex and thereafter

switching to the Levenberg-Marquardt method.

35

Estimated parameters and the standard errors are presented in Table 2, while the results

are displayed in Figure 12. In general, a good correspondence between the experimental

and calculated data was found with a degree of explanation 99.26%. The majority of

parameters were rather well identified and the large error related to few of them is

understandable taking into account the relatively limited set of experimental data and

the fact that some products can be formed through different routes, preventing more

reliable identification of the rate constants.

0 50 100 150 200 250

0,00

0,05

0,10

0,15

0,20

0,25

AG and humins

Oligomers

Galactose

Arabinose

Ethylene glycol

Arabitol

Galactitol

HMF

Furfulal

Unknown

co

nce

ntr

atio

n, w

t%

time, min

Figure 12. Comparison between experimental data for Ru(5)/USY-15 and calculations.

Table 2. Values for the parameters. Kinetic constants correspond to reactions in

Scheme 2.

Rate coefficient,

min-1

Value Error, %

k1 1.53 8.5

k2 0.84 25.3

k3 0.41 >100

k4 1.66 >100

k5 0.52 96

36

k6 0.68 12

k7 0.14 >100

k8 0.45 >100

k9 5.93 >100

k10 0.55 14.8

k11 1.38 7.7

k12 0.61 >100

k13 1.8 6.7

k14 2.0 31.5

Conclusions

The one-pot hydrolytic hydrogenation of arabinogalactan was herein investigated over

bifunctional Ru-modified USY zeolites. Parameters including the Ru content (1 to

5 wt.%) and the acidity of the support were varied in order to maximize the yield of

arabinitol and galactitol. Remarkably, yields up to 23% to the desired alditols and a

limited amount of high molecular-weight compounds (aggregates of sugars and humins)

were attained. Accordingly, Ru/USY zeolites stand as more effective catalyst than

Ru/beta materials previously reported for this application. Monitoring the product

distribution obtained from different known reaction intermediates, insights into the

hydrolytic hydrogenation mechanism were gathered in terms of the nature of the

reactions determining the formation of aggregates, humins, and oligomers. Based on the

advanced reaction network derived, a kinetic model closely fitting the experimental data

was successfully developed.

37

AUTHOR CONTRIBUTIONS

The manuscript was written through contributions of all authors. All authors have given

approval to the final version of the manuscript.

ACKNOWLEDGMENTS

This work is part of the activities at the Åbo Akademi University Process Chemistry

Centre. The Swiss National Science Foundation (Project Number 200021-140496) is

acknowledged for financial support. Dr. S. Mitchell is thanked for TEM analyses.

38

REFERENCES

1. P. Mäki-Arvela, B. Holmbom, T. Salmi, D. Yu. Murzin, Catal. Rev. 49 (2007)

197-340.

2. D. Yu. Murzin, I. L. Simakova, Catalysis in biomass conversion, in

Comprehensive Inorganic Chemistry II, vol. 7, From Elements to Applications,

R. Schlögl and J. W. Niemantsverdriet (Eds.) 2013, Elsevier, vol. 7, pp. 559-

586.

3. M. Yabushita, H. Kobayashi, A. Fukuoka, Appl. Catal., B145 ( 2014) 1-9.

4. A. A. Balandin, N. A. Vasyunina, G. S. Barysheva, S.V. Chepigo, Bull. USSR

Acad. Sci., Div. Chem. Sci. 6 (1957), 392.

5. A. Fukuoka, P. L. Dhepe, Angew. Chem. Int. Ed. 45 (2006) 5161-5163.

6. Y. B. Huang, Y. Fu, Green Chem. 15, 2013, 1095-1111.

7. J. Geboers , S. Van de Vyver , K. Carpentier, P. Jacobs, B. Sels, Green Chem. 13

(2011) 2167-2174.

8. A. Negoi, K. Triantafyllidis, V. I. Parvulescu, S.M. Coman, Catal. Today 223

(2014) 122-128.

9. J. Geboers, S. Van de Vyver, K. Carpentier, P. Jacobs, B. Sels, Chem. Comm. 47

(2011) 5590-5592.

10. M. Käldström, N. Kumar, D. Yu. Murzin, Catal. Today 167 (2011) 91-95.

11. M. Käldström, N. Kumar, M. Tenho, M. V. Mokeev, Y. E. Moskalenko, D. Yu.

Murzin, ACS Catal. 2 (2012) 1381-1393.

12. .P. Mäki-Arvela, T. Salmi, B. Holmbom, S. Willför, D. Yu. Murzin, Chem. Rev.

111 (2011) 5638-5666.

13. B. T. Kusema, G. Hilmann, P. Mäki-Arvela, S. Willför, B. Holmbom, T. Salmi,

D. Yu. Murzin, Catal. Lett. 141 (2011) 408-412.

39

14. P. Demma Carà, M. Pagliaro, A. Elmekawy, D. R. Brown, P. Verschuren, N. R.

Shiju, G. Rothenberg, Catal. Sci. Technol. 3 (2013) 2057-2061.

15. V. A. Sifontes Herrera, O. Oladele, K. Kordas, K. Eränen, J. -P. Mikkola, D.

Yu. Murzin, T.Salmi, J. Chem. Techn. Biotech. 86 (2011) 658-668.

16. V. A. Sifontes Herrera, F. Salem, B. Kusema, K. Eränen, T.Salmi, Top. Catal.,

55 (2012) 550-555.

17. A. Aho, S. Roggan, O. Simakova, T. Salmi, D. Yu. Murzin, Catal. Today 241

(2015), 195-199.

18. B. T. Kusema, L. Faba, N. Kumar, P. Mäki-Arvela, E. Díaz, S. Ordoñez, T.

Salmi, D. Yu. Murzin, Catal. Today 196 (2012) 26-33.

19. B. T. Kusema, L. Faba, N. Kumar, P. Mäki-Arvela, E. Díaz, S. Ordoñez, T.

Salmi, D. Yu. Murzin, Microporous Mesoporous Mater. 189 (2014) 189-199.

20. I.V. Sumerski, S. M. Krutov, M. Ya. Zarubin, Russ. J. Appl. Chem. 83 (2010)

321-328.

21. S. K. R. Patil, C. R. F. Lund, Energy Fuels 25 (2011) 4745-4755.

22. S. K. R. Patil, J. Helzel, C. R. F. Lund, Energy Fuels 26 (2012) 5281-5293.

23. I. van Zandvoort, Y. Wang, C. B. Rasendra, E. R. H. van Eck, P. C. A.

Bruijnincx, H. J. Heeres, B. M. Weckhuysen, ChemSusChem 6 (2013) 1745-

1758.

24. Y. Nakagawa, M. Tamura, K. Tomishige, ACS Catal. 3 (2013) 2655-2668.

25. T. D. Swift, C. Bagia, V. Choudhary, G. Peklaris, V. Nikolakis, D. G. Vlachos,

ACS Catal. 4 (2013) 259-267.

26. A. H. Øygarden, J. Pérez-Ramírez, D. Waller, K. Schöffel, D. Brackenbury,

WO2004/110622, 2004.

27. M. Santiago, A. Restuccia, F. Gramm, J. Pérez-Ramírez, Microporous

Mesoporous Mater. 146 (2011) 76-81.

40

28. F. Frechard, P. Sautet, Surf. Sci. 389 (1997) 131-146.

29. A.S. Boymirzaev, Sh. Shomuratov, A.S. Turaev, Khimija Rastitel’nogo Syr’ja, 2

(2013) 51-55.

30. B. T. Kusema, C. Xu, P. Mäki-Arvela, S. Willför, B. Holmbom, T. Salmi, D.

Yu. Murzin, Int. J. Chem. Reactor Eng. 8 (2010) 1-18.

31. S. Liu, M. Tamura, Y. Nakagawa, K. Tomishige, ACS Sustain. Chem. Eng. 2

(2014) 1819-1827.

32. J. Dijkmans, D. Gabriëls, M. Dusselier, F. de Clippel, P. Vanelderen, K.

Houthoofd, A. Malfliet, Y. Pontikes, B. F. Sels, Green Chem. 15 (2013) 2777 -

2785.

33. D. Kubicka, N. Kumar, P. Mäki-Arvela, M. Tiitta, V. Niemi, H. Karhu, T.

Salmi, D. Yu. Murzin, J. Catal, 227 (2004) 313-327.

34. D. Kubicka, N. Kumar, T. Venäläinen, H. Karhu, I. Kubickova, H. Österholm,

D.Yu. Murzin, J. Phys. Chem. B, 110 (2006) 4937-4946.

35. J.I. Villegas, D. Kubicka, H. Karhu, H. Österholm, N. Kumar, T. Salmi, D.Yu.

Murzin, J. Molec. Catal. A. Chem., 264 (2007) 192-201.

36. L.A. Pedersen, J.H. Lunsford, J Catal. 61 (1980) 39-47.

37. S. Van de Vyver, J. Geboers, W. Schutyser, M. Dusselier, P. Eloy, E.Dornez, J.

Won Seo, C. M. Courtin, E. M. Gaigneaux, P. A. Jacobs, B. F.

Sels, ChemSusChem, 5 (2012) 1549-1558.

38. S. Van de Vyver, J. Geboers, P. A. Jacobs, B. F. Sels, ChemCatChem, 3 (2011)

82–94.

39. R. Palkovits, K. Tajvidi, A. M. Ruppert, J. Procelewska, Chem. Commun., 47,

(2011) 576–578

40. R. M. Ravenelle, F. Schüβler, A. D’Amico , N. Danilina, J. A. van Bokhoven, J.

A. Lercher, C. W. Jones, C. Sievers, J. Phys. Chem. C, 114 (2010) 19582–

19595.

41

41. T. Ennaert, J.Geboers, E.Gobechiya, C.M.Courtin, M.Murttepeli, K.Houthoofd,

C. E.A. Kirschhock, P. C.M.M. Magusin, S.Bals, P. A. Jacobs, B. F. Sels, ACS

Catal.,5 (2015) 754–768.

42. K. Tajvidi, P. Hausoul, R. Palkovits, ChemSusChem, 7 (2014) 1311-1317.

43. R. Sun, T. Wang, M. Zheng, W. Deng, J. Pang, A. Wang, X. Wang, T. Zhang,

ACS Catal. 5 (2015) 874-883.

44. B. Danon, G. Marcotullio, W. de Jong, Green Chem. 16 (2014) 39-54.

45. R. –J. V. Putten, J. N. M. Soetedjo, E. A. Pidko, J. C. van der Waal, E. J. M.

Hensen, E. de Jong, H. J. Heeres, ChemSusChem 6, (2013) 1681-1687.

46. D. J. Liu, E. Y. -X. Chen, ACS Catal. 4 (2014) 1302-1310.

47. H. Haario, ModEst 6.0, User Guide, Helsinki, 2010.