Hydrogenation of Glucose to Sorbitol over Nickel and Ruthenium Catalysts

Article abstract of DOI:10.1002/adsc.200390024

Hydrogenation of aqueous glucose solution was performed in batch and continuous reactors using supported nickel and ruthenium catalysts. Preparation methods were precipitation, impregnation, solgel and template syntheses, and SiO₂, TiO₂, Al₂O₃ and carbon were used as support materials. A procedure for the one-step synthesis of templated metal on support catalysts was established. The influence of support material and preparation methods was studied and the results were compared with those of an industrial nickel catalyst. In a detailed study taking more than 1100 h time on stream we investigated the deactivation of the industrial catalyst and a supported ruthenium catalyst, and evaluated the deactivation mechanisms. As yet unpublished side products of the glucose hydrogenation were identified and a scheme of the reaction network was set up.

Full text of DOI:10.1002/adsc.200390024

FULL PAPERS
Hydrogenation of Glucose to Sorbitol over Nickel and
Ruthenium Catalysts
Burkhard Kusserow, Sabine Schimpf, Peter Claus*
Ernst-Berl-Institute of Chemical Engineering and Macromolecular Chemistry, Darmstadt University of Technology,
Petersenstr. 20, 64287 Darmstadt, Germany
Fax: ( ‡ 49)-6151-16-5369, e-mail: claus@ct.chemie.tu-darmstadt.de
Received: August 2, 2002; Accepted: October 27, 2002
Abstract: Hydrogenation of aqueous glucose solution more than 1100 h time on stream we investigated the
was performed in batch and continuous reactors using deactivation of the industrial catalyst and a supported
supported nickel and ruthenium catalysts. Prepara- ruthenium catalyst, and evaluated the deactivation
tion methods were precipitation, impregnation, sol- mechanisms. As yet unpublished side products of the
gel and template syntheses, and SiO2, TiO2, Al2O3 and glucose hydrogenation were identified and a scheme
carbon were used as support materials. A procedure of the reaction network was set up.
for the one-step synthesis of templated metal on
support catalysts was established. The influence of
support material and preparation methods was stud- Keywords: deactivation; glucose hydrogenation;
ied and the results were compared with those of an leaching; mesoporous molecular sieves; nickel cata-
industrial nickel catalyst. In a detailed study taking lyst; ruthenium catalyst
Introduction
65 wt %, either in discontinuous (autoclave) or contin-
uous (trickle bed) reactors under high pressure.
Up to now, most catalysts for sorbitol hydrogenation
Sorbitol is the polyol with the most widespread use in
nutrition, cosmetics, medical and industrial applications. are based on nickel as active metal. Historically, Raney
It is used as a low calorie sweetener and sugar substitute nickel was used because of its economic price, later
for diabetics, as a humectand in cosmetic and pharma- supported nickel catalysts were more frequently used
ceutical products, in paper and tobacco and other because theyare more active. To increase activitysome
applications. It has no cariogenic activity, so most investigators used promoted nickel catalysts. Li et al.[1]
toothpastes are based on sorbitol. Additionally, sorbitol used nickel-boron amorphous alloycatalysts promoted
does not undergo the Maillard reaction. It is used as a with chromium, molybdenum and tungsten and found
feedstock for vitamin C and for diol components for the highest activities for catalysts promoted with tungsten.
production of polymers. The world-wide production In another article, Li et al.[2] used Raneynickel
volume exceeded 1,000,000 tons/a in 2000.
Sorbitol is produced bycatalytic hydrogenation of
promoted with phosphorus. Gallezot et al.[3] studied
Raneynickel promoted with molybdenum, chromium,
aqueous glucose solutions with glucose contents of up to iron and tin. The kinetics of the continuous glucose
hydrogenation with a supported nickel catalyst in a
trickle bed reactor was studied byDe champ et al.[4] But
¬
nickel has some disadvantages: the nickel leaches,
resulting in a loss of activityand a high nickel content
in the sorbitol solution. For use in food, medical and
cosmetic applications, the nickel must be removed
completelyfrom the sorbitol, resulting in high addi-
tional costs. Therefore, catalysts based on other active
metals were evaluated, including cobalt, platinum,
palladium, rhodium and ruthenium.[1,5,6] The best activ-
ities were found for supported ruthenium catalysts.
Furthermore, ruthenium is not dissolved under the
reaction conditions of the glucose hydrogenation.[5,6] K.
van Gorp et al.[7] used ruthenium on carbon supports for
discontinuous hydrogenation of glucose, Heinen et al.[8]
CH2OH
OH
CHO
H
HO
H
H
HO
H
OH
H
H2
H
OH
OH
OH
OH
Catalyst
H
H
CH2OH
Sorbitol
CH2OH
D-Glucose
Figure 1. Hydrogenation of glucose to sorbitol.
Adv. Synth. Catal. 2003, 345, No. 1+2
¹ 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1615-4150/03/3451+2-289 299 $ 17.50+.50/0
289

Burkhard Kusserow et al.
FULL PAPERS
used the same combination for the hydrogenation of ruthenium template catalyst. The catalysts were tested
fructose to mannitol. in an autoclave, the best ruthenium catalyst was used for
Continuous hydrogenation of glucose with ruthenium a deactivation studyin a trickle bed reactor for 1155 h
catalysts in a trickle bed reactor was performed by TOS.
Arena[9] and Gallezot et al.[10] Arena focused his interest
The characterisation of the catalysts was done by the
on the deactivation mechanisms of catalysts composed following techniques: atomic emission spectroscopy
of ruthenium on alumina supports. He found physical (ICP-OES) to estimate the metal contents of the
changes in the support material and poisoning by catalysts, temperature programmed reduction or oxida-
gluconic acid, iron and sulphur as causes for deactiva- tion (TPR/TPO), transmission electron microscopy
tion. Gallezot used ruthenium on carbon supports. A (TEM) to determine the metal particle size as well as
catalyst with a ruthenium load of 1.6 wt % had a X-rayphotoelectron spectroscopy(XPS). H
chemi-
2
conversion rate of 98.6% after 52 69 h time on stream sorption was used to estimate the metal surface of some
(TOS), the conversion rate dropped to 94.4% after 312 h catalysts. ICP-OES was also used to investigate the
TOS, indicating a remarkable deactivation of the metal leaching bydetermining the metal content in the
catalyst.
The aim of our studywas toovercome the deactivation
product solution. The elemental composition of one
catalyst was analysed with X-ray fluorescence (XRF).
and leaching problems associated with the present
commercial nickel catalysts. One way was to change
support materials and preparation methods to achieve a Results and Discussion
better anchoring of the nickel crystallites on the support.
This was tried bysol-gel and template methods for the
preparation of the catalysts. Additionally, we investi-
Reaction Network
gated impregnation methods using a complexing agent
(ethylenediamine) for nickel on various supports.
The hydrogenation of glucose produced some by-
The other method was the exchange of the active products. Our experiments yielded up to 8 wt % of
metal. Ruthenium, alreadyknown as a veryactive
these compounds. The highest amounts of by-products
catalyst metal for glucose hydrogenation, was used on (~8 wt %) were produced bysupports based on TiO ,
2
different supports. Preparation was done byprecipita-
tion and impregnation. We additionallyprepared a
followed bysupports based on Al O3 (5.5 7.5 wt %)
and carbon (3.4 6.3 wt %). Supports based on SiO2
2
Cleavage
Lobry de Bruyn-Alberda van Ekenstein rearrangement
OH OH
H
H
OH
CHOH
OH
O
O
O
O
HO
Glycol
aldehyde
Dihydroxy-
HO
H
H
OH
Form-
OH
Glycerin
Endiol
acetone
aldehyde
aldehyde
H
OH
CH2OH
[OH–]
[OH–]
[OH–]
O
O
[OH–]
H+; –3 H2O
HOH2C
Dehydration
H
CHO
H
CH2OH
O
CHO
5-Hydroxymethylfurfural
OH
HO
HO
H
H
OH
H
H
HO
H
H
HO
H
H
OH
OH
OH
×2
H
O
H
OH
H
OH
H
OH
Dimerisation
HO
HO
H
H
CH2OH
CH2OH
CH2OH
H
OH
O
OH
H
D-Fructose
H2
D-Mannose
D-Glucose
H
O
D-Maltose
H
HO
H
H2
OH
H2
H
OH
OH
H
CH2OH
H2
CH2OH
H
HO
HO
H
H
H
O
H
HO
H
OH
H
HO
HO
H
H
H
OH
O
Hydrogenation
OH
OH
OH
OH
H
H
OH
H
H
H
OH
OH
HO
CH2OH
CH2OH
Sorbitol
H
Maltitol
OH
H
Mannitol
H
Figure 2. Reaction network for the hydrogenation of glucose.
290
Adv. Synth. Catal. 2003, 345, 289 299

Hydrogenation of Glucose to Sorbitol over Nickel and Ruthenium Catalysts
FULL PAPERS
produced less by-products (2.1 4.1 wt %). Most of ymethylfurfural were not found. The hydrogenation
these by-products had been unknown so far. As sorbitol products of these compounds were mannitol and
is used for nutrition and as some of the by-products were maltitol. Hydrogenation products of the other carbonyls
responsible for deactivation of the catalysts,[9] it is were not found, probablybecause theyreact slower than
important to identifythese compounds. We investigated
these products bycomparison with standards, after
glucose. Heinen et al.[8] reported that glucose adsorbs
much stronger on Ru-catalysts than fructose, leading to
derivatisation with hydroxylamine (both HPLC) and an inhibition of the hydrogenation of fructose under
with GC subsequent to the derivatisation with hexame- mild reaction conditions. According to this, the mannitol
thyldisilazane. We identified mannose, fructose, maltose, yield is low compared to fructose even when glucose
glycerin aldehyde, dihydroxyacetone, glycol aldehyde, conversion to sorbitol is high in our experiments. This
formaldehyde, 5-hydroxymethylfurfural, mannitol, mal- mayalso be alimiting factor for the hydrogenation of the
titol and iditol. From these, fructose, mannitol and iditol other by-products.
were alreadyreported byother research groups.
The formation of iditol is not yet explained. In the
The primary by-products were built by non-catalytic literature, the common explanation is isomerisation of
pathways. Mannose and fructose were products of the sorbitol. But it still cannot be explained whyiditol
Lobryde Bruyn À Alberda van Ekenstein rearrange- should be the onlyproduct of the isomerisation. Some
ment, glycerin aldehyde, dihydroxyacetone, glycol al- minor by-products remained unidentified. The identi-
dehyde, formaldehyde were formed by alkaline cleav- fication of most of these compounds allows the formu-
age of sugars and 5-hydroxymethylfurfural was formed lation of a reaction network, see Figure 2.
by dehydration. 5-Hydroxymethylfurfural is a marker
substance for non-enzymatic browning under acidic
conditions in absence of amines.[11] In our experiments, Characterisation and Catalytic Results of the Nickel
5-hydroxymethylfurfural was detectable before a yellow Catalysts Prepared with Ethylenediamine
colour of the product solution occurred, indicating its
usefulness as a marker. It seems to be surprising that The catalysts were prepared by impregnation (marked
alkaline cleavage of sugars could take place at a pH of bythe suffix ™en∫) and incipient wetness (marked with
3 4, depending on the experiment. We consider the the suffix ™Ren∫) with aqueous solutions of [Ni(ethyle-
alkaline properties of the surface of the support as nediamine)n(H2O)6 2n](NO3)2. The catalysts, the sup-
source for this reaction. Acidic properties of supports ports, the formation method used and the metal contents
mayalso be the reason for the formation of 5-hydrox-
are displayed in Table 1. All catalysts were characterised
ymethylfurfural. No experiment yielded both products bytheir reaction results, metal contents and bythe
of the alkaline decay and 5-hydroxymethylfurfural at analysis of the product solution for nickel leaching.
once. This conjecture is supported bythe observation
that these compounds were also produced bythe pure
Reactions were performed with 50 wt % glucose solu-
tion in an autoclave at 393 K, 0.5 g of catalyst and at a
supports of the corresponding catalysts. In blank runs we hydrogen pressure of 120 bar. Selected catalysts were
found fructose, mannose and maltose in small quantities further investigated with XRD for nickel crystallite
(< 2 wt %) but products of the cleavage and 5-hydrox- sizes, with TPR and chemisorption.
Table 1. Nickel catalysts prepared via ethylenediamine, catalyst name, support, thermal treatment, Ni content of the catalyst
and of the product solution after the reaction.
Catalyst
Support
Thermal
Nickel content
[wt %]
Leaching
treatment[b]
[mg  mlÀ1]
Ni68T[a]
Ni/Al2O3 C-en
Ni/Al2O3 C-Ren
Ni/TiO2-en
Ni/TiO2-Ren
Ni/SiO2 F10-en
Ni/SiO2 F10-Ren
Ni/SiO2 F11-en
Ni/SiO2 F11-Ren
Ni/C(I)-en
SiO2
Al2O3 C
Al2O3 C
TiO2 P25
66.78
5.07
9.11
3.26
6.87
16.92
12.93
5.48
12.83
4.81
5.55
5.62
125.87
66.32
68.77
22.29
30.61
79.94
100.23
40.47
34.10
1.61
A
A
A
A
A
A
A
A
B
TiO2 P25
Silica gel large pore 58 micron
Silica gel large pore 58 micron
Silica gel large pore 99.5%
Silica gel large pore 99.5%
Black Pearls 2000 GP3755
Black Pearls 2000 GP3755
Vulkan XC72R GP3759
Ni/C(I)-Ren
Ni/C(II)-en
B
B
7.04
6.11
[a]
Industrial nickel catalyst for comparison.
A: calcinated in air at 773 K, reduced in H2 at 773 K; B: heated in He at 623 K, reduced in H2 at 773 K.
[b]
Adv. Synth. Catal. 2003, 345, 289 299
291

Burkhard Kusserow et al.
FULL PAPERS
Table 2. Nickel crystallite sizes for selected catalysts pre-
pared with ethylenediamine.
Catalyst
Ni particle size [nm]
Ni/Al2O3 C-en
Ni/Al2O3 C-Ren
Ni/TiO2-en
2.2
4.7
9.7
Ni/SiO2 F10-en
Ni/SiO2 F10-Ren
2.5/10.2
2.6/11.6
Figure 3. Conversion of glucose, selectivityand iyeld to
sorbitol of the nickel catalysts prepared with ethylenedi-
amine, discontinuous hydrogenation.
Figure 5. Temperature programmed reduction of selected
catalysts prepared with ethylenediamine.
Figure 4. Specific activityof the catalsyts prepared with
ethylenediamine, discontinuous hydrogenation.
At the beginning, we used different reduction temper- for Ni/Al2O3 C-Ren (4.7 nm). This explains the higher
atures for these investigations. Since catalysts reduced at activityfor the catalsyt prepared byimpregnation.
773 K had a higher activitythan those reduced at 573 K,
we used the higher reduction temperature for further large crystallite size (9.7 nm). The catalysts Ni/SiO2 F10-
investigations. en and Ni/SiO2 F10-Ren had a bimodal crystallite size
Ni/TiO2-en had a high activityin spite of its relatively
Regarding the catalytic results, displayed in Figure 3, distribution with one population of a mean size of
alumina seems to be the best support for catalyst approximately2.5 nm diameter, the other of about 11 nm.
preparation. Ni/Al2O3 C-en shows conversion compa-
TPR analysis of these catalysts (Figure 5) showed that
rable with the industrial catalyst, but the selectivity to the reduction temperature depends on the nature of the
sorbitol is only88%, whereas the industrial catalyst
reached 92% selectivityat same conditions. The cata-
support. The bimodal crystallite distribution of Ni/SiO2
F10-Ren is also shown as two maxima at Tˆ 613 K and
lysts prepared on titania and silica had lower conversion, Tˆ 793 K in the TPR diagram.
the carbon supports used were not suited for this
H2 chemisorption was performed to estimate the Ni
reaction. Regarding the specific activities of the cata- surface. For the 5 catalysts shown in Figure 6, Ni surfaces
lysts, expressed as mmol sorbitol per gram nickel and per were in the range between 1.3 and 17 m2/gcatalyst. Figure 6
second, displayed in Figure 4, the activity follows the shows the Ni surface, the Ni content, the specific activity
sequence Al2O3 > TiO2 > SiO2 > C. In most cases, and overall yield for the catalysts. These results suggest
catalysts prepared by impregnation had a higher activity that catalytic results depend on the nature of the support
than those prepared byincipient wetness.
rather than on nickel content or nickel surface of the
Some of these catalysts were also investigated with catalyst. A possible explanation is that the support has
XRD to estimate the nickel crystallite size, the results an influence on the balance between the tautomeric
are given in Table 2. This demonstrates that the nickel forms of glucose. Makkee et al.[12] reported for the
crystallites are smaller for Ni/Al2O3 C-en (2.2 nm) than hydrogenation of fructose over Cu and Ni catalysts that
292
Adv. Synth. Catal. 2003, 345, 289 299

Hydrogenation of Glucose to Sorbitol over Nickel and Ruthenium Catalysts
FULL PAPERS
catalysts. One of these strategies involved the suspen-
sion of an organic metal precursor (e.g., nickel acetyl-
acetonate) in the template solution. The precursor was
dissolved in the micelles of the surfactant and the
support precipitated after addition of tetraethyl ortho-
silicate (TEOS) around the micelles. We added different
amounts of toluene to the synthesis mixture to influence
the pore diameter of the resulting materials. These
catalysts were indexed with the suffix Org. The template
was removed in the calcination step.
Another strategyis to synthesise templated supports
with organic anchor groups which serve as ligands for
ions of the active metal. Now we used trimethylsilyl-
propyl-ethylenediamine as complexing agent. The an-
chor precursor was added to the TEOS and both were
Figure 6. Ni surface, Ni content, specific activity, yield and Ni
leaching of selected catalysts.
the different tautomeric forms had different reactivities, added to the template solution. After precipitation of
so a change of this balance will result in a change of the the support, either a metal salt was added directlyto the
overall reactivity. Glucose reacts predominantly as the reaction mixture (Ni20Sp1), or, for use as a stock for
open-chain form, thus a higher concentration of this different metals on support catalysts, the template was
form will result in higher conversion rates.
All tested catalysts were leaching (see Table 1). to a slurryof the support (Ni20S and Ru2S).
Together with the relativelylow nickel content of the We synthesised catalysts with nickel, ruthenium and
extracted from the support and the metal salt was added
best catalysts, this will result in a rapid deactivation. For silver and used them in different reactions. Here we
this reason no experiment in the trickle bed reactor was report onlythe results in the hydrogenation of glucose.
performed.
Characterisation of the catalysts was done with N2
physisorption and by analysis of the nickel content of the
catalyst and the product solution. The results are
Characterisation and Catalytic Results of Sol-Gel and displayed in Table 3.
Template Catalysts All template catalysts showed a very narrow pore size
distribution, BET surfaces ranged from 366 to
The discontinuous hydrogenation of glucose was per- 803 m2 gÀ1. For the series of Org-catalysts, we expected
formed with 50 wt % glucose solution in an autoclave at an increase in pore diameter with increasing toluene
393 K and at a hydrogen pressure of 120 bar. The sol-gel concentration. Surprisingly, we registered the opposite.
catalysts gave no conversion to sorbitol. In BETanalysis, A competition between the nickel acetylacetonate and
a BET surface of 350 m2 gÀ1 was measured, the BJH pore the toluene on the place in the micelles of the template
size was <2 nm. We deduced from this that the pore size can explain this behaviour; accordingly, we found a
was too low for an effective transport of glucose in the decrease of nickel content in the resulting catalyst with
pores. Therefore we do not give explicit data on these increasing toluene content in the synthesis mixture.
experiments. So we made the following syntheses using a
templating strategyto reach wider pores.
There is a strong correlation between toluene con-
centration and pore diameter for these catalysts, repre-
Mesoporous molecular sieves synthesised with sur- sented in Figure 7. This now enables us to synthesise
factants as templates have been investigated since catalysts with predefined pore sizes.
1992.[13] Despite their use as supports for catalytic active
Nickel leaching was estimated byanalsyis of the
[14]
metals, onlyfew applications
deal with the direct product solution for the catalysts Ni20Org1 and
synthesis of templated metal on support catalysts of this Ni20Sp1, the results were 87.26 and 32.74 mg nickel
type. We tried to develop generalised procedures for mlÀ1 product solution, respectively. This indicates a
one-step syntheses of templated metal on support relativelyhigh leaching for both catalysts. In Figure 8,
Table 3. Properties of the template catalysts.
Catalyst
Ni20Org1
Ni20Org2
Ni20Org3
Ni20Sp1
Ni20S
Ru2S
BET [m2 gÀ1]
367
5.5
21.4
7.7
803
4.3
18.5
4.2
708
3.3
10.4
6.8
464
5.1
20.3
2.9
604
1.5
2.1
0.5
509
1.5
2.0
4.1
BJH[a] Pore size [nm]
Metal content [wt %]
Sorbitol yield [%]
[a]Pore size calculation method from Barret, Joyner and Halenda.
Adv. Synth. Catal. 2003, 345, 289 299
293

Burkhard Kusserow et al.
FULL PAPERS
amine yield more sorbitol (up to 29 wt %). Ni20Org1,
with the highest nickel content and the widest pores
yielded most sorbitol (7.7%), Ni20S with the lowest
nickel content and the narrowest pores gave least
sorbitol (0.5%). We think that the low overall conversion
is caused bya transport limitation in the long and
relativelynarrow pores of these materials. This is
supported byhigh conversion rates for these catalysts
in the hydrogenation of smaller molecules (e.g., acrolein
and citral). Ni20Sp1 showed a different behaviour, in
spite of its high nickel content and pore diameter, the
sorbitol yield was very low. The present data are
inadequate for an explanation. Insufficient reduction
of the catalyst may be the reason, however. The reduced
catalyst was grey, Ni20Org1 with comparable nickel
content and pore diameter was black. The higher
sorbitol yield of Ru2S (4%) compared to Ni20S (0.5%)
reflected the higher activityof ruthenium compared to
nickel. Taking into account the low conversion rate and
the relativelyhigh leaching of nickel, these catalysts
were not suited for the hydrogenation of glucose.
Figure 7. Dependence of the pore diameter from the toluene
content of the synthesis mixture for the catalysts Ni20Org1,
Ni20Org2 and Ni20Org3.
Characterisation and Catalytic Results of the
Ruthenium Catalysts
Because of the high activityof ruthenium compared to
nickel, the ruthenium catalysts were prepared with
lower metal contents than the nickel catalysts. Metal
contents for the Ru catalysts were between 0.1 and
2.8 wt %. The catalysts are listed in Table 4, preparation
methods and supports are also given. In Figure 9 the
catalytic results are displayed.
Figure 8. Dependence of the nickel content from the pore
diameter for the templated nickel catalysts.
The discontinuous hydrogenation of glucose was
the correlation between pore diameter and nickel performed with 50 wt % glucose solution in a autoclave
content of all template catalysts is shown. at 393 K and at a hydrogen pressure of 120 bar.
Glucose hydrogenation results, displayed in Table 3, Ruthenium catalysts with 1 wt % or more ruthenium
are insufficient for all template catalysts compared to had conversion rates equal or superior to the commer-
the industrial catalyst, which exhibited a sorbitol yield of cial catalyst Ni68T. The selectivities were slightly lower,
35%. Also, most catalysts prepared with ethylenedi- caused bythe formation of side products. The produc-
Table 4. Catalyst name, preparation method, support, metal content and metal particle size of the ruthenium catalysts.
Catalyst
Preparation method
Support
Metal content
[%]
Particle size
[nm]
Ru3Ap
Ru1Ap
Ru1Ai
precipitation
precipitation
impregnation
impregnation
impregnation
impregnation
impregnation
impregnation
impregnation
precipitation
Al2O3 C
Al2O3 C
Al2O3 C
Al2O3
Al2O3
Al2O3
TiO2 P25
TiO2 P25
TiO2
2.75
1.04
0.95
0.97
0.84
0.47
0.27
0.11
1.01
66.78
1.0
1.4
1.6
Ru09Ai
Ru08Ai
Ru05Ai
Ru03Ti
Ru01Ti
Ru1Ti
1.9
[b]
1.8
Ni68T[a]
SiO2
[a]
Industrial nickel catalyst for comparison.
Not measured.
[b]
294
Adv. Synth. Catal. 2003, 345, 289 299

Hydrogenation of Glucose to Sorbitol over Nickel and Ruthenium Catalysts
FULL PAPERS
Figure 9. Conversion of glucose, selectivityand iyeld to
sorbitol of the ruthenium catalysts compared with the
industrial nickel catalyst Ni68T.
Figure 11. Conversion of glucose, selectivityto sorbitol and
nickel leaching of the industrial catalyst Ni68T during more
than 1100 hours time on stream in continuous hydrogenation.
pressure of 80 bar, reaction temperature was 353 K. The
results are represented in Figure 11 for the nickel
catalyst and will be given later for the ruthenium
catalyst. At the beginning of the experiment, the
conversion of glucose was 99.3% for the nickel catalyst
and 99.9% for the ruthenium catalyst. Selectivity to
sorbitol was 98.8% and 99.0%, respectively. The con-
version dropped below 98% after 770 h TOS for the
nickel catalyst, after 1080 h TOS for the ruthenium
catalyst. The selectivity for both catalysts remained
unchanged during the experiments.
Figure 10. Specific activity, given as mmol sorbitol per gram
catalyst and hour of the ruthenium catalysts compared with
the industrial nickel catalyst Ni68T.
Deactivation
tion of side products depended on the support and the After 250 h TOS the commercial nickel catalyst showed
contact time. In experiments performed in the trickle a nearlyconstant level of leaching over the whole
bed reactor we found smaller amounts of side products. reaction time (see Figure 11). Higher leaching at the
The specific activityof the catalysts, shown in Figure 10,
beginning of the experiment was probablydue to the
is highest for ruthenium contents between 0.3 and loss of small particles of the catalyst remaining after the
0.9 wt %, catalysts with higher or lower ruthenium initial rinsing step. In TEM pictures of the fresh and the
contents displayed lower specific activities.
used catalyst, see Figures 12 and 13, we registered a
change in particle size distribution to higher crystallite
We could not detect anyleaching of ruthenium in our
experiments, ruthenium detection limit was 80 mg LÀ1 of sizes resulting in a loss of active surface after 400 h TOS.
product solution.
These results were promising, so a larger quantityof a
The particle size distribution changed from a mono-
modal to a bimodal distribution. The particles sized
ruthenium catalyst was prepared on a alumina support between 1 9 nm had completelydisappeared, particles
and used for a long-time experiment in the trickle bed with a diameter > 10 nm had been built up. In XRD
reactor.
analysis, the mean particle size increased from 10 nm to
29 nm. The nickel content of the catalyst decreased by
25% (from 66.78 wt % to 50.53 wt %) during the experi-
ment. There appeared also a visible disintegration of the
support, not surprising for silica under hydrothermal
Continuous Hydrogenation
We investigated the commercial comparison catalyst conditions. The BET surface decreased by60% (from
Ni68Tand the ruthenium catalyst Ru05Ai for more than 180 to 85 m2 gÀ1) during this time. Additionally, the pore
1100 h time on stream (TOS) in the trickle bed reactor size distribution changed considerably. We registered an
under standard conditions: 40 wt % glucose solution increase in the fraction of mesopores, the fraction of
was delivered at a flow rate of 40 mL/h to the reactor micropores decreased. Therefore the stabilityproblems
packed with 40 mL of catalyst, H2 flow was 23 L/h at a were caused bythe hot and acidic reaction media (pH 4
Adv. Synth. Catal. 2003, 345, 289 299
295

Burkhard Kusserow et al.
FULL PAPERS
Figure 14. Conversion of glucose, selectivityto sorbitol and
ruthenium leaching of the catalyst Ru05Ai during 1150 h
TOS in continuous hydrogenation.
Figure 12. TEM image of the unused industrial catalyst
Ni68T.
Figure 13. TEM image of the industrial catalyst Ni68T after
400 h TOS. A: carbon film, B: SiO2, C: nickel crystallites
> 10 nm, D: nickel crystallites about 1 nm.
Figure 15. TEM image of the catalyst Ru05Ai after use in the
continuous hydrogenation of glucose for 1155 h TOS.
in the product solution). Furthermore glucose and distribution, shown in Figure 16, is monomodal with a
sorbitol are chelating molecules. The result was a maximum at 3 4 nm. Within the analytical tolerance we
permanent drag-out of nickel and silica with the product found no difference in particle size distribution between
solution.
fresh and used catalyst.
Despite this, there is some deactivation after approx-
The ruthenium catalyst was investigated for 1155 h
TOS. The results are displayed in Figure 14. As men- imately900 h TOS. After 1155 h TOS we tried to restore
tioned above, the conversion of glucose dropped from the activitybyrinsing with distilled water for 72 h under
the initial 99.9% below 98% after 1080 h TOS whereas standard conditions to remove possible water-soluble
the selectivityremained unchanged. No leaching of
ruthenium could be detected byICP-OES for Ru05Ai
contaminants. The activitydecreased further over the
rinsing step. A possible reason could be fouling due to
(see Figure 14). BET surface and pore size distribution high molecular products of the non-enzymatic brown-
remained unchanged during this time, XRF spectra of ing, but we did not find the typical marker substance 5-
the fresh and the used catalyst indicated no difference in hydroxymethylfurfural in the product solutions of this
ruthenium content.
experiments. Furthermore, the activityshould increase
The TEM image (see Figure 15) of the used catalyst or remain constant over the rinsing step because no
shows that most of the ruthenium is present in the form further contamination byproducts of the reaction is
of crystallites with diameters below 5 nm. Some crys- possible. Arena[9] identified gluconic acid, iron (from the
tallites are bigger, in a detailed evaluation we found stainless steel reactor) and sulphur as deactivating
crystallites up to a diameter of 75 nm. The crystallite size compounds. We excluded oxygen by degassing of the
296
Adv. Synth. Catal. 2003, 345, 289 299

Hydrogenation of Glucose to Sorbitol over Nickel and Ruthenium Catalysts
FULL PAPERS
reactor, is the main deactivation mechanism for the
ruthenium catalyst. Teflon liners in the reactor should
prevent the poisoning of the catalyst.
Another charge of this catalyst was tested in an
industrial pilot plant with a catalyst load of 575 mL over
890 h TOS. The 40 wt % glucose solution was delivered
at a flow rate of 0.5 L hÀ1, reaction temperatures were
between 353 K and 393 K, pressure was 24 MPa. Al-
though the reaction conditions were different our results
were confirmed.
The catalyst behaves differently in batch and contin-
uous hydrogenation. In batch hydrogenation the cata-
lyst showed a conversion of 26% and a selectivity to
sorbitol of 83% compared to the industrial nickel Ni68T
catalyst with 41% and 92%, respectively. In continuous
reaction, the results with respect to conversion and
selectivitywere nearlythe same for Ru05Ai and Ni68T.
These results were confirmed bythe industrial applica-
tion of these catalysts.
Figure 16. Crystallite size distribution of the catalyst Ru05Ai
after use in the continuous hydrogenation of glucose for
1155 h TOS.
Conclusions
Nickel catalysts were deactivated in glucose hydro-
genation byleaching of the metal, bychanges in the
crystallite size distribution and by degradation of the
support. A high nickel content is obviouslynecessaryfor
high activityand long working life. The leaching of
nickel in the acidic and chelating reaction media could
not be prevented bythe preparation methods used here.
Ruthenium catalysts are superior to nickel catalysts in
terms of specific activityand lifetime. Conversion of
glucose and selectivityto sorbitol were comparable at
ruthenium loads of 1% of the nickel content of a
commercial nickel catalyst with 66.78 wt % nickel.
Furthermore, no ruthenium leaching could be detected.
The higher costs of the metal are reduced bya much
lower metal content and minor downtime due to the
prolonged lifetime of the catalyst. The main deactiva-
tion mechanism for the ruthenium catalyst was the
poisoning bymetals leached out from the reactor
material.
Figure 17. Parts of the XRF spectra showing the elemental
composition of Ru05Ai before and after reaction. Picture a)
before reaction, b) after reaction, both with corundum target,
c) before reaction, d) after reaction, both with HOPG target.
feed and storage under helium, in the absence of oxygen
there should be no reaction to gluconic acid. Never-
theless we analysed some samples of the product
solution for gluconic acid. No gluconic acid could be
detected. The sulphur content claimed bythe manufac-
turer of the used glucose is ꢀ0.005%, expressed as SO42 .
We used XRF to detect changes in the elemental
composition of the catalyst. The resulting spectra are
shown in Figure 17. We found no sulphur, the iron
content of the catalyst remained unchanged during
1155 h TOS. Instead, we found an increase in the content
of chromium, silicon and zirconium. Variations for the
other elements were within the analytical uncertainty.
The standard deviation of the XRF was 2.2% of the peak
area (6 experiments). The material of the reactor and
tubing was stainless steel 316. This could explain the
content of chromium and silicon, but zirconium is not
used in SS316. It is also remarkable that some palladium
is present in the catalyst, probably because of impurities
in the delivered ruthenium acetylacetonate. Whether
this had a positive or a negative influence on the
catalytic results, is unknown so far.
In autoclaves a higher ruthenium load of the catalyst is
needed to reach catalytic results comparable with indus-
trial nickel catalysts, compared to continuous reactors.
The pore structure of the catalyst had a significant
influence on the catalytic results. In catalysts with pores
width ꢀ 5 nm the transport of glucose was significantly
hindered.
Experimental Section
Reagents and Supports
Therefore we concluded that poisoning bymetal
Ni(NO3)2, nickel acetylacetonate, ethylenediamine, tetraethyl
impurities, leaching out of the working material of the orthosilicate, NH4OH solution and HPLC water were deliv-
Adv. Synth. Catal. 2003, 345, 289 299
297

Burkhard Kusserow et al.
FULL PAPERS
ered byMerck, RuCl 3, Ru(NO)(NO3)3, and ruthenium acetyl-
acetonate byAlfa Aesar. Glucose was purchased from Fluka
(BioSelect), N-[3-(trimethoxysilyl)propyl]ethylenediamine by
Aldrich, Pluronic 64 was a gift from BASF.
The used commercial supports were TiO2 P25, Al2O3 C
(Degussa), silica gel large pore 58 micron, silica gel large pore
99.5% (Alfa Aesar), Black Pearls 2000 GP3755, Vulkan
XC72R GP3759 (Cabot) and Ai (KataLeuna).
room temperature to 773 K in 2 h and held at this temperature
for 2 h.
The ruthenium catalysts were prepared by precipitation
using RuCl3 and byimpregnation using Ru(acac) 3 as precursor.
For precipitation, the calculated amount of RuCl3 was
dissolved in distilled water (approx. 100 mL per g RuCl3).
The support was filled in a 1 L round-bottom flask and
suspended in water until stirring was possible. The RuCl3
solution was added and stirred for 1 h. The slurrywas heated
(353 K) and 5 N NH4OH solution (3 mL per g RuCl3) was
added through a dropping funnel. The catalyst was filtered off
and dried in an oven (393 K). Calcination was carried out in
flowing air (473 K, 3 h), reduction in a hydrogen flow (573 K,
3 h).
Preparation of Catalysts
The preparation of catalysts using ethylenediamine (en) was
The catalyst Ru05Ai was prepared as follows: the support
(73 g) was suspended in toluene (150 mL) and stirred for
15 min bya magnetic stirrer. Ruthenium acetlyacetonate
(2.907 g) was dissolved in toluene (150 mL) and the solution
was added to the slurryof the support. The slurrywas stirred for
1 h and the toluene was allowed to evaporate at room
temperature. From time to time, the slurrywas swirled to
homogenise it. The catalyst was heated in a stream of helium
within 4 h to 250 8C, this temperature was held for 2 h. For
reduction, the catalyst was heated in a hydrogen stream within
1 h to 350 8C, the temperature was held for 3 h. The other
ruthenium catalysts made by impregnation were synthesised
bythe same procedure using different supports and ruthenium
loads. The catalysts are listed in Table 4.
done byimpregnation using aqueous solution of [Ni(en)
n
(H2O)6-2n](NO3)2].[15,16] As support material several commer-
cial oxides, see Table 1, were used. The catalysts supported on
oxides were dried at 363 K, afterwards calcinated (773 K), both
in flowing air and reduced (773 K) in flowing hydrogen. In
Table 1 this procedure is marked as A. The carbon supports
were dried as before and treated with flowing helium (623 K)
before reduction. This procedure is marked as B.
Three synthesis procedures were used for the template
catalysts. The template was in all cases 2% w/w Pluronic 64 in
distilled water. Reactions were performed in a thermostatted
vessel at 45 8C, equipped with a reflux condenser and a
dropping funnel.
Procedure 1: The template solution (150 mL) was thermo-
statted under stirring. Different amounts of toluene, 2 (the
catalyst named Ni20Org1), 4 (Ni20Org2) or 6 mL (Ni20Org3)
and nickel(II) acetylacetonate (4.37 g) were added. After
30 min of stirring, tetraethyl orthosilicate (17.3 g) was added
dropwise through the dropping funnel. After about 20 min a
green precipitation was formed. The mixture was stirred for
24 h, then the precipitate was removed byfiltration, washed
4 times with distilled water and dried in air overnight.
Procedure 2: The template solution (100 mL) was stirred at
45 8C. Tetraethyl orthosilicate (13.84 g) and N-[3-(trimethox-
ysilyl)propyl]ethylenediamine (3.78 g) as an anchor group for
the metal salt were mixed and added dropwise to the template
solution. The mixture was stirred for 48 h. Ni(NO3)2 ¥ 6 H2O
(4.95 g) was dissolved in distilled water (10 ml) and added
dropwise to the mixture. The colour of the precipitate changed
to blue. This mixture was stirred overnight, the precipitate was
filtered off and washed with 4 portions of water. The resulting
catalyst was named Ni20Sp1.
Procedure 3: The synthesis of the precipitate was carried out
like procedure 2, but with a 5-fold charge of the reagents. The
precipitatewas filtered off, dried inair overnight, andextracted
with ethanol in a Soxhlet extractor for 8 h. Afterwards the
precipitate was dried again overnight. The precipitate (2 g) was
suspended in a mixture of distilled water (20 mL) and ethanol
(1 mL). The metal salt [1.93 g Ni(NO3)2 ¥ 6 H2O for the Ni
catalyst Ni20S or 0.126 g Ru(NO)(NO3)3 for the Ru catalyst
Ru2S] was dissolved in distilled water (20 mL) and added to
the suspension. After 2 h of stirring, the precipitate was filtered
off and washed 4 times with water.
The commercial nickel catalyst Ni68T (KataLeuna, Leuna,
Germany) was used as delivered.
Characterisation of the Catalysts
The characterisation of the catalysts was done using the
following techniques.
The metal contents of the catalysts were determined by
atomic emission spectroscopywith inductivelycoupled plasma
(ICP-OES, Perkin Elmer Optima 3000XL) after dissolving the
materials in a mixture of HF/HNO3 bymeans of an MDS-2000
microwave unit (CEM). The metal leaching was investigated
bydetermining the metal content of the product solution in the
same way. The solution was therefore in most cases also
dissolved in a mixture of HF/HNO3 using an MDS-2000
microwave unit (CEM).
Transmission electron microscopy(TEM) for qualitative
and quantitative characterisation of the catalysts was carried
out using a JEM 100C operating at 100 kV for transmission
electron microscopy(TEM) in bright field and dark field
modes. For electron microscopyexamination the catalsyt
samples were dissolved in 2-propanol, dispersed carefullyin an
ultrasonic bath and then deposited on carbon-coated copper
grids.
BET surface and BJH pore size were analysed on a
Sorptomatic (Fisons) using nitrogen as adsorbent gas. The
average load was 200 mg catalyst. The isotherms were eval-
uated with WinADP (CE Instruments). Standard isotherm for
the BJH analyses was hydroxylated silica. BET calculations
were performed between p/p0 ˆ 0 0.4, BJH calculations
between p/p0 ˆ 0.001 1. Onesample(Ni20Org1) wasanalysed
also on an ASAP 2010 (Micromeretics) to confirm the results
from the Sorptomatic.
All catalysts were dried afterwards in a stream of air (363 K,
24 h). For calcination, the catalysts were heated up in an air
stream from room temperature to 773 K in 4 h, this temper-
ature was held for 2 h. The catalysts were reduced in a
hydrogen stream, the temperature was programmed from
298
Adv. Synth. Catal. 2003, 345, 289 299

Hydrogenation of Glucose to Sorbitol over Nickel and Ruthenium Catalysts
XRF analyses were performed on a SPECTRO X-Lab
FULL PAPERS
Acknowledgements
(SPECTRO Analytical instruments, Kleve). Targets were
HOPG for the range between 0 and 12 keVand Al2O3 for the
range between 0 and 50 keV.
The authors thank the Federal Ministry of Research and
Education (BMBF,Germany) for financial support (contract
03C0290C). Further we thank Dr. C. Mohr for TEM analyses,
Dr. R. Schˆdel,KataLeuna for XRD,TPR and chemisorption
analyses,Dr. M. Heck,TU Darmstadt for XRF analyses.
Hydrogenation of Glucose
The discontinuous hydrogenation was performed in an 300 mL
autoclave (Parr Instruments) at the following conditions:
temperature Tˆ 393 K, total pressure p ˆ 120 bar, glucose
concentration: 40 wt % glucose in aqueous solution (120 mL),
reaction time t ˆ 5 h, stirring at 850 rpm. The catalysts were
ground to < 63 mm, 0.5 g were used for the experiments. The
apparatus was equipped with a mass flow controller to deliver
in addition the amount of hydrogen which was converted
during reaction to carrythe reaction out at constant pressure.
The product distribution was analysed by HPLC/RID detec-
tion (column: Aminex HPX-87C operated at 80 8C, eluent
water, 0.6 mL/min.). Gluconic acid was determined byHPLC
using an Aminex HPX 87H column operated at 45 8C with
0.6 mL/min. 0.0005 M H2SO4 as eluent, detectors were RID
and DAD.
Continuous hydrogenation was performed in a trickle bed
reactor (Catatest, Vinci Technologies, France). The glucose
solution was stored under helium pressure in the feed vessel
and delivered byan HPLC pump to the reactor. Hydrogen was
delivered bya mass flow controller. Both flowed downstream
through the reactor. The reactor was packed with the catalyst
between two plugs of glass wool. Having passed the reactor the
product solution flowed through a heat exchanger to the gas-
liquid separator. Gas and product flows were depressurised to
10 bar. An effluent weir controlled the hydrogen effluent to
maintain system pressure, another directed the liquid to the
product vessel. Standard reaction conditions were: feed 40 mL/
h 40 wt % glucose in distilled water degassed under vacuum
(15 min), 23 L/h hydrogen, pressure of 80 bar, 40 mL of
catalyst, 373 K reactor temperature. The catalyst pellets were
ground and sieved, the fraction between 0.8 and 1.2 mm was
used for filling the reactor. The catalysts were reduced with a
hydrogen flow (40 L/h) under atmospheric pressure (2 h,
473 K). Afterwards theywere rinsed with deionised water
under standard conditions until no further particles were
registered in the effluent. The reaction products were analysed
as indicated before. Some samples were analysed for gluconic
acid.
References
[1] H. Li, H. Li, J.-F. Deng, Catal. Today 2002, 74, 53 63.
[2] H. Li, W. Wang, J.-F. Deng, J. Catal. 2000, 191, 257 260.
[3] P. Gallezot, P. J. Cerino, B. Blanc, G. Fleche, P. Fuertes, J.
¡
Catal. 1994, 146, 93 102.
¬
[4] N. Dechamp, A. Gamez, A. Perrand, P. Gallezot, Catal.
Today 1995, 24, 29 35.
[5] J. Wisniak, R. Simon, Ind. Eng. Chem. Prod. Res. Dev.
1979, 18, 50 57.
[6] J. Wisniak, M. Heshkowitz, S. Stein, Ind. Eng. Chem.
Prod. Res. Dev. 1974, 13, 232.
[7] K. van Gorp, E. Boerman, C. V. Cavenaghi, P. H. Ber-
ben, Catal. Today 1999, 52, 349 361.
[8] A. W. Heinen, J. A. Peters, H. van Bekkum, Carbohydr.
Res. 2000, 328, 449 457.
[9] B. J. Arena, Appl. Catal. A 1992, 87, 219 229..
¡
[10] P. Gallezot, N. Nicolaus, G. Fleche, P. Fuertes, A.
Perrand, J. Catal. 1998, 180, 51 55.
[11] R. Liedke, PhD thesis, Universit‰t M¸nster, Germany,
1999.
[12] M. Makkee, P. G. Kieboom, H. van Bekkum, Carbohydr.
Res. 1985, 138, 225 236.
[13] C. T. Kresge, M. E. Leonovicz, W. J. Roth, J. C. Vartuli,
J. S. Beck, Nature 1992, 359, 710 712.
[14] S. Valange, J. Barrault, A. Derouault, Z. Gabelica,
Micropor. Mesopor. Mater. 2001, 44 45, 211 220.
[15] Z. X. Cheng, C. Louis, M. Che, Stud. Surf. Sci.; 1995, 91,
1027 1036.
[16] S. Schimpf, C. Louis, P. Claus, in preparation.
Adv. Synth. Catal. 2003, 345, 289 299
299