Вы находитесь на странице: 1из 16

Applied Catalysis A: General 253 (2003) 437452

The role of the active phase of Raney-type Ni catalysts in the


selective hydrogenation of d-glucose to d-sorbitol
B.W. Hoffer a , E. Crezee a , F. Devred b , P.R.M. Mooijman a ,
W.G. Sloof b , P.J. Kooyman b , A.D. van Langeveld c, ,
F. Kapteijn a , J.A. Moulijn a
b

a Reactor & Catalysis Engineering, Delft University of Technology, Julianalaan 136, 2628 BL Delft, The Netherlands
Laboratory of Materials Science, Delft University of Technology, Rotterdamseweg 137, 2628 AL Delft, The Netherlands
c Charged Particle Optics, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands

Received 7 April 2003; received in revised form 30 June 2003; accepted 30 June 2003

Abstract
Bulk and surface properties of homemade Raney-type Ni catalysts and their NiAl alloy precursors have been studied
and compared with commercial samples, both promoted and unpromoted. The two starting alloys had a different Ni2 Al3
concentration, resulting in Raney-type Ni catalysts with different NiAl composition. All catalysts have been screened in the
hydrogenation of an aqueous solution of d-glucose (10 wt.%) using a three-phase slurry reactor at 4.0 MPa and 393 K. The
promoters are segregated at the surface of the catalysts and have a beneficial effect on the reaction rate, essentially due to
an increased surface area and stability of the active phase while an enhanced interaction of d-glucose with Ni seems to play
a secondary role. The Raney-type Ni catalysts lose Ni and Al at the applied reaction conditions. Mo does not leach at all.
Fe leaches severely from the catalyst surface, but about 30% of the Fe at the surface is present as inactive, bulk iron. The
major cause of deactivation of the Raney-type Ni catalysts is the presence of d-gluconic acid formed during the reaction. The
effect of d-gluconic acid is two-fold: firstly it blocks the Ni sites making them inactive in d-glucose processing. The poisoned
catalytic sites can be recovered again by a regeneration treatment in hydrogen at 393 K. Secondly, the formation of d-gluconic
acid leads to a severe loss of Ni, which makes product purification necessary.
2003 Published by Elsevier B.V.
Keywords: d-Glucose; d-Sorbitol; d-Gluconic acid; Raney Ni; Promoters; Slurry reactor; Deactivation; XPS; XRD; TPD

1. Introduction
The hydrogenation of d-glucose (dextrose) to
d-sorbitol (Fig. 1) is of great industrial importance,
because d-sorbitol is a valuable additive in food,
Corresponding author. Tel.: +31-152-789176;
fax: +31-152-783760.
E-mail address: a.d.vanlangeveld@tnw.tudelft.nl
(A.D. van Langeveld).

0926-860X/$ see front matter 2003 Published by Elsevier B.V.


doi:10.1016/S0926-860X(03)00553-2

drugs, and cosmetics. Moreover, d-sorbitol is an intermediate in Vitamin C production. Excellent settling
properties, a high activity, and low cost price justify
the choice for Raney-type Ni catalysts in this reaction. Although it is claimed that Ru/C catalysts are a
promising alternative because of their non-leaching
behavior and high activity [13], Raney-type Ni catalysts still dominate industrial d-sorbitol processing.
Even though this class of Ni catalysts has already
been applied in industry for decades, many properties

438

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

Nomenclature

d-glucose [2,47]. Main reasons for deactivation of


Raney-type catalysts are considered to be:

Ca
dp
D
Deff
k
kL a

loss of active Ni surface by sintering [4,5,7];


leaching of Ni and promoter metal into the acidic
and chelating reaction mixture [57];
poisoning of the active Ni surface by organic species
produced by side reactions [4,6].

M
SBET
Sh
SNi

Carberry number
catalyst particle diameter (m)
equimolar diffusion coefficient (m2 s1 )
effective diffusion coefficient (m2 s1 )
1
reaction rate constant (kg1
cat s )
volumetric gasliquid mass transfer
coefficient (s1 )
sum of all metals at the surface
catalyst surface area (m2 g1
cat )
Sherwood number
active Ni surface area (m2 g1
cat )

Greek letters

catalyst effectiveness factor

WheelerWeisz modulus (=i 2 )

of Raney-type Ni are not yet fully understood. Also,


characterization of these multi-component materials
is difficult, and many discrepancies in literature exist.
The key to reproducible synthesis can be found in absolute control of preparation of the starting materials,
the alloys, and the leaching process. A detailed study
of these processes, coupled with the determination of
the catalytic properties is then required. In this way,
the complete lifetime of a Raney-type Ni catalyst can
be followed in all its different stages. Therefore, the
relations between the structure, activity and stability
of the Raney-type Ni catalysts have been investigated in this work. The industrially attractive hydrogenation of d-glucose was used to establish these
relations.
It is known that Ni catalysts show deactivation after recycling of the catalyst in the hydrogenation of

Addition of Mo and Cr promoters was found to


be beneficial for the catalyst activity and stability in
this reaction [2,4,6,8,9], whereas Fe and Sn promoted
Raney-type Ni catalysts deactivate very rapidly [6].
Bizhanov et al. studied the promoting effect of Pt,
Pd, Ru and Rh on Raney-type nickel and found that
the activity was enhanced by about 2030% [10].
According to these authors Raney-type Ni/Ru was
the most promising catalyst. It was also established
that addition of the promoters did not lead to any
notable increase in the stability of the Raney-type
Ni catalysts [11]. The combined action of Cr and Fe
on Raney-type Ni is known to be effective in nitrile
hydrogenation [12], but is a novel system for the
hydrogenation of d-glucose.
Although Ni is responsible for the hydrogenation
capacity, it is suggested in the literature that residual
aluminum (in combination with promoters) plays a
key role in the performance of Raney-type Ni catalysts, because it is able to act as an electron donor
to nickel, which renders the d-band of nickel less
electron deficient and thus influences the adsorption
of the reacting species [13]. In the present work,
the influence of residual Al, and the effect of Mo,
and a combination of Cr and Fe promoters on the
performance of Raney-type Ni catalysts in the hydrogenation of d-glucose has been investigated. Special
attention has been paid to the stability of the catalysts in successive hydrogenation runs. Hydrogen

Fig. 1. Hydrogenation of d-glucose to d-sorbitol.

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

chemisorption, temperature-programmed desorption (TPD) and quasi in situ techniques like X-ray
diffraction (XRD), X-ray photoelectron spectroscopy
(XPS), and energy dispersive X-ray analysis (EDX)
were used to characterize the catalysts. The expression quasi in situ in this work implies that for the
characterization of the catalysts transport has been
performed under protective atmosphere. A preliminary report on d-glucose hydrogenation data has been
published previously [2].
2. Experimental
2.1. Catalysts
The applied commercial Raney-type Ni catalysts
were provided by Engelhard de Meern (The Netherlands). Unpromoted as well as promoted Raney-type
Ni catalysts were used in this study. The promoters
were Mo and a combination of Cr and Fe, the latter sample being a lab-sample. To vary the residual
Al content, two samples from different starting alloys
(Engelhard) were leached at mild conditions. Both alloys (A and B) have a 50/50 (w/w) Ni/Al composition
but were prepared via different proprietary processes.
The alloy powders (dp < 80 m) were exposed to an
excess of 20 wt.% NaOH. Leaching was performed at
353 K for 30 min. After leaching, the catalysts were
washed six times with distilled water. A detailed description of the leaching procedure has been given
elsewhere [14].
2.2. Ni and total surface area
The active Ni surface area was determined by means
of volumetric hydrogen chemisorption. The catalysts
were washed with ethanol (three times) to remove
most of the water. The wet Ni catalysts were dried at
323 K for 2 h in He followed by evacuation at 473 K
for 3 h to remove the hydrogen initially present on the
catalyst. After cooling down in vacuum the hydrogen
isotherms were measured at 323 K. The nickel surface
area was evaluated from extrapolation of the isotherm
representing strong adsorption to zero pressure (thus
corrected for weakly adsorbed hydrogen). The nickel
surface area was calculated under the assumption that
one Ni surface atom chemisorbs one hydrogen atom.

439

After the chemisorption measurement the BET surface was measured in the same equipment using N2
physisorption at 77 K. In between these steps contact
with air was avoided.
2.3. TPD
Desorption of residual hydrogen present on the
commercial catalysts (formed during the leaching
procedure) was studied by TPD in an atmospheric
plugflow reactor. The catalysts were weighed under
protective atmosphere (about 0.2 g of material was
used). The catalyst was introduced into the quartz
TPD reactor and a few droplets of ethanol were added
to the catalyst, creating a protective layer of liquid
around the catalyst particles. In this way, the reactor
could be transported to the TPD set-up without exposing the catalyst to air. After mounting the reactor in
the set-up, the sample was allowed to dry for a night
in an Ar flow of 0.42 ml s1 . Next, the dry sample
was heated according to a linear temperature-program
with a heating rate of 0.167 K s1 . In the reactor effluent the H2 -desorption was monitored with a thermal
conductivity detector (TCD).
2.4. Quasi in situ XPS
The Raney-type Ni catalyst samples were washed
with ethanol three times and dried in a glove box
under protective atmosphere. The dry catalyst was
pressed into a soft indium foil mounted on a flat sample holder. The sample holder was put into a transfer
vessel where the sample remains under the protective
atmosphere and subsequently the sample was transported into the vacuum chamber of the instrument
for XPS analysis. The XPS analysis was performed
with a PHI 5400 ESCA system equipped with a dual
anode X-ray source (Mg/Al) and a spherical capacitor analyzer (SCA). The energy scale of the SCA
was calibrated according to a procedure described in
[15]. The instrument was set at a constant analyzer
pass energy of 35.75 eV and unmonochromatized incident Al X-ray radiation (Al K1,2 = 1486.6 eV)
was used for excitation of the sample. The electrons
emitted from the sample were detected at an angle
of 45 with respect to the sample surface. An elliptic
area of 1.1 mm 1.6 mm was analyzed. Spectra were
recorded from the Ni 2p, Ni 3s, Al 2s (including Al

440

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

2p), Fe 3s, Cr 2p and Mo 3d photoelectron lines with


a step size of 0.2 eV. The Ni 2p3/2 binding energy of
metallic Ni (852.80 eV) was used as an internal reference. The spectra shown in this work were corrected
for satellites caused by the non-monochromatic nature
of the incident X-ray source [16] and the background
was subtracted from the spectra using the iterated
Shirley-method [17]. The surface composition of the
catalyst was determined from the integrated intensity
of the Ni 3s, Al 2s, Fe 3s, Cr 2p1/2 , and Mo 3d5/2
photoelectron lines by adopting the elemental sensitivity factors [16] of the Multipak software package.
It should be noted that in this calculation the effective
probe depth, which correlates closely to the inelastic mean free path of the electrons, is accounted for.
Since the Multipak software assumes the sample to be
homogeneous over the volume analyzed, the analysis
results in an average composition of the surface phase,
which has a thickness of about 1.5 nm. Consequently,
the amount of thin oxidic layers of components present
at the surface of the catalyst will be underestimated
in the calculated results, whereas the occurrence of
massive components as metallic Ni an Al will be
overestimated.
2.5. Quasi in situ XRD
The catalyst powder was introduced into a thin glass
capillary under protective atmosphere in a glove box.
The capillary was sealed and in this way it could be
transported to the diffractometer and analyzed without exposure to air. Powder XRD was performed in
a Bruker-Nonius D5005 diffractometer using Cu K
radiation.
2.6. Quasi in situ HRTEM coupled with EDX
High-resolution transmission electron microscopy
(HRTEM) was performed using a Philips CM30T
equipped with a LaB6 filament, operated at 300 kV.
Elemental analysis was performed using a Link EDX
system. Samples were applied on a micro grid carbon
polymer supported on a copper grid by placing a few
droplets of a suspension of the sample in n-hexane on
the grid, followed by drying at ambient conditions,
all in an Ar glove box. Samples were transferred to
the microscope in a special vacuum-transfer sample
holder under exclusion of air [18].

2.7. Hydrogenation reactions


The catalysts were screened in a 300 ml three-phase
slurry reactor with a gas-induced stirrer (Premex AG)
at 393 K and 4.0 MPa using 10% d-glucose (0.56 M)
in Milli-Q water. A low concentration of d-glucose
was used in order to minimize mass transfer limitations, which easily occur in this fast reaction [19]. The
Raney-type catalysts were weighed as dry material in
a glove box under protective atmosphere. The empty
reactor vessel was transferred into the glove box and
the desired amount of catalyst was introduced in the
reactor. To prevent oxidation during transportation
of the vessel to the experimental set-up, 100 ml of
solvent was poured on the catalyst. Above the reactor an injection vessel was situated. The reactor
and injection vessels were purged with N2 and H2
at room temperature. When the solvent with catalyst
was at the desired temperature in the reactor (at about
1.0 MPa), the injection vessel, containing 100 ml of
concentrated d-glucose solution, was opened. The
exact starting time of the reaction could in this way
be assured. The temperature, total pressure and hydrogen consumption were recorded during reaction.
The starting pH of the reactant solution was about
6. High performance liquid chromatography (HPLC)
was used to analyze samples of the product mixture.
The HPLC was equipped with a refractive index (RI)
detector and a RCM monosaccharide column operated at 358 K to separate the components. To study
the stability of the catalysts, several consecutive hydrogenation runs were performed. After each run the
reactor was emptied through a filter (5 nm) under
hydrogen atmosphere. The catalyst was washed in
situ with 150 ml water and a fresh reaction mixture
was introduced for the following run. In this way,
it was assured that the catalyst was not exposed to
air.
2.8. Leaching of the catalysts
The amount of metal ions present in the reaction
mixture after the hydrogenation run was determined
by means of inductively coupled plasma-optical emission spectroscopy (ICP-OES). The analysis was carried out with a Perkin-Elmer 5100PC apparatus. A
10 wt.% d-sorbitol solution was used as a matrix in
the calibration samples.

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

3. Results and discussion


3.1. Bulk composition of the NiAl alloys and
Raney-type Ni catalysts
Fig. 2 shows the XRD patterns of the alloy samples.
The two starting alloys A and B, used for the preparation of the homemade catalysts, contain the same type
of phases (Ni2 Al3 , NiAl3 and pure Al), but in different
concentrations. Although quantification of the phase
distribution within the alloy has not been performed,
Fig. 2 indicates that the Ni2 Al3 phase is predominant
in both alloys. More Ni2 Al3 and less metallic Al is
present in alloy B in comparison with alloy A. The
Ni2 Al3 ratio of the two alloys could be estimated from
the intensity ratio of the {1 0 1} and {0 0 1} reflections:
(Ni2 Al3 )B-alloy
= 1.2
(Ni2 Al3 )A-alloy
This suggests that the A-type alloy can be transformed
into activated catalyst more easily than the B-type
alloy, because Ni2 Al3 is reported to be the most difficult phase to leach [20]. A previous electron microscopy study of the alloys showed that both alloys
have a kernel of NiAl3 , surrounded by Ni2 Al3 [14].
Fig. 3 shows the XRD patterns of the resulting homemade catalysts and the commercial samples. The XRD
patterns of the leached alloys (referred to as RaNi-A
and RaNi-B) show, apart from elemental Ni, traces of
bayerite, Al(OH)3 . The presence of metallic Ni indi-

441

cates that the alloys have been activated relatively fast


at the applied leaching conditions, since the original
NiAl phases are not detected, but traces of intermediates like bayerite point out that the leaching process was not yet in its the final state. Formation of
bayerite is unwanted because its presence could cause
pore blocking. Composition analysis of the homemade
catalysts with EDX supports the suggestion that alloy B was more difficult to leach due to the larger
amount of Ni2 Al3 : after 30 min leaching the B-catalyst
has a larger amount of residual Al than the A-catalyst
(Table 1). It has been shown that at the applied conditions already after 10 s of leaching, the leaching of aluminum from the alloy has finished and the Al is present
as alumina [14]. Extended leaching time is required to
completely dissolve the aluminum oxide species and
optimize the catalytic performance. This phenomenon
is more pronounced for the B catalyst compared to
the A catalyst: due to the lower leaching rate, more
bayerite is present for RaNi-B after 30 min of leaching
as was shown with XRD (Fig. 3). For the commercial samples, only the Ni reflections can be observed
in XRD. The absence of Al reflections could indicate
that the residual Al consists of small particles undetectable using XRD. A more plausible explanation is
the existence of a solid solution of Al in the Ni lattice.
If that were the case, the lattice parameters would be
larger than bulk Ni, leading to a shift to lower 2 values in the XRD diagram. Fig. 3 shows that indeed the
lattice parameter of the Ni is enlarged by the presence
of Al, indicated by the small shift of the Ni reflections

Fig. 2. X-ray diffraction patterns of the starting alloys: () Ni2 Al3 , () Al; the remaining peaks are NiAl3 reflections.

442

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

Fig. 3. X-ray diffraction patterns of the catalysts: () Ni, () Al(OH)3 . The dashed lines represent the Ni {1 1 1} and {2 2 0} reflections
of bulk Ni.

of the catalyst samples compared to bulk Ni, which


position is indicated by the dashed lines [21].

separation (17.2 eV), which is typical for metallic


nickel [16]. The positions of the Ni 2p3/2 emission
lines of all catalysts indicate the presence of Ni0 , but
also other Ni species can be observed. The emission
line at 854.6 eV corresponds to NiO, with a doublet
separation of about 18.2 eV [22]. For the homemade
catalysts, a third Ni contribution is observed at a
binding energy of 858.8 eV. In literature, Ni2 O3 [23],
Ni(OH)2 [24] or NiAl2 O4 [24] have been claimed
to be present at the surface of Raney-type Ni catalysts. Both Ni2 O3 and Ni(OH)2 compounds have a Ni
2p3/2 binding energy at about 856.0 eV [25] and the
Ni 2p3/2 binding energy of NiAl2 O4 , is reported to
be 857.0 eV [16]. Thus, the photoelectron contribution at 858.8 eV does not correspond with one of the
aforementioned phases. However, this high binding

3.2. Surface composition of the Raney-type Ni


catalysts
The nature and the occurrence of the different
phases at the catalyst surface can be evaluated from
the position of the XPS emission lines and their (relative) intensities. Fig. 4 shows the XPS spectra of the
Ni 2p region of the Raney-type Ni catalysts and a
reference material, which is a single crystal of NiAl
(-NiAl), cleaned in situ by sputtering with Ar. The
position of the Ni 2p3/2 emission line of the single
crystal is found at 852.8 eV, corresponding to metallic nickel (Ni0 ). This is confirmed by the doublet
Table 1
Metal distribution of the catalysts (at.%)
Catalyst

RaNi
RaNi-Mo
RaNi-Cr/Fe
RaNi-A
RaNi-B

Bulk (from EDX)

Surface (from XPS)

Ni

Al

Mo

Cr

Fe

Ni

Al

Mo

Cr

Fe

84
79
83
87
74

16
19
11
13
26

1.4

2.5

3.6

69
66
57
60
62

31
25
26
40
38

8.7

3.8

5.9

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

Fig. 4. Ni 2p spectra of the various samples of Raney-type Ni and


a -NiAl single crystal.

energy suggests that it is related to a Ni2+ containing


species (referred to as NiAlO), which strongly interacts with its environment. Fig. 5 gives some more
details. At the high binding energy side in the Al 2s
region of the homemade catalysts, a large contribution can be seen at 122 eV. Because the spectrum of
-NiAl shows that the Al 2s line of metallic Al can
be found at 117.6 eV, and aluminum oxides and hydroxides can be found around 119 eV [26], the high
Al 2s binding energy contribution can be assigned
to Al3+ in a strong ligand field. All catalyst samples have a contribution of metallic Al (117.6 eV).
Deconvolution of the Al 2s region shows that RaNi
contains two types of aluminum oxides (most likely
Al(OH)3 and Al2 O3 ). In contrast, XPS analysis of the
promoted RaNi catalysts shows that only one type of
aluminum oxide is present. The Al 2s binding energy
of 119 eV suggests that this species is rather an aluminum hydroxide than an oxide, because aluminum
oxides have a higher Al 2s binding energy than hydroxides [26]. The overlapping of the Al contributions in the spectra of the homemade catalysts does
not allow an unambiguous identification of all the
phases.

443

Fig. 5. Al 2s and Ni 3s spectra of the various samples of Raney-type


Ni and a -NiAl single crystal.

The Mo in the Mo promoted RaNi is mainly oxidic,


although some metallic Mo is present. Hamar-Thibault
and Masson studied the RaNi-Mo system and concluded that molybdenum was in an oxidized state of
the type MoO2 corresponding with a Mo 3d5/2 binding
energy of 229.9 eV [27]. In the present study, the Mo
3d5/2 binding energy was found at 232.0 eV, making
MoO3 more likely to be present at the catalyst surface
than MoO2 [16]. The Cr 2p3/2 line at 577.2 eV corresponds to Cr2 O3 [28]. Because the Fe 2p lines overlap
with Ni Auger lines, the Fe 3s emission line has been
used in this study. From its binding energy position
(93.5 eV), it can be concluded that iron is present as
an oxide (most likely Fe2 O3 ) rather than as metallic
Fe (91.3 eV) [16].
The surface composition of the catalysts has been
estimated from the recorded XPS spectra and is reported in Table 2. Note that in the current work it is
not the aim to establish the dispersion of the various
phases present in the catalyst, but to acquire information of the relative amount of the various species.
Since the software assumes the sample to be homogeneous over the volume analyzed, the analysis results
in an average composition of the surface phase with a
thickness of about 1.5 nm. Consequently, the presence

444

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

Table 2
Phase distribution at the catalyst surface (at.%) from XPS

Ni0
NiO
NiAlO
Al0
Al2 O3
AlO(OH)
Mo0
MoO3
Cr2 O3
Fe2 O3
a

RaNi

RaNi-Mo

RaNi-Cr/Fe

RaNi-A

RaNi-B

48.4
20.3
0
12.4
10.8
8.2

57.1
8.8
0
17.0

8.2
0.8
7.9

57.1
7.1
0
17.9

8.3

3.8
5.9

19
18
23

15
16
31

40a

38a

This number represents the sum of all Al species.

of thin oxidic layers at the surface will be underestimated in the calculated results, whereas the occurrence of massive components as metallic Ni an Al
will be overestimated, depending on the thickness of
the oxide layers. At the surface of the RaNi catalyst a
relatively large amount of oxides is present, both NiO
and aluminum oxides. The presence of promoters apparently leads to suppression of the formation of NiO
and Al2 O3 , leaving larger amounts of metallic Ni and
Al at the surface.
Table 1 shows that the surface of the catalysts is
enriched with Al. Also, the promoters are segregated
at the surface of the catalyst at the expense of Al.
The total amount of Ni at the surface of the commercial samples is comparable. The amount of aluminum
species at the surface of the homemade catalysts is
large compared to the commercial RaNi catalysts, as
was expected since the leaching time was relatively
short. Although the bulk compositions of the homemade catalysts differ significantly in Ni/Al ratio, the
surface composition is comparable.

3.3. Surface morphology of the Raney-type Ni


catalysts
Table 3 gives a summary of the characteristics of
the studied catalysts. The amount of metallic Ni at the
catalyst surface as determined by XPS for the Ni-only
materials follows the trend of the active Ni surface
area. The homemade catalysts have a lower Ni surface area as measured by hydrogen chemisorption than
the commercial RaNi due to incomplete leaching, resulting in a relatively large amount of Al at the surface. Also, the strongly interacting third Ni species,
found in the homemade samples, might explain the
relatively small amount of metallic Ni. Thus, for the
unpromoted catalyst systems the two applied methods
to estimate the fraction Ni at the catalyst surface, i.e.
hydrogen chemisorption and XPS, give the expected
trends. However, the presence of promoters gives contradictory results for the determination of the Ni surface fraction with the different techniques. Saliently,
the surface area of active material as determined by
hydrogen chemisorption (SNi ) is at most 34% of the
total surface area (SBET ). Addition of Mo results in
an increased total surface area, but SNi is comparable
to that of unpromoted RaNi. Promotion with Cr and
Fe lowers SNi dramatically, but gives a more porous
structure to the catalyst, since the SBET increases with
a factor 2. These observations are remarkable because
a higher BET surface area is likely to correspond to a
higher Ni surface since the catalysts consist essentially
of Ni (85%). Possible explanations are (i) the segregation of the promoter atoms to the nickel surface,
thus hindering hydrogen chemisorption (Table 1) and
(ii) an underestimation of the hydrogen chemisorption
surface area due to the presence of strongly bound
hydrogen formed during catalyst preparation, which

Table 3
Properties of the Raney-type Ni catalysts
Catalyst

a
SNi (m2 g1
cat )

b
SBET (m2 g1
cat )

SNi /SBET

(Ni0 /M)surf c

Activity (kg1 s1 )

RaNi
RaNi-A
RaNi-B
RaNi-Mo
RaNi-Cr/Fe

19
10
7
21
10

56
56
62
77
112

0.34
0.18
0.11
0.27
0.09

0.48
0.19
0.15
0.57
0.57

0.35
0.16
0.14
0.50
0.90

Determined from hydrogen chemisorption.


Determined from nitrogen physisorption.
c Determined from XPS, M is the sum of all metals at the surface.
b

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

445

Fig. 6. TPD profiles of the commercial Raney-type Ni catalysts. The profiles are normalized for the amount of sample used.

could not be removed during the catalyst pretreatment in the chemisorption experiment. To resolve this,
TPD was performed. Fig. 6 shows the TPD profiles
of the commercial catalysts. Martin et al. [29] demonstrated, by measuring the saturation magnetization of
Raney-type Ni in an electromagnetic field, that the
evolved hydrogen during TPD cannot be the result of
the reaction of water with metallic Al, as was proposed
by Mars and coworkers [30]. Thus, the hydrogen detected during the TPD, originates from adsorbed hydrogen on the Ni, formed during preparation of the
catalysts. The profiles of the promoted catalysts show
one large peak at about 460 K, while the unpromoted
RaNi has its largest contribution peaking at 435 K, and
smaller peaks at 480 and 525 K. These small peaks
could be related to a different type of Ni species, but
additional experiments are required to understand this
phenomenon. The Cr/Fe promoted catalysts releases
the largest amount of hydrogen. Possibly, the applied
temperature and time (473 K, 2 h) to evacuate the samples in the determination of the nickel surface area by
hydrogen chemisorption was insufficient to remove all
the strongly bonded hydrogen initially present on the
catalyst. Consequently, the hydrogen chemisorption
results in an underestimation of the availability of the
Ni sites, most pronounced for the RaNi-Cr/Fe sample.
3.4. Performance of the Raney-type catalysts in the
hydrogenation of d-glucosemass transfer effects
Preliminary measurements were performed with
the Raney-type Ni catalysts to study whether mass

transfer or the chemical reaction controls the overall


reaction rate. To study the impact of gasliquid mass
transfer, the stirrer speed was varied between 1600
and 2500 rpm. The difference in reaction rate constant
was less than 5%, indicating the absence of gasliquid
mass transfer limitations. In addition, previous experiments in the applied three-phase batch reactor showed
that gasliquid mass transfer coefficient (kL a) values
up to 0.13 s1 could be reached in a 10% d-glucose
solution at reaction conditions [19]. Knowing this
value, it is possible to calculate the ratio between the
observed reaction rate and the gas to liquid maximum transfer rate, defined as the Carberry number,
CaGL , and the utilization of the catalyst considering
gasliquid mass transfer, GL [31]. In the calculations, the initial reaction rate has been considered
in order to deal with a worst-case scenario. Table 4
shows that in all cases, the effectiveness factor GL is
close to unity, confirming that gasliquid limitations
Table 4
Evaluation of the presence of H2 transport limitationsa
Gasliquid

RaNi
RaNi-Mo
RaNi-Cr/Fe
RaNi-A
RaNi-B

Liquidsolid

Internal

Ca

Ca

0.038
0.056
0.062
0.023
0.015

0.96
0.94
0.94
0.98
0.99

0.020
0.029
0.032
0.011
0.008

0.98
0.97
0.97
0.99
0.99

0.232
0.338
0.377
0.136
0.092

0.87
0.82
0.78
0.92
0.95

a Reaction conditions: 10 wt.% d-glucose, 4.0 MPa, 393 K,


0.9 gcat .

446

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

can be neglected. The liquidsolid mass transfer can


be estimated by means of the dimensionless Sherwood number, Sh [31]. The effectiveness factor for
liquidsolid mass transfer LS is above 0.95.
The absence of external mass transfer does not rule
out diffusion limitations in the pores of the Raney-type
catalysts. The average particle diameter of the applied
catalysts is in the order of 40 m. The WheelerWeisz
criterion was used to evaluate the absence of internal
diffusion limitations based on experimental data. The
effective diffusivity Deff was estimated to be 0.1D.
For d-glucose no pore diffusion limitations occurred,
but in most cases the reaction is partly controlled by
internal mass transfer limitation of hydrogen. Only
the experiment performed with RaNi-B is completely
governed by kinetics, due to the low activity of this
catalyst, as will be discussed further on. To check the
reproducibility of the experiments, the RaNi-Mo catalyst was tested three times. The reaction rate was reproduced with an error of only 2%.
3.5. Performance of the Raney-type catalysts in the
hydrogenation of d-glucoseactivity
All catalysts show a high selectivity to d-sorbitol
(>99%). Main by-products during reaction are
d-fructose and d-mannitol. d-Fructose is formed due
to isomerization of d-glucose. d-Mannitol is produced by hydrogenation of d-fructose (also yielding
d-sorbitol). The catalytic hydrogenation of d-glucose
can usually be described by LangmuirHinshelwood
kinetics with a transition from zero-order dependency
in d-glucose at high concentrations to first-order behavior at low concentrations [19]. At the applied
d-glucose concentrations (10 wt.%) in this study, the
reaction can be described by a first-order dependency
in d-glucose for all catalysts [2]. The activity of the
catalysts can therefore be expressed by the pseudo
1
first-order reaction rate constants (kg1
cat s ) and is
given in Table 3. The mildly leached RaNi samples
have lower activity than the commercial RaNi catalyst, in line with the XPS analysis and hydrogen
chemisorption data, from which it was established
that the Ni surface area is smaller for the mildly
leached catalysts. RaNi-A contains more metallic
Ni after 30 min leaching and less of the intermediate NiAlO species due to lower amount of initial,
less reactive, Ni2 Al3 in comparison with RaNi-B.

Although RaNi-B has a significantly different bulk


composition than RaNi-A (Table 1), the activity data
show that the surface properties determine the catalytic performance. Unfortunately, the role of metallic
Al remains unknown since its surface quantity could
not be determined by means of XPS. The promoted
catalysts show higher reaction rates than unpromoted
RaNi. The Cr/Fe promoted Raney-type Ni performs
better than the RaNi-Mo catalyst, which is the traditional catalyst used in this process. To understand
the role of the promoters on the catalytic activity, the
results have been expressed per unit surface area as
determined by hydrogen chemisorption (SNi ), nitrogen physisorption (SBET ) and a combination of SBET
and XPS (Fig. 7). It can be seen that the activity of
the commercial Raney-type catalysts correlates with
the specific surface area as determined by nitrogen
physisorption. This result suggests than the main
role of the promoters is to increase and stabilize the
specific surface area of the catalyst. Based on the
chemisorption data, however, it can be seen that the
enhanced BET surface area of the promoted catalysts
cannot fully account for the increase in activity. Especially, the activity of the Cr/Fe promoted Raney-type
Ni catalyst does not correlate with SNi . Therefore,
the following working hypothesis is proposed. The
promoters are present at the surface in the form of
oxide rafts or patches [32]. At the edges of the oxidic
rafts, a positive charge is induced on the neighboring
Ni, thus enhancing the adsorption of d-glucose. The
hydrogen adsorbs dissociatively on the Ni sites and
migrates by a spill over effect to the Ni sites with the
adsorbed d-glucose molecules, initiating the reaction.
For the Cr and Fe promoted catalysts it is difficult to
conclude on their role as promoter [6].
3.6. Performance of the Raney-type catalysts
in the hydrogenation of d-glucosestability
Stability of catalyst performance is of extreme
importance in industrial d-sorbitol processing. The
catalysts should be resistant to metal sintering and
poisoning. Moreover, due to formation of small
amounts of d-gluconic acid during the reaction, which
is a strong chelating agent, metal atoms tend to leach
from the catalyst surface into the liquid phase. This
phenomenon is highly undesirable, firstly because of
the resulting metal impurities in the product, which

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

447

Fig. 7. Activity of the Raney-type catalysts expressed per unit surface area as determined by hydrogen chemisorption, nitrogen physisorption
(SBET ) and a combination of SBET and XPS (393 K, 4.0 MPa, 10 wt.% d-glucose).

renders product purification necessary (d-sorbitol is


used in food and personal care products) and, secondly, loss of active catalyst sites. The amount of
dissolved metal ions in the reaction mixture after
each hydrogenation run was determined by ICP-OES.
Table 5 shows the results, expressed as the loss of
original metal content (wt.%). About 0.8 wt.% Ni has
leached from the catalyst surface for all Raney-type
Ni catalysts (ca. 350 mg kg1
sorbitol ). Mo does not leach
at all, within limits of accuracy, while Al and Cr
leach a little, comparable with Ni. Fe dissolves quite
easily in the reaction mixture. This loss of Fe has as
consequence that also Ni and Al leach more severely
as compared to RaNi. RaNi-Mo is in this perspective
the most stable catalyst; Mo does not leach and both
Ni as well as Al leaching is reduced in the presence
of Mo. This is in agreement with work of Gallezot
et al. [6]. They found that iron promoted Raney-type
Ni showed deactivation after successive runs due to
leaching of the iron from the Ni surface. Leaching of

molybdenum or chromium was less pronounced. The


combined effect of Cr and Fe was not studied.
The loss of Ni from the catalyst into the reaction
mixture is an indication for the catalysts structural
stability. To test the catalyst performance stability, successive experiments were carried out. As an example,
the activity of successive recycle runs of the commercial RaNi sample is shown in Fig. 8. The successive
runs show a steady decrease in activity. After three
consecutive runs 48% of the initial activity is lost.
Promoting Raney-type Ni gives initially a better catalyst stability (Table 6). The activity of the Raney-type
Ni-Mo catalyst has lost 30% of the initial activity after three runs and RaNi-Cr/Fe 16%. Hence, although
the promoters make Raney-Ni catalysts more stable,
still a significant loss in activity after a few successive
runs is observed. van Gorp et al. also studied the stability of the commercial RaNi-Mo and observed a decrease in activity of about 17% after three runs, which
is much less than the present results [3]. The tests were

Table 5
Amount of leached metal ions (wt.%)a

Table 6
Activity of the catalysts after successive recycles (kglucose ,
kg1 s1 )

RaNi
RaNi-Mo
RaNi-Cr/Fe
RaNi-A
RaNi-B
a

Ni

Al

Mo

Fe

Cr

0.8
0.6
1.1
0.9
0.7

0.8
0.5
1.5
0.5
0.1

0.0

27.7

1.2

Relative to the composition of the starting catalyst material.

RaNi
Run 1
Run 3
Run 5
Loss run (1 3) (%)
Loss run (3 5) (%)

0.35
0.18

48

RaNi-Mo
0.46
0.32
0.25
30
21

RaNi-Cr/Fe
1.02
0.86
0.61
16
29

448

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

Fig. 8. Conversion of d-glucose as function of time in three successive hydrogenation runs for RaNi.

performed at identical temperature and hydrogen pressure, but with a 2.78 M d-glucose solution instead of
the present 0.56 M solution. A first-order reaction rate
1 was found, which is 20 times lower
of 0.025 kg1
cat s
than the value found in this work. Most likely, these
data were disguised by mass transfer limitations of
hydrogen, which masked the deactivation. The question arises what is causing the decrease in activity.
Fig. 9 shows the loss in activity of the Cr/Fe promoted
Raney-type Ni during the stability tests as a function
of iron content in the catalyst. Initially a large amount
of Fe is lost (almost 30%), whereas only 10% loss in
activity is observed. After run 3, the decrease in hydrogenation activity is proportional to the decrease in
Fe content. Apparently, about 30% of the Fe is present

Fig. 9. Relation between leached Fe and catalytic activity after each


hydrogenation run relative to starting conditions. The pre-leached
sample is indicated with ().

as catalytically inactive, cluster material. This result


supports the proposed mechanism of the promoting
effect of Fe. The promoting Fe is relatively strongly
interacting with the matrix of the catalyst, whereas
less interacting bulk-like Fe-species are catalytically
inactive. Also Table 6 shows that the deactivation is
accelerated once the bulk Fe has been removed. This
shows that although chromium is expected to have a
stronger promoting effect, the importance of the presence of iron at the surface in this system cannot be disregarded. To further investigate the influence of the Fe
content on the activity, also a pre-leached Raney-type
Ni-Cr/Fe catalyst was prepared by treatment in 0.01 M
HCl under reflux conditions. Whereas after this treatment hardly any Cr, Ni, and Al was leached, about
14% of Fe was removed from the catalyst. The activity
was only slightly lowered (see Fig. 9), in agreement
with the stability experiment. Hence, for the Cr/Fe
promoted catalyst leaching of Fe could be causing the
deactivation. However, also the bare Raney-type Ni
and Mo promoted catalyst show deactivation, so Fe
leaching is not the only reason for the loss in activity.
Other possible explanations of deactivation are metal
sintering or poisoning of the active sites. The Ni surface area of the Mo promoted RaNi was therefore examined before and after reaction (Fig. 10). The BET
surface area remains intact after repeated use, indicating that significant sintering is not occurring. Apparently, the active Ni sites are poisoned by strongly
adsorbed species. The hydrogen chemisorption results
show that after run 5, a large part of the Ni surface

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

449

Fig. 10. Ni surface area from hydrogen chemisorption and total surface area for nitrogen physisorption as a function of the condition of
the RaNi-Mo catalyst.

area is not accessible for hydrogen anymore. Remarkably, the decrease in Ni surface area (85%) is not proportional to the loss of activity (46%), which suggests
that not all sites are identical in terms of activity. One

explanation could be that there are Ni clusters in close


vicinity to promoter metal and Ni-only clusters, more
sensitive to poisoning. Regeneration by reduction for
2 h at 573 K and consequent hydrogen chemisorption

Fig. 11. Formation of d-gluconic acid via transfer hydrogenation.

450

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

Fig. 12. Influence of initial d-gluconic acid on the d-glucose hydrogenation rate of RaNi and RaNi-Mo (0.56 M d-glucose, 393 K and
4.0 MPa). The black bars represent glucose only and the gray bars with additional d-gluconic acid (0.1 M).

shows that most of the original Ni surface area is regained (Fig. 10); the adsorbed species have been removed resulting in a clean surface. The compound
most likely to be poisonous is d-gluconic acid, formed
via a Cannizzaro-type of reaction [8]. This reaction
is an oxido-reduction, induced by an alkaline environment in the presence of nickel and leads to the
formation of a mixture of d-gluconic acid (50%),
d-sorbitol, and d-mannitol (Fig. 11) [33]. This dehydrogenation reaction involves transfer hydrogenation
with d-glucose as hydrogen donating agent [33,34].
It is known that carboxylic acids have a strong tendency to bind to Ni. The effect of d-gluconic acid formation is suspected to be two-fold: (i) the acidity of the
reaction mixture increases, resulting in dissolution of
metals; (ii) blocking of active sites by stronger adsorption. To prove this hypothesis, experiments were carried out in which d-gluconic acid (0.1 M) was added
to the reaction mixture. Fig. 12 shows the results for
both RaNi and RaNi-Mo. The activities of d-glucose
hydrogenation have dramatically decreased for both
types of catalyst. The product mixture was completely
green, indicating large amounts of dissolved Ni. This
was confirmed by ICP-OES: almost 20% Ni and 3% Al
of the original catalyst had dissolved into the product
mixture. The BET surface area of the spent RaNi-Mo
catalyst was examined and showed that the area was
decreased from 77 to 53 m2 g1 due to the presence
of d-gluconic acid. Thus, although large amounts of

Ni and Al were leached, a large part of the original


porous structure has not been destroyed. It can therefore be concluded that blocking of the active sites by
d-gluconic acid is the main cause for the loss in activity. Crezee et al. showed that also Ru/C catalysts
deactivate due to poisoning of active species, although
to a lesser extent [35]. The authors were able to regenerate part of the catalyst activity by treatment of
the catalyst in vacuum thereby removing the adsorbed
products. Apparently, for RaNi additional hydrogenation at elevated temperature is required. Industrially,
this procedure can be performed, but the regeneration
must be preceded by a very severe washing procedure.
The regeneration treatment is not possible in the presence of d-glucose, because the high temperature of
the regeneration step will cause formation of polysaccharides, which will strongly adsorb on the catalyst,
resulting in even further deactivation, as was observed
experimentally.
In summary, it can be concluded that poisoning of
the active sites by d-gluconic acid is the main cause
of deactivation of (promoted) Raney-type nickel in the
hydrogenation of d-glucose. d-Gluconic acid is also
responsible for leaching of Ni into the product mixture. It can be expected that the loss of Ni and Al after
long-term processing eventually leads to a decrease in
surface area resulting in loss of activity. The Cr/Fe promoted system loses Fe, which also contributes to the
deactivation. Considering the constant loss of both Ni

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

and Al, this could eventually lead to significant loss of


surface area in long-term industrial applications, especially when more concentrated d-glucose solutions
are used. It should, however, be noted that in industrial d-sorbitol processing higher hydrogen pressures
(typically 20 MPa) are applied [36], possibly advantageous for the stability of the catalyst with respect to
poisoning. Buffering the reaction mixture to keep the
pH neutral may minimize leaching of metals into the
product, while maintaining reasonable selectivity.

451

catalyst surface, but only once the promoting, more


refractory, Fe has been leached into the product mixture the activity in d-glucose hydrogenation decreases
significantly. The major cause of deactivation of the
commercial Raney-type Ni catalysts is the presence of
d-gluconic acid formed during the reaction. The effect
of d-gluconic acid is two-fold: firstly, it poisons the Ni
sites making them inactive in d-glucose processing.
The sites can be recovered again by a regeneration
treatment in hydrogen at 393 K. The formation of
d-gluconic acid also leads to a relatively large loss of
Ni, which is highly undesired in d-sorbitol processing.

4. Conclusions
The phase composition of the starting alloys in
Raney-type Ni manufacturing is highly important, as
was demonstrated by using relatively short digestion
times in the leaching process of Al from the NiAl
alloy. The alloy with a high initial amount of Ni2 Al3
gives an Al-rich catalyst after leaching. But independent of the starting alloy, the surface properties are
comparable. The homemade catalysts have a low Ni
surface area and contain remains of bayerite, formed
during the leaching, as compared to commercial RaNi.
As a result, these catalysts are less active in d-glucose
hydrogenation. Metallic Ni and Al are present at the
surface of all examined catalysts. The surfaces of Mo
and Cr/Fe promoted catalysts contain a small amount
of Ni and Al oxides, but the promoters are essentially
present as oxides. The promoters and Al are segregated to the surface in the form of their oxides, which
results in a lower chemisorption area as compared to
SBET . Furthermore, the evacuation temperature and
time (473 K, 2 h) as applied for the chemisorption experiments might be insufficient to remove the strongly
bonded hydrogen initially present on the catalyst.
All catalysts show high selectivity towards d-sorbitol
(>99%). The main role of the promoters is to enhance
and stabilize the BET surface area. The activity of the
commercial Raney-type of catalysts is mainly dependent on their specific surface area. Additionally, there
is a promoting effect based on the enhanced interaction of d-glucose with Ni due to neighboring oxidic
promoter metal. The Raney-type Ni catalysts lose Ni
and Al at the applied reaction conditions. Mo does
not leach at all. About 30% of the Fe at the surface
of the examined RaNi-Cr/Fe catalyst is present as
inactive, bulk iron. This Fe leaches severely from the

Acknowledgements
This work has been supported by The Netherlands
Foundation for Technical Research (STW), Engelhard de Meern, and DSM Research. Johan Groen
(TUDelft) is gratefully acknowledged for performing
the hydrogen chemisorption and nitrogen physisorption measurement, and Niek van der Pers (TUDelft)
for the XRD analysis. We also thank Michiel Makkee
(TUDelft) for the fruitful discussions about d-glucose
hydrogenation.

References
[1] P. Gallezot, N. Nicolaus, G. Flche, P. Fuertes, A. Perrard,
J. Catal. 180 (1998) 51.
[2] B.W. Hoffer, E. Crezee, P.R.M. Mooijman, A.D. van
Langeveld, F. Kapteijn, J.A. Moulijn, Catal. Today 7980
(2003) 35.
[3] K. van Gorp, E. Boerman, C.V. Cavenaghi, P.H. Berben,
Catal. Today 52 (1999) 349.
[4] P.J. Cerino, G. Flche, P. Gallezot, J.P. Salome, in: M. Guisnet,
J. Barrault, C. Bouchoule (Eds.), Heterogeneous Catalysis
and Fine Chemicals II, Elsevier, Amsterdam, 1991, p. 231.
[5] B. Kusserow, S. Schimpf, P. Claus, Adv. Synth. Catal. 345
(2003) 289.
[6] P. Gallezot, P.J. Cerino, B. Blanc, G. Flche, P. Fuertes, J.
Catal. 146 (1994) 93.
[7] J.-P. Mikkola, H. Vaino, T. Salmi, R. Sjoholm, T. Ollonqvist,
J. Vayrynen, Appl. Catal. A 196 (2000) 143.
[8] R. Albert, A. Strtz, G. Vollheim, Chem. Eng. Technol. 52
(1980) 582.
[9] T. Koscielski, J.M. Bonnier, J.P. Damon, J. Masson, Appl.
Catal. 49 (1989) 91.
[10] F.B. Bizhanov, D.V. Sokolskii, N.I. Popov, N.Y. Malkhina,
A.M. Khisametdinov, Kinet. Catal. 10 (1969) 655.

452

B.W. Hoffer et al. / Applied Catalysis A: General 253 (2003) 437452

[11] F.B. Bizhanov, D.V. Sokolskii, N.I. Popov, N.Y. Malkhina,


A.M. Khisametdinov, Kinet. Catal. 8 (1967) 529.
[12] S.B. Ziemecki, in: M. Guisnet, J. Barbier, J. Barrault (Eds.),
Heterogeneous Catalysis and Fine Chemicals III, Elsevier,
Amsterdam, 1993, p. 283.
[13] P. Fouilloux, Appl. Catal. 8 (1983) 1.
[14] F. Devred, B.W. Hoffer, A.D. van Langeveld, P.J. Kooyman,
H.W. Zandbergen, Appl. Catal. A. 244 (2003) 291.
[15] ASTM E902-88, Surf. Interface Anal. 17 (1991) 4709.
[16] J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben,
Handbook of X-ray Photoelectron Spectroscopy, PerkinElmer, Wellesley, 1992.
[17] D.A. Shirley, Phys. Rev. B 5 (1972) 4709.
[18] H.W. Zandbergen, P.J. Kooyman, A.D. van Langeveld, in:
H.A. Calderon, M.J. Yacaman (Eds.), Electron Microscopy,
second ed., Institute of Physics, 1998, p. 491.
[19] E. Crezee, B.W. Hoffer, R.J. Berger, M. Makkee, F. Kapteijn,
J.A. Moulijn, Appl. Catal. A, in press.
[20] M.L. Bakker, D.J. Young, M.S. Wainwright, J. Mater. Sci.
23 (1988) 3921.
[21] W.B. Pearson, A Handbook of Lattice Spacings and Structures
of Metals and Alloys, Pergamon Press, London, 1958.
[22] K. Kishi, J. Electron. Spectrosc. Relat. Phenom. 46 (1988)
237.
[23] T. Yoshino, T. Abe, I. Nakabayashi, J. Catal. 118 (1989) 436.
[24] F. Hochard-Poncet, P. Delichere, B. Moraweck, H. Jobic, A.
Renouprez, J. Chem. Soc., Faraday Trans. 91 (1995) 2891.
[25] B.W. Hoffer, A.D. van Langeveld, J.-P. Janssens, R.L.C.
Bonn, C.M. Lok, J.A. Moulijn, J. Catal. 192 (2000) 432.

[26] O. Bse, E. Kemnitz, A. Lippitz, W.E.S. Unger, Fres. J. Anal.


Chem. 358 (1997) 175.
[27] S. Hamar-Thibault, J. Masson, J. Chim. Phys. 88 (1991)
219.
[28] F.M. Capece, V. Di Castro, C. Furlani, G. Mattogno, C.
Fragale, M. Gargano, M. Rossi, J. Electron. Spectrosc. Relat.
Phenom. 27 (1981) 119.
[29] G.A. Martin, P. Fouilloux, J. Catal. 38 (1975) 231.
[30] P. Mars, J.J.F. Scholten, P. Zwietering, Proceedings of the
Deuxime Congrs International de Catalyse, Technip, Paris,
1961, p. 1245.
[31] F. Kapteijn, G.B. Marin, J.A. Moulijn, in: R.A. van Santen,
P.W.N.M. van Leeuwen, J.A. Moulijn, B.A. Averill (Eds.),
Catalysis: An Integrated Approach, Elsevier, Amsterdam,
1999, p. 375.
[32] J. Pardillos-Guindet, S. Metais, S. Vidal, J. Court, P.
Fouilloux, Appl. Catal. 132 (1995) 61.
[33] G. De Wit, J.J. De Vlieger, A.C. Kock-Van Dalen, R. Heus, R.
Laroy, A.J. Van Hengstum, A.P.G. Kieboom, H. Van Bekkum,
Carbohydr. Res. 91 (1981) 125.
[34] A.J. Van Hengstum, A.P.G. Kieboom, H. Van Bekkum, Starch
36 (1984) 317.
[35] E. Crezee, B.W. Hoffer, P.R.M. Mooijman, B. van der Linden,
J.C. Groen, M.L. Toebes, K.P. de Jong, F. Kapteijn, J.A.
Moulijn, in preparation.
[36] R.P. Verma C.S.L. Narasimhan, in K.D.P. Nigam, A. Schumpe
(Eds.), Three-Phase Sparged Reactors, Gordon and Breach,
Amsterdam 1996, p. 461.

Вам также может понравиться