New synthesis method for nickel phosphide hydrotreating catalysts

Article abstract of DOI:10.1039/b507940c

Nickel phosphide particles on silica and alumina support were prepared from metal or metal oxide particles by treatment with phosphine and hydrogen at moderate temperature, resulting in small particle sizes equivalent to that of the precursor particle size. The Royal Society of Chemistry 2005.

Full text of DOI:10.1039/b507940c

View Online / Journal Homepage / Table of Contents for this issue
COMMUNICATION
www.rsc.org/chemcomm | ChemComm
New synthesis method for nickel phosphide hydrotreating catalysts
Shaofeng Yang and Roel Prins*
Received (in Cambridge, UK) 7th June 2005, Accepted 7th July 2005
First published as an Advance Article on the web 27th July 2005
DOI: 10.1039/b507940c
Fig. 1 shows the XRD patterns of the reduced N-13 Ni/SiO2
precursor, prepared from nickel nitrate, after reaction with 10%
PH3/H2 at different temperatures. After reaction at 298 K, the
diffraction characteristics of Ni particles are still present, similar to
those of the starting Ni/SiO2 precursor. The Ni peaks almost
disappeared after reaction at 323 K, only a broad Ni peak at
ca. 44.6u remained. In the XRD pattern of the Ni/SiO2 precursor
after reaction with phosphine at 373 K intense peaks were present
at 40.7u, 44.6u and 47.4u and weaker peaks at 54.2u and 55.0u, all
attributed to Ni2P.4 Very weak peaks were present at 41.8u, 42.8u,
43.6u and 45.3u, corresponding to Ni3P.4 When the reaction
temperature was increased to 423 and 523 K, only the Ni2P
diffraction peaks were present, indicating that pure Ni2P can be
prepared on silica by treatment with phosphine and hydrogen
above 423 K.
Nickel phosphide particles on silica and alumina support were
prepared from metal or metal oxide particles by treatment with
phosphine and hydrogen at moderate temperature, resulting in
small particle sizes equivalent to that of the precursor particle
size.
The need to deeply reduce the sulfur level of fuels has led to a high
interest in materials different from the classic hydrotreating
catalysts based on MoS2. Metal carbides and nitrides are initially
highly active, but are not stable in a sulfur-containing atmosphere.1
Supported metal phosphides of W, Mo, Co and Ni can be
prepared from the corresponding metal salts and (NH4)2HPO4 by
reduction in H2 and the resulting materials have very promising
hydrotreating activities.2–8 Under hydrodesulfurization (HDS)
conditions the surfaces of molybdenum and nickel phosphide take
up sulfur atoms, but the kernels of the metal phosphide particles
stay intact.8–10 Unfortunately, the dispersion of thus-prepared
supported metal phosphides is low, because the P–O bond is
strong and its reduction requires high temperature. We have
therefore looked for phosphor-containing species that contain P–X
bonds that are easier to break than the P–O bond in phosphates
and phosphites. We have found that supported metal phosphide
particles of high dispersion can be obtained by treating reduced
metal particles as well as metal oxide particles on a support by
phosphine. In this method there is no need to go to high
temperatures, and as a consequence the size of the metal phosphide
particles is equivalent to that of the metal particles and can be kept
small.
Metallic Ni/SiO2 precursors, prepared from different metal salts,
were treated in situ with 10% PH3/H2 at 523 K. The crystallite sizes
of the samples corresponding to the [111] diffraction at 40.7u were
calculated using the Scherrer equation (D 5 0.9l/b cosh). Fig. 2a
shows that the XRD pattern of the catalyst prepared by the
classical phosphate reduction method had sharp peaks; large Ni2P
particles with an average crystallite size of 30 nm had formed. The
N-13 and N-5 Ni2P/SiO2 catalysts exhibited broader XRD peaks
(Figs. 2b and c) with average crystallite sizes of 13 and 11 nm,
respectively (the Ni particle sizes of their Ni/SiO2 precursors were
Silica-supported nickel phosphide was prepared by reaction of
supported Ni and NiO particles with PH3. The supported metal
precursor was prepared by pore-volume impregnation. Aqueous
solutions of nickel nitrate and nickel acetate were used in the
impregnation. The corresponding samples are denoted as N–X
(nitrate) and A–X (acetate) respectively, where X is the weight
percent of nickel in the samples. The dried, impregnated N–X and
A–X solids were reduced in flowing H2 at 673 K for 2 h, followed
by cooling down to 298 K in flowing H2. The dried N–X
precursors were also calcined at 773 K for 3 h before reduction in
H2. The reduced Ni/SiO2 samples were exposed to a flowing 10%
PH3/H2 mixture (20 ml min21) between 298 and 523 K; the final
temperature was maintained for 2 h. Then the samples were cooled
down to 298 K in flowing H2, flushed with He for 20 min and
passivated in a flow of 1 mol% O2/He. In addition, a sample was
prepared by the classical phosphate reduction method2 for
comparison (NP-13, nickel loading 13 wt%).
Institute for Chemical and Bioengineering, Swiss Federal Institute of
Technology (ETH), 8093, Zu¨rich, Switzerland.
E-mail: roel.prins@chem.ethz.ch; Fax: +41-44-6321162;
Tel: +41-44-6325490
Fig. 1 XRD patterns of the 13 wt% Ni/SiO2 precursor, prepared from
nickel nitrate, after reaction with phosphine at different temperatures. x,
Ni3P.
4178 | Chem. Commun., 2005, 4178–4180
This journal is ß The Royal Society of Chemistry 2005

View Online
obtained by comparing spectra measured at different spinning
rates. The Ni2P structure contains two crystallographic P sites and
consequently gives two resonances with isotropic chemical shifts
of 1487 and 4076 ppm.4 31P MAS-NMR spectra of the A-5
Ni2P/SiO2 catalysts prepared at 423 K showed peaks around 1478
and 4060 ppm, confirming that Ni2P had formed.
To examine if PH3 also reacted with supported NiO, the
calcined 13 wt% NiO/SiO2 sample was treated with a 10% PH3/H2
mixture at different temperatures. The XRD patterns show that,
after reaction with phosphine at 523 K for 2 h, the NiO peaks had
almost disappeared and that weak peaks at 40.7u, 44.6u and 47.4u
attributed to Ni2P were present. When the reaction temperature
was increased to 623 K, only the diffraction peaks corresponding
to Ni2P were present. This indicates that silica-supported Ni2P can
also be achieved by treating a NiO/SiO2 precursor with phosphine
above 623 K.
In 1996 Robinson et al. described several methods to prepare
Ni2P and Co2P and studied the HDN activity of quinoline of the
resulting materials.11 One of their preparation methods was the use
of phosphine, but only a limited number of experiments was
described. Our results demonstrate that metal phosphide particles
on a support can very easily be prepared from metal or metal oxide
particles by treatment with phosphine and hydrogen. In the case of
Ni, the metal particles already became fully phosphided at 423 K,
while NiO particles needed 623 K. These temperatures are very
much lower than the temperatures required in the phosphate
method of preparing metal phosphide. Because of the very strong
P–O bond, temperatures of about 1000 K are needed.2–6 Such high
temperatures lead to the almost exclusive formation of phosphorus
(phosphine is not stable at high temperature), which diffuses only
slowly into the metal particles. All this leads to sintering and loss of
dispersion.
Fig. 2 XRD patterns of Ni2P/SiO2 catalysts prepared from different
precursors.
10 and 8 nm). The XRD pattern of sample A-5 showed only the
features of amorphous silica, implying that the nickel phosphide
particles were too small to be detected by XRD. This could be
further confirmed by the STEM image of Ni2P/SiO2 catalyst of
sample A-5 (Fig. 3).
For the highly dispersed A-5 Ni2P/SiO2 catalyst, for which no
XRD lines could be measured, NMR spectroscopy proved a
powerful tool. The metallic character of metal phosphides leads to
large 31P NMR Knight shifts, which makes it easy to distinguish
metal phosphides from diamagnetic phosphates.4,6 31P MAS-
NMR spectra were obtained with an Advance 400 WB Bruker
spectrometer at room temperature. Samples were packed into a
4 mm diameter rotor in an inert atmosphere and measured with
spinning at 12 kHz. The isotropic shifts of the signals were
At these high temperatures phosphate also reacts with supports
like alumina and therefore a large excess of phosphate has to be
added in order to make metal phosphide particles.5,12 Thus, only
alumina-supported nickel phosphide catalysts with low dispersions
and loadings above 20 wt% have been reported.12 The present
phosphine method allows preparing highly dispersed metal
phosphide particles even on an alumina support. This is an
important result, because alumina is industrially preferred to silica
since it is thermally much more stable. Fig. 4e shows the XRD
pattern of the alumina-supported nickel phosphide catalyst (Ni
loading 10 wt%) prepared by treating the metallic Ni/Al2O3
precursor (Fig. 4c) with a 10% PH3/H2 mixture at 523 K. In situ
NMR measurements on the c-Al2O3 support after treatment with
PH3 at 573 K for 2 h showed no phosphate peaks around 0 ppm.
This indicates that the alumina support was not phosphided.
The hydrotreating activities of the A-5 and N-13 Ni2P/SiO2
catalysts were tested in the HDN of o-methylaniline (OMA) and
HDS of dibenzothiophene (DBT) at 613 K and 3.0 MPa. The
passivated catalysts (0.2 g, diluted with 8 g SiC) were activated
in situ with H2 at 673 K and 0.1 MPa for 3 h. Reactions were
carried out in a continuous–flow microreactor as described
before.6,13 The composition of the gas-phase feed consisted of
130 kPa octane or toluene (used as the solvents for HDN and
HDS, respectively), 17 kPa heptane (as reference for OMA and its
derivatives in the GC analysis) or 8 kPa dodecane (as reference for
DBT and its derivatives in the GC analysis), and about 2.8 MPa
H2. For each of the catalysts, the HDS reaction was carried out
Fig. 3 STEM image of A-5 Ni2P/SiO2 catalyst.
This journal is ß The Royal Society of Chemistry 2005
Chem. Commun., 2005, 4178–4180 | 4179

View Online
Table 1 Hydrotreating performance of Ni2P/SiO2 catalysts
Conversion (%) Selectivity (%)
Reaction
A-5
N-13
82
Product
A-5
N-13
HDS of DBT 65
HDN of OMA 76
Biphenyl
Cyclohexylbenzene
TH-DBT
Methylcyclohexane 75
Methylcyclohexene 10
93
6
1
85
14
1
59
5
90
Toluene
Ethylcyclopentane
12
3
34
2
92% at the highest weight time) is much higher than that over
sulfided catalysts, which is 70y80% in general.13 This could be due
to the small particle size of the A-5 Ni2P/SiO2 catalyst, which
enhances the s adsorption of DBT molecules on the active sites.
The amount of tetrahydro-DBT was very low, showing that it is
easily desulfurized. The further hydrogenation of cyclohexyl-
benzene to bicyclohexyl did not take place.
The phosphine method allows preparing metal phosphide
particles at moderate temperature. As a result, the particle size is
equivalent to that of the precursor particle size and small
supported metal phosphide particles can be made of any metal
or metal oxide precursor that can be made with high dispersion.
Even on alumina, low-loading metal phosphide catalysts with high
dispersion and high activity can be made.
Fig. 4 XRD patterns of alumina support (a), Ni/Al2O3 precursor (c),
Ni2P/Al2O3 catalysts prepared by treating (c) with phosphine at 523 K (e),
and calculated patterns of Ni (b) and Ni2P (d).
subsequent to the HDN reaction, without changing the catalyst.
The reaction products were analyzed by off-line gas chromato-
graphy. The catalyst was stabilized at 613 K and 3.0 MPa for at
least 48 h before samples were taken for analysis.
The authors thank Dr. Changhai Liang for performing the
hydrotreating measurements.
The hydrotreating reaction results for the A-5 and N-13
Ni2P/SiO2 catalysts are summarized in Table 1. The data were
taken at the weight time13 of 40 g min mol21 for OMA HDN and
20 g min mol21 for DBT HDS. The HDN activities of both Ni2P
catalysts were very good and much higher than that of the metal
phosphide catalysts determined before.6 Also the activities of both
Ni2P catalysts in the HDS of DBT were very high. They are higher
than that of the Ni2P/SiO2 catalyst described by Wang et al.14
Wang et al. observed a conversion that was 1.4 times higher than
ours. Assuming a first order rate equation, taking into account
that they used 11.7 times more catalyst, and that the HDS of DBT
was carried out at 613 K in our study and at 643 K in their study,
we estimate that our catalyst is about one order of magnitude
more active than that of Wang et al.
Notes and references
1 R. Prins, Adv. Catal., 2001, 46, 399.
2 P. Clark, W. Li and S. T. Oyama, J. Catal., 2001, 200, 140.
3 D. C. Phillips, S. J. Sawhill, R. Self and M. E. Bussell, J. Catal., 2002,
207, 266.
4 C. Stinner, Z. Tang, M. Haouas, Th. Weber and R. Prins, J. Catal.,
2002, 208, 456.
5 P. A. Clark and S. T. Oyama, J. Catal., 2003, 218, 78.
6 V. Zuzaniuk and R. Prins, J. Catal., 2003, 219, 85.
7 S. T. Oyama, J. Catal., 2003, 216, 343.
8 Y. Shu and S. T. Oyama, Chem. Commun., 2005, 1143.
9 Z. Wu, F. Sun, W. Wu, Z. Feng, C. Liang, Z. Wei and C. Li, J. Catal.,
2004, 222, 41.
10 K. A. Layman and M. E. Bussell, J. Phys. Chem. B, 2004, 108, 10930.
11 W. R. A. M. Robinson, J. N. M. van Gestel, T. I. Koranyi, S. Eijsbouts,
A. M. van der Kraan, J. A. R. van Veen and V. H. J. de Beer, J. Catal.,
1996, 161, 539.
12 S. J. Sawhill, K. A. Layman, D. R. Van Wky, M. H. Engelhard,
C. Wang and M. E. Bussell, J. Catal., 2005, 231, 300.
13 M. Egorova and R. Prins, J. Catal., 2004, 225, 417.
14 X. Wang, P. Clark and S. T. Oyama, J. Catal., 2002, 208, 321.
The product distribution (the selectivity of biphenyl is very high)
shows that DBT is mainly desulfurized by the DDS route over
phosphide catalysts. This is similar to that over sulfided CoMo and
NiMo catalysts. But it should be noted that the selectivity of
biphenyl over the A-5 Ni2P/SiO2 catalyst (95% at the lowest and
4180 | Chem. Commun., 2005, 4178–4180
This journal is ß The Royal Society of Chemistry 2005