DOI: 10.1002/cctc.201801430
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

Full Papers

Ensemble Design in Nickel Phosphide Catalysts for Alkyne
Semi-Hydrogenation
Davide Albani,[a] Konstantin Karajovic,[b] Bharath Tata,[a] Qiang Li,[b] Sharon Mitchell,[a]
Núria López,*[b] and Javier Pérez-Ramírez*[a]
Modification of transition metals with p-block elements is
known to be effective to tune the ensemble characteristics of
catalysts for the semi-hydrogenation of alkynes. To further
explore this approach, here we prepare two nickel phosphides,
namely Ni2P and Ni5P4. Assessment in the semi-hydrogenation
of 1-hexyne and 2-methyl-3-butyn-2-ol shows that the phosphides present higher rate and selectivity than unmodified
nickel catalysts. While no activity and selectivity differences are
displayed in the semi-hydrogenation of 1-hexyne over Ni2P and
Ni5P4, in the case of 2-methyl-3-butyn-2-ol a higher rate and

lower selectivity to 2-methyl-3-buten-2-ol are observed over
Ni2P. Density functional theory reveals that the hydroxyl group
facilitates the reaction, but also increases the barrier for product
desorption. Detailed analyses of the ensemble show the
potential of phosphorus to create spatially-isolated nickel
trimers that surpass the performance of unmodified nickel, but
also its limited ability to modulate the electronic properties and
related binding energies of organic intermediates, which is key
to preventing undesired side reactions.

Introduction

adjacent alkyne adsorption sites (i. e., comprising at least 5–6
neighboring atoms).[11,12] To reduce the ecological footprint of
hydrogenation processes, new materials have been investigated
including other metals, alloys and intermetallic compounds, metal
oxides, hybrid nanoparticles, supported single atoms, and p-block
modified noble metals.[6,13] As lately shown by an integrated
experimental and theoretical work that classified prominent
families of hydrogenation catalysts, the modification of palladiumbased catalysts with sulfur-containing molecules, which yields
nanostructured palladium sulfide (i. e., Pd3S and Pd4S) emerged as
one of the most effective ways of tuning the ensemble properties
to obtain a robust catalyst.[13,14] To benefit from this design
strategy, we have explored if the approach applies to other pblock elements.
Transition metal phosphides are an intriguing class of materials
that have been applied in different fields from materials science to
catalysis. In the latter domain, attention has primarily focused on
the activity of Ni, Co, W, and Fe phosphides in reactions such as
hydrotreating,[15] hydrodenitrization,[16] aminoborane decomposition,[17] and electrochemical hydrogen evolution.[18] Several crystalline phosphide phases exist with different phosphorus-to-metal
ratios,[19] potentially offering a material versatility similar to that of
sulfides. Thus, it seems possible to optimize the geometric and
electronic properties of the ensemble and consequently the
catalytic function within this family of compounds. In view of the
low cost of nickel (ca. 2000 times cheaper than palladium)[20,21] and
of the abundance and hydrogen splitting ability of this metal,
nickel phosphides have also attracted interest for alkyne semihydrogenation, although only a few contributions have evaluated
their activity. In this respect, Corma and coworkers[22] assessed
colloidal nickel phosphides in the semi-hydrogenation of functionalized alkynes, while bulk NixPy phases were prepared by thermal
treatment of hydrotalcite-like compounds containing Ni and
tested in phenylacetylene hydrogenation by Chen et al.[23] These
two contributions showed that the well-known unselective

The need for sustainable processes in today’s society steers
research in heterogeneous catalysis towards developing more
environmentally friendly and efficient systems.[1,2] Since 1952, the
three-phase semi-hydrogenation of alkynes, which is key for the
industrial synthesis of drugs, vitamins, fragrances, and agrochemicals, has been conducted over lead-poisoned palladium
(5 wt.%) supported on calcium carbonate (the well-known Lindlar
catalyst).[3,4] Severe drawbacks of the state-of-the-art catalyst (i. e.,
low noble metal utilization, and use of hazardous Pb) have
prompted a search for alternative materials.[5,6] The replacement of
the archetypal catalyst has not been an easy task due to the
challenges associated with precisely tailoring the active site to
achieve selective performance. On active metals such as Ni and
Pd, to prevent the over-hydrogenation to the alkane, the
electronic properties of the catalyst need to be modulated with
suitable modifiers, to weaken the adsorption of the desired semihydrogenated product[7–10] so that it can easily detach from the
surface improving selectivity. More importantly, oligomerization
paths must be suppressed by downsizing the ensemble size to 3
metal atoms, thus avoiding CÀ C couplings which require two
[a] D. Albani, B. Tata, Dr. S. Mitchell, Prof. J. Pérez-Ramírez
Department of Chemistry and Applied Biosciences
Institute for Chemical and Bioengineering
ETH Zurich
Vladimir-Prelog-Weg 1, 8093 Zurich (Switzerland)
E-mail: jpr@chem.ethz.ch
[b] K. Karajovic, Dr. Q. Li, Prof. N. López
The Barcelona Institute of Science and Technology
Institute of Chemical Research of Catalonia (ICIQ)
Av. Països Catalans 16, 43007 Tarragona (Spain)
E-mail: nlopez@iciq.es
Supporting information for this article is available on the WWW under
https://doi.org/10.1002/cctc.201801430
This manuscript is part of the Anniversary Issue in celebration of 10 years of
ChemCatChem.

ChemCatChem 2019, 11, 457 – 464

457

© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

behavior of bare nickel[9,10] can be modulated by treatment with
phosphorus. Similar findings have been recently extrapolated by
Anderson et al.[24] to nanostructured palladium phosphide catalysts
for gas-phase applications. However, the activity of phosphides
could not match that of a nanostructured Pd4S catalyst. Overall,
these investigations postulated that the presence of phosphorus
on the surface is key in hindering side paths. On the other hand,
the origin of the selective performance and the link to ensemble
design have not been rationalized at the molecular level.
Herein, we provide fundamental understanding of the
performance of nickel phosphide catalysts in alkyne semihydrogenation. This is achieved by combining state-of-the-art
Density Functional Theory (DFT), syntheses of bulk nickel
phosphides with different stoichiometries (Ni2P and Ni5P4), and
catalytic tests of linear and functionalized alkynes in continuous-flow mode. In contrast to batch setups commonly
employed for laboratory testing,[25] continuous-flow reactors
can operate efficiently at steady state and offer key advantages,
such as cost reduction, efficient heat transfer, rapid mixing, and
small volume usage.[26–28] Overall, an emphasis is placed on the
role of the ensemble properties on the complex reaction
network, revealing the limitations of phosphorus as an effective
electronic modifier.

Results and Discussion
Catalysts Preparation and Characterization
Nickel phosphides are generally prepared by reacting the pure
elements (i. e., Ni and yellow or red phosphorus) or alternatively
using nickel salts in combination with phosphorus-containing
compounds such as, phosphines and phosphonic acid.[29] Many
works have investigated the use of yellow phosphorus, which
enables precise control over the final stoichiometry, but its
highly pyrophoric behavior and the need for water-free
conditions make it unattractive.[16] In this contribution we
synthesized two of the most robust NixPy structures: Ni5P4 and
Ni2P. The former can be readily prepared from more stable red
phosphorus, by treatment with metallic nickel powder under an
inert atmosphere at 773 K, while it is also straightforward to
isolate the latter by the reaction of nickel chloride hexahydrate
and sodium hypophosphite at 573 K.[30]
Following these approaches, two nickel phosphides with
different stoichiometry were obtained. X-ray diffraction analyses
confirmed the presence of a single crystalline phase corresponding to the hexagonal structure of Ni2P or Ni5P4, respectively (Figure 1a). Also, Ni5P4 does not display any reflections
from the nickel powder precursor. The experimentally determined lattice parameters agree well with the calculated values
(Table S1). In comparison, Ni presents the typical diffractogram
of metallic nickel, while the commercial Ni/SiO2-Al2O3 displays
reflections arising from both metallic and oxidic nickel phases.
Consistent with the expected stoichiometries, assessment of the
bulk composition by X-ray fluorescence (XRF) evidenced an
atomic Ni/P ratio of 2 (Ni2P) or 1.2 (Ni5P4), and excluded the
presence of trace impurities. Analysis by N2 physisorption
ChemCatChem 2019, 11, 457 – 464

www.chemcatchem.org

Figure 1. (a) X-ray diffraction patterns and (b) 31P magic angle spinning
nuclear magnetic resonance spectra of the catalysts. Vertical lines below the
diffractograms in (a) show the reference patterns of Ni2P (PDF-Number 98005-2967), Ni5P4 (PDF-Number 98-004-8850), Ni (green, PDF-Number 98-0060941), and Ni15O16 (blue, PDF-Number 98-008-8918).

confirmed the nonporous nature of these phosphides and of
the nickel precursor, which display Brunauer-Emmett-Teller
(BET) surface areas lower than 1 m2 gÀ 1, while the commercial
catalyst featured a high surface area (i. e., 190 m2 g– 1) due to the
high porosity of the silica-alumina support. Solid state magic
angle spinning nuclear magnetic resonance spectroscopy (MASNMR) was applied to further probe the atomic structure. The
analysis focused on 31P, considering the high sensitivity and
abundance of this nucleus (Figure 1b). The spectra agree well
with previously reported data.[31,32] In particular, the high
chemical shifts observed are expected due to the metallic
character of the two phases, which is responsible for the socalled Knight shift (Table 1). Ni2P displays two inequivalent
phosphorus sites with very different chemical shifts, which are
assigned to a phosphorus atom shared by two NiP4 tetrahedral
and one at the common edge of a NiP4 tetrahedral and a NiP5
pyramid, respectively. On the other hand, the case of Ni5P4 is
more complex because, from inspection of the spectra, 5 types
of inequivalent P sites are identified in the region 800–

458

© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

Table 1. NiÀ Ni, NiÀ P, and PÀ P bond distances on the catalyst surfaces.
Bond

Ni(111)

Ni3P(0001)

Ni3P2(0001)

Ni4P3(0001)

NiÀ Ni [Å]
NiÀ P [Å]
PÀ P [Å]

2.492
–
–

3.177
5.876
2.268; 3.652

2.716
3.393
2.188

2.413; 2.619
3.088; 3.711
2.322; 3.174

2200 ppm at 1910, 1816, 1340, 930, and 720 ppm with the
corresponding sidebands. The five sites arise due to the specific
atomic structure of Ni5P4 in which P atoms are surrounded by
different numbers of nearest neighbor Ni atoms.[33] The
appearance of distinct bands associated with each P position
may vary depending on both the extent of structural disorder
in the phase and on the measurement conditions.[32] The latter
are particularly critical due to the metallic character of nickel
phosphides that limits the spinning speed that can be applied
and can produce spurious signals related to temperature
effects. Also, the origin of the band at 4000 ppm is unclear.

Geometric Properties of NixPy Ensembles
A selective semi-hydrogenation catalyst requires spatially-isolated ensembles to avoid the occurrence of CÀ C coupling,
which leads to oligomerization. This is assessed by identifying
the distribution of nickel and phosphorus on the surface and
associated ensemble motifs. Upon inspection of Ni(111) and the
(2 × 2) supercells of two terminations of Ni2P (Ni3P(0001) and
Ni3P2(0001)), and the lowest energy termination of Ni5P4
(Ni4P3(0001)) (Figure 2), it is possible to identify a pattern of
trimeric ensembles in all cases. In particular, we can determine
that the distribution and relative organization of fcc (green) and
hcp (yellow) sites change from the Ni(111) to Ni4P3(0001)
surface depending on the surface concentration of phosphorus.
For example, while the green ensembles, which are arranged in
larger triangular motifs, on Ni3P(0001) are interrupted by one
phosphorus, on Ni3P2(0001) and Ni4P3(0001) the number of
phosphorus atoms between the nickel ensembles increases.
This implies a higher degree of separation and lower density of
Ni3 ensembles on the surface compared to the palladium

ensembles on Pd3S,[13] pointing out the high efficiency of
phosphorus in imparting geometric modifications.
To understand the phosphorus-rich nature of the surfaces,
the segregation energy, Eseg, which is a robust parameter for
evaluating the tendency of bulk P and/or Ni atoms to migrate
to the surface, and the island formation energy, Eisl, which is
used to describe the possibility of forming surface homoatomic
domains, have been calculated (Tables S2 and S3). In all cases
phosphorus atoms are identified to be prone to migrate to the
surface, while nickel prefers to move to the bulk. However, as
liquid phase applications at mild temperatures are conducted in
the current study, the extent of diffusion of phosphorus atoms
towards the surface is expected to be minimal.

Adsorption Properties of NixPy Ensembles
Evaluation of the adsorption energies of C2H2, C2H4, and C2H6 on
the investigated surfaces (Figures S1–S4) provides a means of
determining the impact of phosphorus on the electronic
properties of the ensembles. The values can be compared with
other prominent classes of hydrogenation catalysts to derive a
hierarchy of the effectiveness of different approaches to
optimize ensemble design. The adsorption energy of C2H2 is
one of the main activity descriptors for alkyne semi-hydrogenation catalysts and needs to be exothermic.[20] While C2H2
binds strongly to all surfaces, the adsorption geometry of the
activated acetylenic fragment changes as a function of the
nickel ensemble and participating phosphorus atoms. On Ni
(111), C2H2 can bind either to a site containing four Ni atoms
with the CÀ C bond sitting perpendicular (Eads = À 2.90 eV) or
above the Ni3 site with the two C atoms on two NiÀ Ni bridge
positions, (Eads = À 2.50 eV) (Figure S1). On Ni3P(0001), two
ethylene-like structures, PÀ CHÀ CHÀ Ni, and NiÀ CHÀ CHÀ Ni, are
identified with energies of À 2.07 and À 1.66 eV, respectively.
Similarly, on Ni3P2(0001), PÀ CHÀ CHÀ P and NiÀ CHÀ CHÀ Ni can be
distinguished with Eads of À 2.23 and À 2.47 eV, respectively. On
Ni4P3(0001), the most stable structure is PÀ CHÀ CHÀ P with the
two C atoms bound to P atoms (Eads = À 3.07 eV). In contrast to
other p-block-doped metal surfaces, the binding of the alkyne
to phosphorus atoms is much stronger than at any other
possible coordination sites on NixPy.

Figure 2. Top view of the Ni(111), Ni2P (Ni3P(0001) and Ni3P2(0001) terminations) and Ni5P4 (Ni4P3(0001) termination) surfaces. Color code: Ni (blue and green),
P (pink). Triangles highlight the nickel atoms forming surface ensembles. Ni4P3(0001) displays two inequivalent Ni colored in blue and green.

ChemCatChem 2019, 11, 457 – 464

www.chemcatchem.org

459

© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

To obtain a high selectivity to the desired product, it is
important that the intermediate alkene displays a thermoneutral or endothermic adsorption energy, so that it can easily
detach from the surface, suppressing over-hydrogenation. C2H4
binds to Ni(111) via one carbon to the nickel top site and the
other carbon to the NiÀ Ni bridge site[9] with a binding energy of
À 1.14 eV (Figure S1). This binding geometry is not affected by
the surface composition since the adsorption energy on Ni3P
(0001) and Ni3P2(0001) is À 1.13, and À 1.26 eV, respectively
(Figures S2 and S3). On Ni4P3(0001), the adsorption of C2H4 to a
phosphorus site, forming PÀ CH2À CH2À P, is even stronger (Eads =
À 2.23 eV) and more stable than on the Ni3 site of the same
surface (Eads = À 0.98 eV). This shows that having trimeric Ni
ensembles spaced by phosphorus atoms does not necessarily
poison nickel, but it provides new P-containing sites on the
surfaces with stronger adsorption features.

Hydrogenation of Alkynes
The nickel phosphide catalysts were evaluated in the continuous-flow three-phase semi-hydrogenation of 1-hexyne, a substrate widely studied to probe the selective character of alkyne
hydrogenation catalysts, and benchmarked against the nickel
precursor. Figure 3a maps the influence of the reaction temperature and pressure on the rate and selectivity to 1-hexene.
While both Ni2P and Ni5P4 display very similar performance
under every condition evaluated, for instance, at 303 K and
10 bar the rate of 1-hexyne conversion is 1.3 and 1.4 hÀ 1, on
Ni2P and Ni5P4, respectively, Ni present much lower reaction
rate, 0.4 hÀ 1. Remarkably, nickel phosphides are almost fully
selective (higher than 95 %) to the desired alkene even at high
pressure (P = 20 bar) and temperature (T = 343 K), while Ni
displays ca. 70 % selectivity at a much lower conversion level
(Figure S5). This points to the robustness of nickel phosphides
to the formation of β-hydrides, which bare nickel is known to
suffer from, and to the beneficial effect of introducing
phosphorus into the nickel lattice to increase both the activity
and the selectivity. To assess the selectivity patterns of a Nibased catalyst at higher conversion levels, a commercial
60 wt.% Ni/SiO2À Al2O3 catalyst was evaluated in 1-hexyne
hydrogenation. At 343 K and 20 bar the rate of 1-hexyne
conversion on Ni/SiO2À Al2O3 is 3.5 hÀ 1 and the selectivity to 1hexene of 18 %, with n-hexane as the major side product and
traces of isomers of the double bond (ca. 5 %). These results
indicate that supported nickel-based catalyst can match and
surpass the activity of bulk nickel phosphides, but at the
expenses of a very low selectivity to the semi-hydrogenated
product.
The catalysts were additionally evaluated in the continuousflow three-phase semi-hydrogenation of 2-methyl-3-butyn-2-ol
at various temperatures and pressures. Conversely from the
case of 1-hexyne, where Ni2P and Ni5P4 display similar activity
profiles, significant differences are observed in the hydrogenation of the alkynol (Figure 3b). In particular, at 303 K and
5 bar the rate of 2-methyl-3-butyn-2-ol conversion is 1.6 and
1.1 hÀ 1, while the selectivity to 2-methyl-3-buten-2-ol is 92 and
ChemCatChem 2019, 11, 457 – 464

www.chemcatchem.org

90 % on Ni2P and Ni5P4, respectively. These results point to the
higher activity of Ni2P than Ni5P4 when compared under the
same conditions, and call for a molecular-level understanding
of this behavior. In line with the observations for 1-hexyne, Ni
and Ni/SiO2À Al2O3 exhibit lower and similar reaction rates to
NixPy, respectively. Also, at 343 K and 20 bar, the selectivity to 1hexene on Ni and Ni/SiO2À Al2O3 is 60 % and 10 %, respectively.
To verify the robustness on stream, a long-term test was
conducted over the catalysts. As shown in Figure 4, both Ni2P
and Ni5P4 displayed an induction period of 2 h, likely due to the
reduction of the oxidic layer covering the surface,[16] before
reaching stable performance for 28 h on stream, confirming the
suitability of these systems for continuous-flow applications.
Analysis of the used catalysts by X-ray diffraction and X-ray
fluorescence spectrometry confirmed the preservation of the
crystalline structure and the absence of leaching of either Ni or
P.
The recent kinetic study over palladium sulfides evidenced the
role of the p-block element (i. e., sulfur) in stabilizing the activated
H atoms, imparting a bifunctional mechanism that originates
higher activity than unmodified palladium nanoparticles, where H2
and the alkyne compete for the same site.[13] Prompted by these
observations, the mechanism was explored by conducting kinetic
tests to determine the reaction order with respect to hydrogen
and the alkyne over Ni2P, Ni5P4, and Ni/SiO2À Al2O3. In line with
previous investigations,[13] the order with respect to molecular
hydrogen is positive (ca. 1.2) in all cases, while the order to the
alkyne is close to 0 for nickel phosphide (À 0.04 and À 0.20 for Ni2P
and Ni5P4, respectively, Figure S6), and À 1 over Ni/SiO2À Al2O3,
suggesting in the latter situation a competition between the
alkyne and hydrogen on the surface of the unmodified nickel
nanoparticles. Also, this indicates a role of the heteroatom in the
reaction mechanism as now a two-site model is required to
describe the kinetic behavior.
It is noteworthy that in comparison with the nickel phosphides, a nanostructured Pd3S catalyst, the highest-performing Pdbased catalyst identified to date, displayed a two orders of
magnitude higher reaction rate under every condition evaluated.
This indicates that phosphorus is not able to sufficiently promote
the performance of nickel to match that of palladium.

Reaction Mechanism
To understand the structure sensitive or insensitive activity profile
as a function of the substrate, molecular solvation corrected
reaction profiles for the hydrogenation of C2H2 to C2H6, comprising
four hydrogenation steps (CHCH* + 4H*!CHCH2* + 3H*!
CH2CH2* + 2H*!CH2CH3* + H*!CH3CH3*), were computed (Figure 5). On Ni(111), the reaction barriers (energies) are 1.02 (0.61),
0.60 (À 0.02), 0.55 (0.39), 1.17 (À 0.07) eV. For Ni2P, Ni3P2(0001) was
selected to investigate the reaction due to its higher stability.[34]
The corresponding values are: 0.93 (0.20), 1.00 (À 0.04), 1.05 (0.49),
and 1.09 (À 0.05) eV. The situation on Ni3P4(0001) is slightly
different. Since C2H2 can be bound with two different adsorption
structures, PÀ CHÀ CHÀ P and PÀ CHÀ CHÀ Ni (vide supra), with the
former showing stronger adsorption that might potentially poison

460

© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

Figure 3. Reaction rate (top row, in molalkyne molNiÀ 1 hÀ 1) and selectivity (bottom row, in %) in the hydrogenation of (a) 1-hexyne to 1-hexene and (b) 2-methyl3-butyn-2-ol to 2-methyl-3-buten-2-ol over the catalyst as a function of the temperature and pressure. Conditions: Wcat = 0.2 g, FL(alkyne + toluene)
= 0.5 cm3 minÀ 1, FG(H2) = 48 cm3 minÀ 1.

ChemCatChem 2019, 11, 457 – 464

www.chemcatchem.org

461

© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

Figure 4. Reaction rate (solid symbols) and selectivity (open symbols) to 2methyl-3-buten-2-ol over Ni2P (blue) and Ni5P4 (green) versus time-on-stream
in the hydrogenation of 2-methyl-3-butyn-2-ol to 2-methyl-3-buten-2-ol.
Conditions: Wcat = 0.2 g, P = 1 bar, T = 303 K, FL(alkyne + toluene)
= 0.5 cm3 minÀ 1, FG(H2) = 48 cm3 minÀ 1.

Figure 5. Energy profiles for the hydrogenation of acetylene on the Ni(111),
Ni3P(0001), Ni3P2(0001) and Ni4P3(0001) surfaces. Ni4P3-1 and Ni4P3-2 indicate
that the profiles consider two different adsorption configurations of the
acetylene molecule, forming PÀ CHÀ CHÀ P and PÀ CHÀ CHÀ Ni intermediates,
respectively.

the surface, both were taken as starting points to investigate the
reaction. From P-CH-CH-P the energies of the reaction profile are
(Ni4P3-1 in Figure 5) are 0.98 (À 0.75), 2.50 (0.29), 1.40 (À 0.33), 2.20
(0.64) eV; whereas considering NiÀ CHÀ CHÀ P they are 0.54 (0.37),
0.33 (À 0.32), 0.27 (À 0.72), 2.46 (0.56) eV (Ni4P3-2 in Figure 5). From
analysis of the barriers for the hydrogenation steps, it can be
inferred that breaking the PÀ C bond is energetically demanding
and thus the adsorption of acetylene on the P-containing sites
would not lead to a reaction.
The C2H4 desorption barriers on Ni(111), Ni3P2(0001), and
Ni4P3(0001) are 1.14, 1.26, 2.23 (PÀ CH2À CH2À P) and 1.04 eV
(PÀ CH2À CH2À Ni), which are lower (once entropy contributions
are included) or comparable to the corresponding highest overhydrogenation barriers to C2H6 of 1.17, 1.09, 2.20, and 2.46 eV,
respectively. This clarifies why the reaction is selective to the
alkene. The above energies correspond to gas-phase models,
ChemCatChem 2019, 11, 457 – 464

www.chemcatchem.org

and thus are referred to as gas-phase C2H2, H2, and clean
surfaces. To describe the effect of the solvent,[35] we calculated
the adsorption of toluene on all surfaces (À 1.17, À 1.35, À 1.30
and À 1.47 eV, on Ni(111), Ni3P(0001), Ni3P2(0001), and
Ni4P3(0001), respectively). The very negative value indicates that
toluene displaces C2H4 from the surface.
The higher rate of 2-methyl-3-butyn-2-ol conversion experimentally observed over Ni2P (Figure 3) and the similar activity
for 1-hexyne hydrogenation are rationalized based on the
computed reaction profiles. The hydrogenation of 1-hexyne can
be directly mapped to the profile for C2H2 as they are both
linear and unfunctionalized alkynes. The step with the highest
potential energy in the reaction profile (Figure 5), CHCH2* +
H*!CH2CH2*, on Ni3P2(0001) and Ni4P3(0001) differs only by
0.06 eV. The minor differences in the energy landscape for the
reaction on both catalysts shows again the negligible impact on
the overall performance of nickel phosphides by varying the
compositional Ni/P ratios between 2 and 1.2.
A different picture is observed in the hydrogenation of 2methyl-3-butyn-2-ol where the potential energy features a
0.25 eV higher barrier on Ni4P3(0001) compared to Ni3P2(0001)
(Figure S7), which facilitates the reaction path on Ni2P over
Ni5P4. This difference in reaction barrier can be traced back to
the specific interactions between the alkyne and the surface,
which changes the reaction profile. Also, the binding between
the hydroxyl group of the alkynol and the surface increases the
product desorption barrier, which enhances the likelihood of
competitive over-hydrogenation, explaining why the selectivity
levels to the alkene decrease in the hydrogenation of 2-methyl3-butyn-2-ol (70 %) compared to 1-hexyne (95 %) (Figures S7
and S8).
To gain further insight into the highly selective character of
the nickel phosphide catalysts, isomerization paths were
investigated for 1-hexyne. These processes are one of the main
cause of poor selectivity in hydrogenation reactions and require
the formation and cleavage of CÀ H bonds in different carbon
positions. On the unmodified Ni surface the dehydrogenation
step was found to feature similar barriers and thus compete
with the hydrogenation step, while on NixPy the hydrogenation
requires lower barriers (Table S4).
The higher activity in the hydrogenation of 2-methyl-3butyn-2-ol on Ni2P and high and low selectivity to 1-hexene
and of 2-methyl-3-buten-2-ol on both catalyst not only
indicates that the catalyst performance is sensitive to the
substrate, but also the lack of an impact of the Ni/P ratio on the
selectivity patterns within the range investigated. Phosphorus
acts effectively in forming Ni3 triangular ensembles, but its
modification of the electronic properties is limited. Optimal
semi-hydrogenation performance is known to depend on a
complex interplay of parameters, comprising the binding of the
reactants and intermediates, and the d-band shift. By evaluating
the adsorption energies, C2H2 and C2H4 are found to be ca. 1 eV
more strongly bound to NixPy than to Pd3S (Table S5). Comparatively, the adsorption of C2H4 on Ni3 undergoes only minor
variation, 0.25 eV, with respect to unmodified Ni(111) (Figure 6
and Figure S9), while a significant decrease is observed from Pd
(111) to Pd3S(202), from À 0.85 to À 0.22 eV, respectively. This is

462

© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

functionality different activity profiles and selectivity patterns
are displayed on the phosphides. Comprehensive theoretical
analyses including assessment of the adsorption energies and
the reaction profiles over the Ni3 ensembles rationalize that the
hydroxyl group in the alkynol molecules facilitates each reaction
step. While this enhances the activity, the higher desorption
barrier compromises the selectivity. Although the modification
with phosphorus defines trimeric ensembles and imparts a
bifunctional mechanism, it is demonstrated that the electronic
properties of the surface are not sufficiently modulated in the
range of Ni/P ratios investigated leading to lower rate and
selectivity levels than encountered over palladium sulfide.
Overall, this highlights the fundamental importance of balancing the electronic effects to reach optimal alkyne semi-hydrogenation performance.
Figure 6. Scaling relationship for the adsorption energy of C2H2 (blue) and
C2H4 (green) as a function of the adsorption energy of the methyl group on
Ni(111) (squares), Ni3P(0001) (diamonds), Ni3P2(0001) (triangles), Ni4P3 (0001)
(circles). The structures corresponding to each data point are presented in
Figure S9.

a consequence of the relatively small shift in the d-band centers
(0.1 eV) for NixPy compared to up to 0.5 eV for PdxS, which
measures the extent of electronic poisoning. This can be better
understood by plotting the adsorption energy of C2H2 and C2H4
as a function of the methyl group adsorption (Figure 6 and
Figure S9), following the approach proposed by Nørskov and
coworkers.[20,35] By applying this model, we can observe that the
theoretical points for Ni2P and Ni5P4 lie on the same line with
Ni, and the ratio of the slope of the trend lines is identical to
that reported, which points to the impossibility of this system
to break linear scaling relations. In fact, phosphorus enhances
the reactant and product adsorption/desorption to/from the
surface by forming PÀ CHÀ CHÀ P and PÀ CH2À CH2À P complexes,
which increases the difficulty of the intermediate alkene to
detach. Therefore, C2H2 and C2H4 adsorptions almost do not
vary upon including phosphorus, but they even get more
exothermic when the binding occurs via a P-containing site
(points labelled as 1 in Figure 6). The above description sets the
limits for the use of p-block dopants in alkyne hydrogenation.
In particular, it shows that the establishment of a general
doping strategy requires detailed investigation of the electronic
structure of possible new materials.

Conclusions
The potential of phosphorus to modify the properties of nickel
catalysts for alkyne semi-hydrogenation by using nickel phosphides has been explored. Two single phases (i. e., Ni2P and
Ni5P4) were prepared with different nickel to phosphorus ratios
and the properties of the ensembles in the most stable surface
terminations were identified, revealing the presence of spatially
isolated nickel trimers in all cases. Continuous-flow semi-hydrogenation of 1-hexyne and 2-methyl-3-butyn-2-ol over the
catalysts evidenced the higher activity and selectivity of these
trimers over unmodified nickel. Depending on the substrate
ChemCatChem 2019, 11, 457 – 464

www.chemcatchem.org

Experimental Section
Materials: Ni5P4 was synthesized by mixing nickel powder (Sigma
Aldrich, 99 %) and red phosphorus powder (Sigma Aldrich, 99 %) in a
mortar with a P / Ni molar ratio of 2, and heating the resulting mixture
under a nitrogen atmosphere at 773 K, respectively. Ni2P was
synthesized using the method outlined by Guan et al.[30] with nickel
chloride hexahydrate and sodium hypophosphite precursors and a P/
Ni ratio of 5. The reference 60 wt.% Ni/SiO2À Al2O3 (Alfa Aesar, ref:
7440-02-0) catalyst was used as received. Powder X-ray diffraction
(XRD) was carried out using a PANalytical X’Pert PRO-MPD diffractometer with CuÀ Kα radiation (λ = 0.154 nm). Data was recorded in the
10–70° 2θ range with an angular step size of 0.033° and a counting
time of 8 s per step. Solid-state 31P magic angle spinning nuclear
magnetic resonance (31P MAS-NMR) spectroscopy was conducted by a
Bruker Advance III HD spectrometer equipped with a 9.4 T magnet
(Larmor frequency fL = 400.13 MHz). The spectra were recorded by a
2.5 mm double resonance MAS probe at a spinning speed of 10 kHz.
X-ray fluorescence spectroscopy (XRF) was used to quantify the Ni and
P contents by using an Orbis Micro instrument equipped with an Rh
source operated at 35 kV and 500 mA. Nitrogen isotherms were
obtained at 77 K using a Micromeritics Tristar instrument, after
evacuation of the samples at 473 K for 3 h.
Catalytic tests: Continuous-flow hydrogenations of 2-methyl-3butyn-2-ol (Sigma Aldrich, 98 %) and 1-hexyne (Acros Organics,
98 %) were carried out in a fully-automated flooded-bed microreactor (ThalesNano H-Cube Pro™), in which molecular hydrogen
(produced in situ via water electrolysis) and the liquid feed
(introduced by an HPLC pump, Knauer) flow concurrently upward
through the catalyst bed. Note that all catalysts were sieved in the
particle size range 0.2–0.4 mm prior to the testing. The reaction
products were analyzed offline using a gas chromatograph (HP6890) equipped with a HP-5 capillary column and a flame ionization
detector. The conversion (X) of the reactant, which ranges from 5 to
100 % as shown in Figure S5, was determined as the amount of
reacted alkyne divided by the amount of alkyne at the reactor inlet;
the selectivity (S) to a given product was quantified as the amount
of that compound divided by the total amount of reacted alkyne.
Computational details: Density Functional Theory (DFT) calculations were performed using the VASP code.[36,37] The functional of
choice was Pedrew-Burke-Ernzenhof (PBE).[38] The inner electrons
were replaced by projected-augmented wave (PAW)[39] pseudopotentials with a kinetic cutoff energy of 600 eV for bulk optimizations. The Γ-centered k-point grids were sampled according to the
Monkhorst-Pack (MP)[40] scheme and was denser than 0.3 ÅÀ 1
(Table S1). In case of Nickel, spin polarization is considered. For

463

© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57

NixPy systems, spin and non-spin polarization calculations had been
compared and the results showed that there is no magnetization in
NixPy systems, thus the same structures were simulated without
considering spin polarization. The calculated lattice parameters
agree well with experiment and previous DFT results (Table S1). In
the slab calculations, the cutoff energy was 450 eV, Γ-centered kpoint samplings were also denser than 0.3 ÅÀ 1. Vacuum length for
all of the slabs was 15 Å and half of the layers were fixed. DFT-D2
method developed by Grimme et al.[41,42] is used for the van der
Waals interaction with the C6 parameters developed in our
group.[43,44] Gas-phase molecules were relaxed in a box 20 × 20 ×
20 Å3. Segregation and islanding calculations[45] for Ni2P and Ni5P4
were performed by switching Ni and P atoms from the surface to
bulk for segregation calculations, and on the surface for islanding
calculations. Acetylene is taken as surrogate in the calculations for
three main reasons: (i) the adsorption is dominated by the alkyne
moiety, meaning that the chain only binds to the surface via van
der Waals forces; (ii) most of our/previous calculations have been
done for acetylene and thus it is easier to analyze the role of
phosphorus in the catalysts with respect to other modifiers; and (iii)
the number of configurations in the complex NixPy surfaces grows
exponentially when compared to the pure metal in the combined
search of many adsorption sites and organic fragment configurations, which are difficult to compile in a simple form. The
isomerization reaction was investigated for 1-hexyne. Climbing
Image Nudged Elastic Band (CI-NEB) approach[46,47] and the
improved dimer algorithms[48] were employed to locate the
transition states, which were further verified by their single
imaginary frequency character. Solvation (toluene) contributions to
the reaction network have also been considered via the MGCM
method.[49,50] All structures in the present study can be retrieved
from ioChem-BD.[51,52]

[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]

Acknowledgements

[33]

Financial support from ETH Zurich, and MINECO (grant number
CTQ2015-68770-R). The Barcelona Supercomputing Centre (BSCRES) is thanked for access to its facilities. Dr. René Verel is
acknowledged for conducting NMR analyses and for discussions.

Conflict of Interest

[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]

The authors declare no conflict of interest.
Keywords: alkyne semi-hydrogenation · DFT calculations ·
ensemble design · nickel phosphide · structure sensitivity

[1]
[2]
[3]
[4]

R. A. Sheldon, Green Chem. 2014, 16, 950–963.
R. A. Sheldon, Chem. Soc. Rev. 2012, 41, 1437–1451.
H. Lindlar, Helv. Chim. Acta 1952, 35, 446–450.
W. Bonrath, J. Medlock, J. Schütz, B. Wüstenberg, T. Netscher, in
Hydrogenation, InTech, 2012.
[5] M. García-Mota, J. Gómez-Díaz, G. Novell-Leruth, C. Vargas-Fuentes, L.
Bellarosa, B. Bridier, J. Pérez-Ramírez, N. López, Theor. Chem. Acc. 2011,
128, 663–673.
[6] G. Vilé, D. Albani, N. Almora-Barrios, N. López, J. Pérez-Ramírez,
ChemCatChem 2016, 8, 21–33.
[7] M. García-Mota, B. Bridier, J. Pérez-Ramírez, N. López, J. Catal. 2010, 273,
92–102.

ChemCatChem 2019, 11, 457 – 464

www.chemcatchem.org

[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]

A. J. McCue, J. A. Anderson, Front. Chem. Sci. Eng. 2015, 9, 142–153.
B. Bridier, N. López, J. Pérez-Ramírez, J. Catal. 2010, 269, 80–92.
D. L. Trimm, I. O. Y. Liu, N. W. Cant, Appl. Catal. A 2010, 374, 58–64.
B. Bridier, N. López, J. Pérez-Ramírez, Dalton Trans. 2010, 39, 8412–8419.
D. Teschner, J. Borsodi, A. Wootsch, Z. Révay, M. Hävecker, A. KnopGericke, S. D. Jackson, R. Schlögl, Science 2008, 320, 86–89.
D. Albani, M. Shahrokhi, Z. Chen, S. Mitchell, R. Hauert, N. López, J.
Pérez-Ramírez, Nat. Commun. 2018, 9, 2634.
A. J. McCue, A. Guerrero-Ruiz, C. Ramirez-Barria, I. Rodríguez-Ramos, J. A.
Anderson, J. Catal. 2017, 355, 40–52.
S. T. Oyama, T. Gott, H. Zhao, Y.-K. Lee, Catal. Today 2009, 143, 94–107.
R. Prins, M. E. Bussell, Catal. Lett. 2012, 142, 1413–1436.
C.-C. Hou, Q. Li, C.-J. Wang, C.-Y. Peng, Q.-Q. Chen, H.-F. Ye, W.-F. Fu, C.M. Che, N. López, Y. Chen, Energy Environ. Sci. 2017, 10, 1770–1776.
E. J. Popczun, J. R. McKone, C. G. Read, A. J. Biacchi, A. M. Wiltrout, N. S.
Lewis, R. E. Schaak, J. Am. Chem. Soc. 2013, 135, 9267–9270.
S. Carenco, D. Portehault, C. Boissière, N. Mézailles, C. Sanchez, Chem.
Rev. 2013, 113, 7981–8065.
F. Studt, F. Abild-Pedersen, T. Bligaard, R. Z. Sørensen, C. H. Christensen,
J. K. Nørskov, Science 2008, 320, 1320–1322.
A. Borodziński, G. C. Bond, Catal. Rev. 2006, 48, 91–144.
S. Carenco, A. Leyva-Pérez, P. Concepcion, C. Boissière, N. Mézailles, C.
Sanchez, A. Corma, Nano Today 2012, 7, 21–28.
Y. Chen, C. Li, J. Zhou, S. Zhang, D. Rao, S. He, M. Wei, D. G. Evans, X.
Duan, ACS Catal. 2015, 5, 5756–5765.
Y. Liu, A. J. McCue, C. Miao, J. Feng, D. Li, J. A. Anderson, J. Catal. 2018,
364, 406–414.
A. Renken, L. Kiwi-Minsker, in Advances in Catalysis, Vol. 53 (Eds.: B. C.
Gates, H. Knözinger), Academic Press, 2010, pp. 47–122.
K. S. Elvira, X. C. i Solvas, R. C. Wootton, A. J. deMello, Nat. Chem. 2013,
5, 905.
D. Albani, G. Vilé, M. A. Beltran Toro, R. Kaufmann, S. Mitchell, J. PérezRamírez, React. Chem. Eng. 2016, 1, 454–462.
M. N. Kashid, L. Kiwi-Minsker, Ind. Eng. Chem. Res. 2009, 48, 6465–6485.
R. García-Muelas, Q. Li, N. López, J. Phys. Chem. B 2018, 122, 672–678.
Q. Guan, W. Li, M. Zhang, K. Tao, J. Catal. 2009, 263, 1–3.
C. Stinner, Z. Tang, M. Haouas, T. Weber, R. Prins, J. Catal. 2002, 208,
456–466.
E. Bekaert, J. Bernardi, S. Boyanov, L. Moncounduit, M.-L. Doublet, M.
Ménétrier, J. Phys. Chem. 2008, 112, 20481.
X. Feng, M. Tang, S. O’Neill, Y.-Y. Hu, J. Mater. Chem. A 2018,
doi:10.1039/C8TA05433 A.
Q. Li, X. Hu, Phys. Rev. B 2006, 74, 035414.
Y. Segura, N. López, J. Pérez-Ramírez, J. Catal. 2007, 247, 383–386.
G. Kresse, J. Furthmüller, Phys. Rev. B 1996, 54, 11169–11186.
G. Kresse, J. Furthmüller, Comput. Mater. Sci. 1996, 6, 15–50.
J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865–3868.
P. E. Blöchl, Phys. Rev. B 1994, 50, 17953–17979.
H. J. Monkhorst, J. D. Pack, Phys. Rev. B 1976, 13, 5188–5192.
S. Grimme, J. Comput. Chem. 2006, 27, 1787–1799.
T. Bučko, J. Hafner, S. Lebègue, J. G. Ángyán, J. Phys. Chem. A 2010, 114,
11814–11824.
N. Almora-Barrios, G. Carchini, P. Błoński, N. López, J. Chem. Theory
Comput. 2014, 10, 5002–5009.
P. Błoński, N. López, J. Phys. Chem. C 2012, 116, 15484–15492.
N. López, C. Vargas-Fuentes, Chem. Commun. 2012, 48, 1379–1391.
G. Henkelman, B. P. Uberuaga, H. Jónsson, J. Chem. Phys. 2000, 113,
9901–9904.
G. Henkelman, H. Jónsson, J. Chem. Phys. 2000, 113, 9978–9985.
A. Heyden, A. T. Bell, F. J. Keil, J. Chem. Phys. 2005, 123, 224101.
M. Garcia-Ratés, N. López, J. Chem. Theory Comput. 2016, 12, 1331–1341.
M. Garcia-Ratés, R. García-Muelas, N. López, J. Phys. Chem. C 2017, 121,
13803–13809.
M. Álvarez-Moreno, C. de Graaf, N. López, F. Maseras, J. M. Poblet, C. Bo,
J. Chem. Inf. Model. 2015, 55, 95–103.
K. Karajovic, Q. Li, N. López, Database for Ensemble design in nickel
phosphide catalysts for alkyne semi-hydrogenation. DOI: 10.19061/
iochem-bd-1-94

Manuscript received: September 2, 2018
Version of record online: November 20, 2018

464

© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim