Synthesis and hydrodeoxygenation activity of Ni2P/C - Effect of the palladium salt on lowering the nickel phosphide synthesis temperature

Article abstract of DOI:10.1016/j.jcat.2016.05.016

The effect of the nature of Pd salt (PdCl₂, Pd(NO₃)₂ or Pd(acac)₂) in lowering Ni₂P/C synthesis temperature and their effect on guaiacol hydrodeoxygenation reaction were investigated. Pd addition led to a partial lowering of Ni_xP_yO_z/C (precursor) reduction temperature. For each Pd salt, both the decrease in synthesis temperature and the degree of reduction of Ni_xP_yO_z were different. While the latter was related to Pd particle size (the bigger the particles were, the higher was the amount of Ni_xP_yO_z reduced), the former depends both on the contact between Pd and Ni_xP_yO_z (direct contact is needed) and on Pd particle size (the smaller the particles, the higher the decrease in synthesis temperature). Reaction data revealed that Pd addition suppressed guaiacol hydrogenation. Catalysts characterization revealed that the product distribution was dependent upon the relative proportion on the two types of Ni active sites of Ni₂P.

Full text of DOI:10.1016/j.jcat.2016.05.016

Journal of Catalysis 340 (2016) 154–165
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Synthesis and hydrodeoxygenation activity of Ni2P/C – Effect of the
palladium salt on lowering the nickel phosphide synthesis temperature
Leon F. Feitosa a, Gilles Berhault b, Dorothée Laurenti b, Thomas E. Davies c, Victor Teixeira da Silva a,
a Universidade Federal do Rio de Janeiro (NUCAT/PEQ/COPPE/UFRJ), Rio de Janeiro, Brazil
b Institut de Recherches sur la Catalyse et l’Environnement de Lyon (IRCELYON/CNRS/UCBL1), Villeurbanne 69626, France
c Stephenson Institute for Renewable Energy, Chemistry Department, University of Liverpool, L69 7ZD Liverpool, UK
a r t i c l e i n f o
a b s t r a c t
Article history:
The effect of the nature of Pd salt (PdCl2, Pd(NO3)2 or Pd(acac)2) in lowering Ni2P/C synthesis temperature
and their effect on guaiacol hydrodeoxygenation reaction were investigated. Pd addition led to a partial
lowering of NixPyOz/C (precursor) reduction temperature. For each Pd salt, both the decrease in synthesis
temperature and the degree of reduction of NixPyOz were different. While the latter was related to Pd par-
ticle size (the bigger the particles were, the higher was the amount of NixPyOz reduced), the former
depends both on the contact between Pd and NixPyOz (direct contact is needed) and on Pd particle size
(the smaller the particles, the higher the decrease in synthesis temperature). Reaction data revealed that
Pd addition suppressed guaiacol hydrogenation. Catalysts characterization revealed that the product dis-
tribution was dependent upon the relative proportion on the two types of Ni active sites of Ni2P.
Ó 2016 Elsevier Inc. All rights reserved.
Received 20 March 2016
Revised 16 May 2016
Accepted 17 May 2016
Keywords:
Nickel phosphide
Carbon
Palladium
Hydrodeoxygenation
Guaiacol
1. Introduction
phides as HDT catalysts caught the attention of the scientific com-
munity [23,24]. Among the different metal phosphides, nickel
Many studies in the literature have shown the importance of
residual biomass as a renewable source of biofuels and chemicals
[1–8]. Through different processes, biomass may be converted into
solid (bio-char), liquid (bio-oil) or gaseous (bio-gas) products. For
biofuels production, bio-oil is of particular interest [9], but its high
oxygen content does not allow its direct use as fuel. Different
upgrading methodologies may be applied to reduce the bio-oil
oxygen content [10–13], with hydrotreatment (HDT) being one of
them. This technology is well developed and consolidated in oil
industry, mainly for sulfur (hydrodesulphurization – HDS) and
nitrogen (hydrodenitrogenation – HDN) removal. Despite this, its
application in oxygen removal (hydrodeoxygenation – HDO) has
received limited attention until concern about biomass-derived
fuels arose. The interest in HDO started about 10 years ago, usually
employing model oxygenated molecules as bio-oil representatives
[14–18].
phosphide (Ni2P) exhibits catalytic performances similar or
superior to that of conventional HDT catalysts in HDS and HDN
reactions [25–31]. Shu and Oyama [28] showed that carbon-
supported nickel phosphide has a better performance for HDS
and HDN reactions than silica-supported nickel phosphide and
NiMoS/Al2O3 catalysts. Besides that, it has been shown that nickel
phosphide presents good catalytic activity and has no deactivation
issues when employed for the HDO of model molecules using high
hydrogen partial pressures [20,32–35].
Despite being presented in the literature as a good alternative
for conventional HDT catalysts, transition metal phosphides syn-
thesized from the corresponding phosphates have a major disad-
vantage, which is their relatively high synthesis temperature
(ꢀ650 °C) when compared to the traditional HDT catalysts. Hence,
many studies have been carried out in an effort to synthesize tran-
sition metal phosphides at lower temperatures [36–42]. Among
them, Teixeira da Silva et al. [41] were able to lower the nickel
phosphide synthesis temperature by approximately 200 °C
through the incorporation of small amounts of PdCl2 (0.1/ 0.5 /
1.0 wt.% Pd) to the silica-supported nickel phosphate precursor.
The authors attributed the temperature lowering to the occurrence
of hydrogen spillover and activation over Pd particles during
reduction: the formed atomic hydrogen is more reactive than
molecular hydrogen and is able to reduce phosphate particles at
The main problem concerning HDO of bio-oil using conven-
tional Co(Ni)-promoted Mo(W)S2 HDT catalysts is that they pre-
sent severe deactivation [19,20]. Therefore, several studies have
been performed searching for alternative HDO catalysts
[14,21,22]. Two works published during the 90s using metal phos-
Corresponding author. Tel.: +55 21 39388344.
E-mail address: victor.teixeira@peq.coppe.ufrj.br (V. Teixeira da Silva).
http://dx.doi.org/10.1016/j.jcat.2016.05.016
0021-9517/Ó 2016 Elsevier Inc. All rights reserved.

L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
155
lower temperatures. Based on HDS catalytic activity data, they also
proposed that Pd particles were covered by the as-formed Ni2P
phase and therefore had no participation at all in the reaction.
The main objectives of this work are therefore (1) to evaluate
the effect of the Pd source (PdCl2, Pd(acac)2 or Pd(NO3)2) on lower-
ing the synthesis temperature of carbon-supported nickel phos-
phide catalysts and (2) to evaluate the obtained catalysts in
guaiacol HDO reaction as this phenolic molecule is often used as
a representative of refractory oxygenated compounds present in
bio-oils [20,43–51]. More precisely, the main reason for using gua-
iacol as a model compound lay on the fact that this molecule is a
by-product of lignin degradation. It also presents two oxygenated
functions (phenolic and methoxy) making it difficult to be com-
pletely deoxygenated.
– Vetec, PA, 37% solution) and distilled water. Pd(NO3)2 and Pd
(acac)2 were easily dissolved in water and acetone, respectively,
while PdCl2 was solubilized employing the procedure described
elsewhere [41].
Once each Pd source solution was prepared, they were used to
impregnate NixPyOz/C. Successive impregnations followed by 1 h
drying steps at 110 °C were performed until all the solution was
used. The impregnated sample was then dried overnight at
110 °C. In order to decompose the impregnated Pd salts, each sam-
ple was submitted to thermal treatment at 350 °C/2 h under pure
He flow (10 mL minꢁ1 He per gram of sample). Since it has been
concluded in previous studies done by our group that at 350 °C
each one of the palladium salts is totally decomposed, this temper-
ature was used. The obtained samples were labeled as Pd(s)–NixPy-
Oz/C, where (s) stands for the employed Pd source: (C) for PdCl2, (A)
for Pd(acac)2, and (N) for Pd(NO3)2.
2. Experimental
2.1.3. 30 wt.% Ni2P/C and 1 wt.% Pd(s) 30 wt.% Ni2P/C
2.1. Synthesis
Promoted and non-promoted nickel phosphide catalysts were
obtained by temperature-programmed reduction of their precur-
sors under pure hydrogen flow (H2 – Alphagaz, 1 mL minꢁ1 H2
per mg of precursor) and 1 °C minꢁ1 heating rate up to 650 °C/0 h
(NixPyOz/C) or 550 °C/1 h (Pd(s)–NixPyOz/C). After reduction, cata-
lysts were cooled to room temperature in inert atmosphere using
either helium (He – Alphagaz, <5.5 ppm total impurities) or argon
(Ar – Alphagaz, <5.5 ppm total impurities). Reduced samples with
and without Pd were labeled as Pd(s)–Ni2P/C and Ni2P/C, respec-
tively, with (s) denoting the palladium salt employed.
Due to the Ni2P pyrophoric nature, all reduced samples were
passivated for 2 h at room temperature under 50 mL minꢁ1 of a
2% (v/v) O2/Ar gas mixture before ex situ characterization. After
the passivation step, all samples could be manipulated in air and
stored for further characterization.
2.1.1. NixPyOz/C
Ni2P/C precursor (NixPyOz/C) synthesis was based on a two-step
procedure comprising: (1) an incipient wetness impregnation of
activated carbon support (Merck, 772 m2 gꢁ1, pore volume of
0.64 cm3 gꢁ1) with a solution prepared by dissolving adequate
amounts of nickel nitrate and ammonium hydrogen phosphate
salts and (2) thermal treatment of the impregnated sample under
inert atmosphere (He) in order to have the nickel species trans-
formed into nickel phosphate (NixPyOz). For obvious reasons, i.e.
activated carbon total oxidation, this thermal treatment could
not be done under oxidizing atmosphere. Appropriate quantities
were used to obtain catalysts with 30 wt.% Ni2P after the reduction
step.
Initially, 12.12 g of nickel(II) nitrate hexahydrate (Ni(NO3)2-
ꢂ6H2O – Vetec, PA, 97% pure) and 4.36 g of ammonium hydrogen
phosphate ((NH4)2HPO4) – Vetec, ACS, 98% pure) were each solubi-
lized in 20.5 mL of distilled water. Solution A ((NH4)2HPO4) was
then added dropwise to solution B (Ni(NO3)2ꢂ6H2O) under mag-
netic stirring. After complete addition of solution A, approximately
4 mL of nitric acid (HNO3 – Vetec, PA, 65 wt.% solution) was added
dropwise under stirring in order to solubilize the precipitate
formed after mixing solutions A and B. Thereafter, 7 g of the acti-
vated carbon support was impregnated with the prepared solution
by means of successive impregnations (impregnations until wet
point was reached intercalated with 1 h drying steps at 110 °C).
Once all the solution was incorporated into the carbon support,
the sample was dried at 110 °C overnight. The obtained solid was
then submitted to a thermal treatment at 500 °C/6 h in a quartz
reactor under He flow (10 mL minꢁ1 He per gram of sample) (He
– Linde, 99.995% pure). The obtained sample will be hereafter
referred to as NixPyOz/C.
2.2. Characterization
2.2.1. Temperature programmed reduction (TPR)
TPR analyses were carried out in a tubular quartz reactor in line
with a VG 40 Thermo quadrupole mass-spectrometer. For each
test, the reactor was loaded with 0.05 g of either NixPyOz/C or Pd
(s)–NixPyOz/C. Analysis conditions were the same for all cases: a
flow of 50 mL minꢁ1 of pure H2 and a heating rate of 1 °C minꢁ1
up to 650 °C.
The signal due to water (m/z = 18) was recorded during TPR to
follow the reduction process. Other mass to charge (m/z) signals
were also recorded: 2 (H2), 12 (C), 14 (N), 16 (CH4), 28 (CO and/
or N2), 31 (P), 32 (O2), 34 (PH3), 44 (CO2), 62 (P2). Water formation
profiles were deconvoluted using Peak Fit 4 software and each con-
tribution was simulated by means of Gauss-Lorentz functions
(65/35 Gaussian/Lorentzian weight).
2.2.2. X-ray diffraction (XRD)
2.1.2. Pd(s)–NixPyOz/C
XRD analyses of precursors and reduced catalysts (in the passi-
vated form as detailed in Section 2.1.3) were performed using a D8
Bruker diffractometer equipped with a Ni filter (Cu Ka1 radiation,
k = 1.542 Å). Powder patterns were recorded in the 4° < 2h < 90°
range with steps of 0.020°. PDF4 (2013) database was used to iden-
tify the observed crystalline phases. Rietveld refinement was also
applied to calculate Ni2P crystallite size using Topaz software.
The palladium-containing samples (Pd(s)–NixPyOz/C) synthesis
was also based on a two-step procedure comprising: (1) incipient
wetness impregnation of NixPyOz/C sample with palladium solu-
tions prepared from different salts (chloride, nitrate and acetylace-
tonate) and (2) thermal treatment under inert atmosphere. The
amount of palladium added was such in order to obtain a 1 wt.%
Pd loading after the reduction step.
Due to solubility issues, different solvents (water, acetone and
hydrochloric acid) were used for each Pd salt. Palladium nitrate
(Pd(NO3)2ꢂxH2O – Aldrich, PA) was solubilized in distilled water,
palladium acetylacetonate (Pd(C5H7O2)2 – Aldrich, PA) in acetone
((CH3)2CO – Vetec, PA, 99.5% pure) while palladium chloride (PdCl2
– Aldrich, PA, 99% pure) was dissolved with hydrochloric acid (HCl
2.2.3. High resolution transmission electron microscopy (HR-TEM)
HR-TEM analyses of non-reduced samples were performed in a
JEOL JEM-2100 microscope at 200 kV. Energy Dispersive X-ray
(EDX) analysis was performed using Oxford Instruments X-MaxN
analyzer and Aztec software. Samples were prepared by dispersion

156
L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
in methanol with sonication before supporting on holey carbon
film copper grids.
sample was collected, corresponding to reaction time zero t00 at
room temperature. Thereafter, heating was started and a new liq-
uid sample was collected when the reaction temperature (300 °C)
was reached, corresponding to reaction time zero t0. These two
time zero reaction samples were collected in order to verify
whether there was guaiacol conversion either at room temperature
or during the heating step. Thereafter, hydrogen was introduced in
order to reach a pressure of 30 bar of H2 while reaction time mea-
surement was started. Liquid samples were collected after 20 min
(t1) and then hourly up to 6 h of reaction (t2–t7). All collected sam-
ples were analyzed in a Hewlett Packard 5890 SERIES II gas chro-
matograph (GC) equipped with a flame ionization detector (FID)
2.2.4. Inductively coupled plasma-optical emission spectroscopy (ICP-
OES)
ICP was used to determine the Ni, P and Pd contents in reduced
solids. Samples were first dissolved in H2SO4 + aqua regia (Pd con-
taining samples) or H2SO4 + HNO3 (sample without Pd) at 250–
300 °C and the obtained solutions were analyzed using an ACTIVA
– Horiba Jobin Yvon spectrometer.
2.2.5. In situ CO chemisorption and CO temperature-programmed
desorption (TPD)
and a DB-5MS column (30 m ꢃ 0.32 mm ꢃ 0.25
lm). Response fac-
tors for each of the identified products were taken from the litera-
In situ CO chemisorption and TPD analyses were carried out in a
homemade multipurpose unit. A four channels Type 247 MKS mass
flow meter was used to select the gas to be used in each analysis
and the exit gases were analyzed by a QME 200 Pfeiffer mass
spectrometer.
ture, allowing quantitative analysis of the products.
Given that guaiacol is a molecule with 2 oxygenated groups and
an aromatic ring, it can be transformed into different compounds,
oxygenated or not. Therefore, conversion is expressed in this work
either as total conversion (XT) or as HDO conversion (XHDO). The
former is defined as the sum of all of the products obtained during
the test divided by the initial guaiacol, while the latter is defined as
the sum of totally deoxygenated products divided by the initial
guaiacol. Selectivity values were calculated by the ratio between
the molar fraction of each formed product (yi) and the total conver-
sion XT. According to this definition, selectivity for time zero of
reaction cannot be mathematically calculated because XT is null
at the beginning of the reaction. However, extrapolating the selec-
tivity curves to time zero allows to determine whether the primary
products are being formed through parallel (finite values of selec-
tivity for the primary products) or sequential reactions (zero selec-
tivity for secondary products).
In a typical test, a quartz reactor was loaded with 0.2 g of the
sample to be analyzed and connected to the multipurpose unit.
After reduction using the procedure described in Section 2.1.3, H2
flow was changed to He (50 mL minꢁ1) and the reduced catalyst
was cooled to room temperature. Once all signals were stabilized
at ambient temperature, CO (2.4 mL of a 20% (v/v) CO/He mixture)
was pulsed into the reactor until 3 successive peaks presented the
same intensityꢁd1enoting complete saturation of the sample. CO
uptake (
lmol gcat) was calculated using Eq. (1):
ꢀ
ꢁ
P
npulses
i¼1
Ai
nCO
ꢂ
1 ꢁ A
CO
COUPTAKE
¼
ð1Þ
mcat
where nCO (lmol) is the amount of CO injected per pulse, Ai is the
Turnover frequency (TOF) values were caꢁlc1ulated as the ratio
between guaiacol reaction rates (rGUA – mol gcat sꢁ1) and the num-
area of the m/z = 28 peak obtained for the ith pulse, ACO is the area
of m/z = 28 peak after stabilization (no CO chemisorption) and mcat
(g) is the weight of catalyst after reduction.
ber of active sites determined by CO chemisorption (COUPTAKE
–
mol gꢁca1t). For each test, the CO uptake was experimentally deter-
Once the CO signal reached stabilization, the reactor was heated
up to 1000 °C using a 15 °C minꢁ1 heating ramp in order to acquire
the CO TPD profile of the analyzed catalysts. The signal evolution of
m/z = 28 was then deconvoluted using the Peak Fit 4 software.
Based on literature data [52], it was assumed that Ni2P presents
2 different types of active sites and both components were simu-
lated by means of Gaussian functions.
mined (Section 2.2.5) while reaction rates were calculated by Eq.
(2):
n0GUA ꢂ S
rGUA
¼
ð2Þ
mcat
where n0GUA is the amount of substance of guaiacol at reaction time
zero (mol), S is the slope of XT vs. time curve at t0 (sꢁ1) and mcat (g) is
the weight of catalyst.
2.3. HDO catalytic activity tests
Guaiacol (Acros Organics, 99%) HDO was carried out in a 300 mL
Parr autoclave reactor. The reaction mixture consisted of 3 mL of
3. Results
guaiacol, 300
lL of hexadecane (Sigma–Aldrich, 99%) used as inter-
3.1. Precursor characterization
nal standard and 100 mL of dodecane (Sigma–Aldrich, 99%). All
experiments were performed at 300 °C and 30 bar (H2) during
6 h and with a catalyst weight of approximately 0.15 g.
Water formation profiles obtained during TPR of the samples Pd
(s)–NixPyOz/C are presented in Fig. 1. While the Pd-free phosphate
displays one large and asymmetrical reduction peak with maxi-
mum at 550 °C, Pd-promoted samples show an extra reduction
peak in the 350–450 °C temperature range. The additional peak is
too intense and at too low a temperature to be solely attributed
to the reduction of cationic or partially oxidized Pd (1 wt.%) so it
is reasonable to attribute this to the lowering of the NixPyOz reduc-
tion temperature as a result of the noble metal presence/or the
promotional effect of the noble metal. Contrary to what was
observed for the reduction of Pd–NixPyOz/SiO2 [41], in which only
one single peak was found, data in Fig. 1 show that the Pd effect
on lowering the reduction temperature on carbon-supported sam-
ples is only partial. Moreover, depending on the Pd salt used, the
first reduction peak is observed at different temperatures and has
different intensities. It is noteworthy that during the TPR of the
pure support (results not presented) no water formation was
In a typical experiment, 0.2 g of catalyst precursor was reduced
in a U-shaped quartz reactor with stopcock valves, using the con-
ditions described in Section 2.1.3. Once reduction was completed
and the sample cooled to room temperature, the stopcock valves
were closed and the reactor containing the reduced sample was
transferred to an inert glove box constantly fed with argon flow
and equipped with a precision balance. The autoclave and the reac-
tion mixture were also transferred to the glove box and the purging
with argon was performed over 30 min to remove residual oxygen
from the inert glove box. Thereafter, the reduced catalyst was
weighed and loaded into the autoclave. After that, the autoclave
was filled with the reaction mixture, taken out from the inert
box and connected to the Parr reactor system. After a new purging
procedure with nitrogen to remove oxygen from the dead volume
of the reactor, stirring (800 rpm) was initiated and an initial liquid

L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
157
Fig. 2. XRD patterns of NixPyOz/C and Pd(s)–NixPyOz/C samples.
orthophosphate – Ni3(PO4)2) in all samples as seen in other works
from the literature [41]. Because phosphorous excess was used in
the synthesis to guarantee Ni2P formation, then it can be assumed
that the NixPyOz is composed mainly of amorphous Ni3(PO4)2 and
small amounts of P2O5. In this way, the water formed during reduc-
tion is essentially due to the orthophosphate reduction. On the
other hand, Pd-promoted samples presented an additional peak
at 2h = 40° attributed to metallic Pd particles. This result suggests
that all of the different palladium salts were decomposed during
thermal treatment under He and underwent reduction most prob-
ably due to the reducing nature of the carbon support. Moreover,
the data show that the Pd particle size changes depending on the
salt employed. Considering the change in intensity and width of
the peak at 2h = 40°, Pd particle size increased in the order Pd(A)
< Pd(N) ꢄ Pd(C).
Fig. 1. Water formation profiles during TPR of NixPyOz/C and Pd(s)–NixPyOz/C
(reduction conditions: 1 °C minꢁ1 up to 650 °C).
observed, only CO, CO2 and CH4 formation at temperatures around
680, 450, and 670 °C, respectively. While CO and CO2 formation is
due to the decomposition of the superficial groups of the support,
CH4 formation results from carbon methanation. Consequently, the
water profiles observed in Fig. 1 are solely related to nickel phos-
phate reduction.
Deconvolution of the TPR profiles is presented in Fig. 1, where it
can be seen that the reduction profile of the non-promoted
NixPyOz/C catalyst presents three contributions. The first one can
be related to a surface reduction of the nickel phosphate particles
leading to the formation of a partially reduced nickel phosphate
layer that slows down hydrogen diffusion thus resulting in the sec-
ond and third contributions represented by the high temperature
reduction peaks. For Pd-promoted NixPyOz/C samples, the first peak
shifts to lower temperatures depending on the nature of the Pd
source and its relative area (shown in Table 1) changes according
to the type of Pd precursor. The shift of the maximum of the first
reduction peak follows the order Pd(N) > Pd(A) > Pd(C) while the
HR-TEM micrographs of the non Pd-promoted NixPyOz/C and Pd
(s)–NixPyOz/C are presented in Fig. 3. The NixPyOz particles exhibit a
single morphology corresponding to agglomerated quasi-spherical
clusters over the ‘‘orange peel” type morphology typical of poorly
crystalline or amorphous carbon (Fig. 3a inset). Pd-containing sam-
ples presented a similar morphology but upon impregnation parti-
cles of palladium can be seen intimately mixed within the NixPyOz/
C support (Fig. 3b–d). As seen in the XRD (Fig. 2) the particle sizes
increase in the order Pd(A) < Pd(N) ꢄ Pd(C) with larger Pd particles
clearly apparent in the Pd(C)–NixPyOz/C catalyst prepared from
PdCl2 (Fig. 3d).
Elemental EDX mapping results for the non-promoted and Pd-
promoted samples are presented in Fig. 4. In all cases, Ni, P and
O elements are well dispersed all over the carbon support. How-
ever, C maps show that carbon-rich regions are present in some
areas of the Pd(s)–NixPyOz/C samples implying that the carbon sur-
face was not fully covered by the NixPyOz particles. Additionally, Pd
EDX mapping images suggest that the Pd dispersion increases in
the order DPd(C) ꢄ DPd(N) < DPd(A), which is in accordance with the
XRD results concerning the Pd particle sizes (i.e., the smaller the
particles, the higher the dispersion). A comparison between Pd
and C maps also indicates that some noble metal particles are
deposited directly on the NixPyOz phase while the rest is on the car-
bon support. It is also noteworthy that EDX analysis of sample Pd
(C)–NixPyOz/C revealed the presence of chlorine in concentrations
of c.a. 0.5 wt.% (see Fig. A1 in the Supplementary Material).
change in relative area increases according to APd(A) < APd(N) < APd(C)
.
Therefore, Pd particles derived from Pd(NO3)2 led to highest shift in
the synthesis temperature while Pd particles resultant from PdCl2
were able to reduce a bigger amount of phosphate.
XRD patterns of the different precursors are presented in Fig. 2
and all of them present a large and broad peak around 26° and
small one at 45°, which are attributed to the carbon support.
Moreover, the absence of any peak due to nickel phosphate indi-
cates that this species was present as an amorphous phase (nickel
Table 1
Temperature of the maxima and relative area of the deconvoluted water formation
profiles of Fig. 1 (peaks were fitted using Gauss–Lorentzian functions).
Sample
1st Peak
2nd Peak
3rd Peak
T (°C)
Arel. (%)
T (°C)
Arel. (%)
T (°C)
Arel. (%)
3.2. Catalysts characterization
NixPyOz/C
463
443
430
420
15.0
32.7
22.3
23.2
508
490
483
491
25.9
23.0
19.8
18.3
546
533
540
535
59.1
44.3
57.9
58.5
Pd(C)–NixPyOz/C
Pd(A)–NixPyOz/C
Pd(N)–NixPyOz/C
Elemental analysis (Table 2) showed the actual Ni, P and Pd con-
centrations to be close to the theoretical values. The small excess of

158
L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
Fig. 3. NixPyOz/C HR-TEM micrographs obtained at different magnifications showing (a) NixPyOz/C, (b) Pd(N)–NixPyOz/C, (c) Pd(A)–NixPyOz/C, (d) Pd(C)–NixPyOz/C.
phosphorous observed was expected because a surplus was used to
prepare the solution. In fact, it is well known from the literature
[53] that if a stoichiometric P/Ni molar ratio of 0.5 is used an Ni-
rich phase (Ni12P5) is formed, which is inactive for HDT reactions.
To avoid the formation of this undesired phase, a 0.8 P/Ni molar
ratio was employed in this work. TPR data show the formation of
PH3 and P2 in small amounts during reduction (see Fig. A2 in Sup-
plementary Material), thus explaining why the phosphorous excess
reported in Table 2 is smaller than the expected one. Moreover, Pd-
NixPyOz/SiO2 system [41], that palladium addition indeed promotes
a decrease in synthesis temperature of nickel phosphide.
CO chemisorption and desorption data are presented in Table 3
and Fig. 6, respectively. The analysis of the table and the figure
shows some differences in terms of surface sites properties among
the reduced samples. As shown in Table 3, CO uptake values for Pd
(A)–Ni2P/C and Pd(C)–Ni2P/C are smaller when compared to those
obtained for Pd(N)–Ni2P/C and Ni2P/C. The small CO uptakes and
large crystallite sizes reported in Table 3 have been reported in
the literature [28,54,55] for systems containing high loadings of
nickel phosphide and weak interactions with the support as it is
the case with carbon. However, despite these observations the
promoted samples presented lower
promoted Ni2P/C sample.
P excess than the non-
XRD results are presented in Fig. 5 and the crystallite sizes cal-
culated using the Scherrer equation are shown in Table 3. As
observed in Fig. 5, all of the diffraction peaks are characteristic of
the Ni2P phase (JCPDS file number 03-0953) indicating that Pd
addition did not modify the nature of the nickel phosphide phase
formed. Moreover, in comparison with the non-promoted catalyst,
Ni2P crystallite sizes were smaller for Pd-promoted samples when
Pd(NO3)2 and Pd(acac)2 salts were used. Strikingly, the use of PdCl2
leads to a Ni2P crystallite size similar to the non-promoted sample.
This result shows that there is a direct relationship between the Pd
particle size in the Pd(s)–NixPyOz/C samples and the final Ni2P par-
ticle size. This implies a direct influence of the Pd particle size on
the final dispersion of the nickel phosphide active phase.
In order to confirm Pd effect in lowering Ni2P synthesis temper-
ature, the evolution of the crystalline phases as a function of the
reduction temperature was studied. NixPyOz/C and Pd(C)–NixPyOz/
C samples reduced at 400, 450, and 500 °C were analyzed by XRD
and the results are presented in Fig. A3 of the Supplementary
Material. The results show, as previously reported for the Pd(C)–
important point is that palladium addition has promoted
a
decrease of 100 °C in synthesis temperature as seen in Fig. 1.
It could be argued that it is too luxury to add 1 wt.% Pd without
increasing the CO uptake. However, in a previous work [41] it has
been shown that smaller amounts of palladium such as 0.1 or
0.5 wt.% are able to reduce the synthesis temperature of nickel
phosphide with the CO uptake being similar for the different sam-
ples. The low CO uptake is related to the fact that the final synthe-
sis temperature (550 °C) is still high enough to promote the active
phase sintering.
CO desorption data (Fig. 6) reveal that all profiles are composed
of a broad peak that can be deconvoluted in two peaks: a low tem-
perature one (A1) presenting a maximum in the 99–105 °C temper-
ature range and another peak at higher temperature (A2) with
maximum in the range 149–160 °C. Moreover, the area of the
low temperature peak is higher whatever the sample considered.
It can also be seen from Fig. 6 that the A1/A2 ratio is higher in Pd
(s)–Ni2P/C catalysts than in Ni2P/C.

L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
159
Fig. 4. NixPyOz/C and Pd(s)–NixPyOz/C EDX maps showing element distribution in selected particles.

160
L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
Table 2
sented in Fig. A4 in Supplementary Material). Benzene and
Elemental analysis data of all reduced catalysts.
cyclohexane appear as the main final products with phenol, anisole
and methoxycyclohexanol (cis and trans isomers) are the primary
products as their initial selectivity values (extrapolation to zero
conversion) leads to finite values. It should also be noted that the
selectivity ratio anisole/phenol tends to decrease when going from
Pd(C)–Ni2P/C to Pd(A)–Ni2P/C and finally Pd(N)–Ni2P/C showing a
progressive higher relative tendency to demethoxylation, particu-
larly for Pd(N)–Ni2P/C. Finally, methanol, which is a primary pro-
duct formed from the demethoxylation of guaiacol can also be
formed in other steps of the reaction network (as will be discussed
later) and because of this presents a steady selectivity value. For
low conversion values methanol is not identified in the reaction
mixture probably due its adsorption on activated carbon. All the
remaining identified products had selectivity values below 10%
and are presented in Fig. A4 of the Supplementary Material.
Comparing the Ni2P/C selectivity data with those of Pd-
promoted catalysts, a strong decrease in the methoxycyclohexanol
and cyclohexanol formation is observed, which was accompanied
by an increase in selectivity for benzene and anisole. In order to
verify if the observed changes in selectivity was due to a parallel
reaction of guaiacol over Pd sites, a 1 wt.% Pd/C catalyst (reduced
in the same conditions as those used for Pd promoted samples)
was evaluated for guaiacol HDO. As may be seen in the Supplemen-
tary Material (Fig. A5), the total conversion after 6 h was only 3.6%
and this catalyst presented a completely different product distribu-
tion with formation of only catechol, methoxycyclohexanol, phenol
and cyclohexanol.
Sample
Content (wt.%)
Ni2P contentd (wt.%)
P excesse (wt.%)
Nia
Pb
Pdc
Ni2P/C
26.3
27.5
26.5
28.3
7.4
7.5
7.2
7.7
–
33.2
34.7
33.5
35.7
0.4
0.2
0.2
0.2
Pd(C)–Ni2P/C
Pd(A)–Ni2P/C
Pd(N)–Ni2P/C
0.9
1.3
1.1
a
b
c
Theoretical value: 23.7%.
Theoretical value: 6.3%.
Theoretical value: 1.0%.
Assuming that all Ni is as Ni2P.
Theoretical value: 3.8%.
d
e
4. Discussion
Fig. 5. XRD patterns of Ni2P/C and Pd(s)–Ni2P/C samples.
As shown in Fig. 1, the effect of Pd in decreasing the Ni2P syn-
thesis temperature was only partial, contrary to what was reported
in the literature for silica-supported systems [41]. Comparing both
cases, the most probable explanation for such a change in behavior
is the difference in the specific surface area values of the supports:
while the activated carbon used in the present work has a specific
surface area of 772 m2 gꢁ1, the specific surface area of the silica
support used in Ref. [41] was 205 m2 gꢁ1. Since the Ni2P loading
is 30 wt.% in both studies, it is reasonable to assume that due to
the differences in specific surface area, the surface coverage by
the NixPyOz particles prior to the reduction step is different in both
works, being lower in the present study.
Table 3
Ni2P crystallite sizes and CO uptake values of Ni2P/C and Pd(s)–Ni2P/C samples.
Ni2P crystallite size (nm)a
CO uptake (
l
mol g
)
cat
ꢁ1
Sample
Ni2P/C
45
48
38
35
19
12
15
19
Pd(C)–Ni2P/C
Pd(A)–Ni2P/C
Pd(N)–Ni2P/C
a
Calculated using Rietveld refinement.
3.3. HDO catalytic activity evaluation
If one considers the specific surface areas values of the carbon
support of this study (772 m2 gꢁ1) and of silica (205 m2 gꢁ1) [41],
the Ni2P loading necessary to form a theoretical monolayer in each
of the supports can be determined. While this value corresponds to
34 wt.% for SiO2, a 66 wt.% Ni2P loading would be necessary for C.
Given that both catalysts have the same nickel phosphide loading
(30 wt.%), then it can be concluded that while for SiO2 the surface
covering by Ni2P was almost complete, for the carbon support the
covering was only partial. This hypothesis is confirmed by the EDX
data presented in Fig. 4. In fact, EDX mapping shows the presence
of high carbon concentration regions for all samples confirming the
hypothesis that the covering was not complete. Since the SiO2 sur-
face of the work from literature [41] is almost completely covered,
it may be assumed that in this case all Pd particles were in direct
contact with NixPyOz. However, the same assumption cannot be
made for C-supported systems because palladium distribution is
random with some Pd particles being deposited on top and/or in
close contact with the NixPyOz phase, while others are deposited
on the exposed carbon surface and without contact with the nickel
phosphate particles. EDX data support this description: a compar-
ison between Pd and C maps shows that Pd particles are present on
both carbon-poor regions (i.e., Pd in on top and/or in close contact
Reaction rates (rGUA) and TOF values were calculated for the
guaiacol HDO reaction and are presented in Table 4. It is notewor-
thy that independently of the sample, the TOF values are essen-
tially the same if one takes into consideration the experimental
error associated with the CO chemisorption and conversion mea-
surements. In fact, comparison of the reaction rate values shows
a levelling off effect between Pd-promoted samples and Ni2P/C,
with Pd(N)–Ni2P/C presenting a reaction rate 20% higher than that
of the Ni2P/C. However, taking into account that reaction rate val-
ues were calculated using the total conversion of guaiacol (XT) and
that some products can be only partially deoxygenated or not
deoxygenated at all, it is necessary also to compare the results in
terms of HDO activity. Considering this fact, XT and XHDO were com-
pared during the reaction course for each catalyst and the obtained
data are presented in Fig. 7. Since XHDO/XT curves are almost super-
imposed for all samples, it can be concluded that the HDO activity
follows essentially similar tendencies to those observed for the
total conversion.
Main products selectivity data obtained for all catalysts are pre-
sented in Fig. 8 (selectivity data of products below 10% are pre-

L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
161
Fig. 6. CO temperature-programmed desorption profiles of Ni2P/C and Pd(s)–Ni2P/C samples. Each profile was deconvoluted into two contributions, A1 and A2. The A1/A2 ratio
is given for each sample.
Table 4
due to Pd addition were directly affected by the nature of the Pd
salt. Analysis of respectively XRD and EDX allows to conclude that
the Pd crystallite size increases in the order Pd(A) < Pd(N) ꢄ Pd(C),
while the noble metal dispersion trend is Pd(C) ꢄ Pd(N) < Pd(A).
Analyzing the amounts of NixPyOz reduced and the maxima of
the first water formation peak for each of the Pd salts incorporated
into the NixPyOz/C two clear trends are observed. The first one
clearly shows that the smaller the noble metal particle size, the
higher the lowering in the maximum of the first reduction peak;
and the second trend shows that the bigger the palladium particles,
the higher the amount of phosphate reduced at lower
temperatures.
TOF values and reaction rates (rGUA) for Ni2P/C and Pd(s)–Ni2P/C samples during the
guaiacol HDO reaction at 300 °C and 30 bar (H2).
TOF (sꢁ1
)
rGUA (10ꢁ6 mol g sꢁ1
)
ꢁ1
Sample
cat
Ni2P/C
0.22
0.37
0.30
0.26
4.18
4.44
4.50
4.94
Pd(C)–Ni2P/C
Pd(A)–Ni2P/C
Pd(N)–Ni2P/C
In fact, as shown in Table 1, the first peak relative area (i.e., the
relative amount of NixPyOz reduced) increases in the order Pd(A)
< Pd(N) < Pd(C), which follows the increase of the Pd crystallite size.
It is proposed in the literature [41] that the lowering of the reduc-
tion temperature of NixPyOz due to addition of palladium takes
place in two steps: (1) hydrogen spillover and its activation over
Pd particles, which would decrease NixPyOz reduction temperature
and (2) the covering of Pd particles by Ni2P phase during reduction
(or noble metal migration to the interior of Ni2P), which would
stop the hydrogen spillover and activation with the consequent
decrease in phosphate reduction. Taking these hypotheses into
account and considering the observed relationship between Pd
particle size and the NixPyOz amount reduced by addition of Pd, it
is clear that the reduced amount of phosphate seems to be depen-
dent on how long noble metal particles remain exposed during the
reduction step. Assuming that the proposed model is also valid in
the present study, it would be reasonable to expect that bigger
Pd particles would take longer to be covered. Consequently, the
bigger the Pd particles, the longer the spillover effect would last
Fig. 7. HDO conversion (XHDO) and total conversion (XT) relationship for guaiacol
reaction for Ni2P/C and Pd(s)–Ni2P/C samples. Tendency lines are added only to
guide the eye.
resulting in
a higher amount of NixPyOz reduced at lower
with NixPyOz) and carbon-rich regions (i.e., Pd in contact with C
support). Taking into account TPR and EDX data and comparing
SiO2 and C supported systems, the present data suggest that Pd
and NixPyOz particles need to be in close contact in order for the
noble metal to be able to efficiently lower the synthesis tempera-
ture of Ni2P.
temperatures.
The differences in temperature maxima observed for the first
reduction peak as presented in Fig. 1 may be explained if Pd parti-
cle size and the degree of contact between Pd and NixPyOz particles
are considered. If it is accepted that the palladium particles pro-
mote hydrogen activation and spillover, then the formed atomic
hydrogen has to come into contact with the phosphate in order
to reduce it to phosphide. Recombination of atomic hydrogen into
Data reported in Table 1 show that both the temperature
decrease and the amount of NixPyOz reduced at lower temperature

162
L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
Fig. 8. Main products selectivity data as a function of total conversion (XT) for Ni2P, Pd(C)–Ni2P/C, Pd(A)–Ni2P/C and Pd(N)–Ni2P/C.
molecular hydrogen directly depends on how far palladium is from
the phosphate particle leading, in extreme cases, to a complete
absence of the effect on the reduction temperature. In other words,
palladium and phosphate particles have to be in close contact in
order to be possible to observe an effect on the reduction temper-
ature. Since the Pd loading is the same for all of the samples stud-
ied in this work, a better dispersion implies that smaller particles
were formed and, consequently, that a larger number of Pd parti-
cles were obtained. The probability for one of these Pd particles
to be in direct contact with NixPyOz is consequently higher, thus
explaining why PdCl2, with the lowest dispersion, was a less effec-
tive precursor in terms of temperature lowering. This also seems to
directly influence the final Ni2P particle size since a higher disper-
sion of Pd leads to a lower Ni2P particle size. Finally, one should
note that if the noble metal dispersion was homogeneous on the
overall exposed surface (C surface + NixPyOz surface), contact
degree would be directly related to noble metal dispersion and
particle size. However, depending on the respective Pd interaction
with both C and NixPyOz, a preferential deposition over one surface
or another would be possible depending on the Pd salt used. This
last hypothesis may explain why Pd(NO3)2 presents the best tem-
perature lowering while Pd(acac)2 shows the best dispersion. A
schematic representation summarizing the Pd effect on lowering
the NixPyOz/C reduction temperature is presented in Fig. 9.
verted via three different routes: (1) demethoxylation (DMO) to
phenol, (2) direct deoxygenation (DDO) to anisole and (3) hydro-
genation (HYD) to methoxycyclohexanol. In subsequent steps, dif-
ferent additional routes would lead to the formation of benzene
and cyclohexane as the main final products and to some additional
compounds in minor amounts (selectivity below 10%). For Ni2P/C,
guaiacol was initially converted via DMO, DDO and HYD in similar
proportions and cyclohexanol appeared as a significant additional
intermediate. With addition of Pd, the observed decrease in selec-
tivity to methoxycyclohexanol and cyclohexanol showed that the
HYD route was remarkably suppressed on the route forming these
compounds. This fact was followed by an increase of the selectivity
for anisole and benzene, showing that the DDO route was herein
favored at the expense of the HYD pathway for Pd-containing
samples.
Considering that benzene and cyclohexane are guaiacol main
HDO final products, the evaluation of the benzene/cyclohexane
ratio is a simple way to compare the hydrogenating character of
the different catalysts. As shown in Fig. 11, benzene/cyclohexane
ratios are higher for Pd promoted samples than for the non-
promoted catalyst, thus confirming the assumption that Pd addi-
tion really suppressed the Ni2P/C hydrogenating character. Besides,
the benzene/cyclohexane ratios presented in Fig. 11 show that for
low guaiacol conversion the formation of benzene for Ni2P/C is zero
allowing to conclude that for this catalyst the HYD route through
cis/trans methoxycyclohexanol formation is favored. On the other
hand, reaction on the palladium containing catalysts seems to be
favored through the guaiacol DDO or DMO pathways forming
With respect to the reaction data, a significant difference was
observed in products selectivity. A general scheme for the guaiacol
HDO reaction over Ni2P/C is proposed in Fig. 10, based on selectiv-
ity data reported in Fig. 8. Basically, guaiacol may be initially con-

L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
163
Fig. 9. Schematic representation of the Pd effect on lowering the NixPyOz/C reduction temperature: smaller particles promote a more extensive decrease in synthesis
temperature while bigger particles promote a higher reduction.
Fig. 10. Proposed reaction scheme for guaiacol HDO over Ni2P/C and Pd(s)–Ni2P/C. Guaiacol is represented in green, main reaction products are represented in red and
additional main products for the non-promoted sample are represented in purple.
It could be argued that the participation of Pd active sites in
guaiacol HDO reaction could be the cause for the benzene/cyclo-
hexane ratios observed in Fig. 11 for the palladium containing cat-
alysts. However, this hypothesis can be excluded because the
catalytic test using 1 wt.% Pd/C leads to a final conversion below
4% and the selectivity to either anisole or benzene was null (see
Fig. A4 in Supplementary Material). Contrary to some reports from
the literature [56,57,50] that show that Pd/C is a good catalyst for
the guaiacol HDO reaction, the very low conversion attained by the
1 wt.% Pd/C in this work can be explained by the very high reduc-
tion temperature employed (550 °C/1 h). This temperature was
employed to allow the direct comparison with the palladium con-
taining catalysts that were reduced in this condition.
If palladium had any role on the activity of the noble metal con-
taining catalysts, then one would expect to find catechol (interme-
diate product obtained by demethylation of guaiacol) in the
Fig. 11. Benzene/cyclohexane molar ratios in function of the total conversion for
reaction products. Because this compound was not detected, then
Ni2P/C and Pd(s)–Ni2P/C catalysts for the HDO of guaiacol.
the hypothesis that Pd particles in direct contact to NixPyOz are cov-
ered during TPR seems to be in fact possible. Supposing that iso-
anisole and phenol, respectively. Subsequently, anisole and phenol
lead to benzene formation through DMO and DDO reactions as
depicted in Fig. 10. It is noteworthy that with the increase in con-
version the benzene/cyclohexane ratio decreases for the palladium
containing catalysts suggesting that the hydrogenation pathway
occurs with increasing extent.
lated Pd particles are not recovered, one could expect catechol
formation in Pd(s)–Ni2P/C catalytic tests. However, because the
obtained conversion with 1 wt.% Pd was too low, the remaining
(non-covered) amount of Pd probably presented an even lower
activity. Consequently, products due to non-covered Pd in Pd(s)–
Ni2P/C samples, if produced, were below the detection range.

164
L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
Based on Ni2P crystalline structure and EXAFS data, Oyama and
Lee [52] reported the existence of two types of Ni sites on Ni2P:
The ‘‘two-active-site” model proposed here cannot be used to
explain the differences in selectivity between anisole and phenol
since these compounds result respectively from DDO and DMO
steps directly from guaiacol. The two types of nickel sites observed
by Oyama and Lee [52] were considered to be responsible for
changes in the hydrogenating character of nickel phosphide cata-
lysts and can be envisaged here for the only real significant hydro-
genation step occurring on all catalysts (containing or not Pd) and
corresponding to the benzene-cyclohexane transformation.
It is also clear that the deconvolution into two components of
the CO TPD sites remains quite approximate due to the relatively
low signal to noise ratio in Fig. 6. Therefore, the change of the
A1/A2 ratio from 1.68 to 1.73 for the different Pd(s)–Ni2P/C catalysts
cannot be considered as enough significant to finely correlate these
data with the changes of the benzene/cyclohexane ratio in Fig. 11.
However, the change of the A1/A2 ratio going down to 1.55 for Ni2P/
C explains in a satisfactory way the more marked modification of
the benzene/cyclohexane ratio when comparing Ni2P/C with the
series of Pd(s)–Ni2P/C catalysts.
ꢅ Ni(1), which are quasi-tetrahedral sites surrounded by
4
nearest-neighbor P atoms and 8 s-nearest neighbor Ni atoms
and;
ꢅ Ni(2), which are square pyramidal sites surrounded by 5
nearest-neighbor P atoms and 6 next-nearest neighbor Ni
atoms.
Ni(1) has a lower coordination number and was believed to be
the site related to direct desulphurization (DDS) route in HDS,
while Ni(2) presents a higher coordination number and was the
site associated with the HYD route. The authors also reported that
Ni/P molar ratios after reduction were smaller for samples contain-
ing more Ni(2) sites.
Deconvolution of the CO TPD profiles (Fig. 6) resulted in two
peaks, each one related to different active sites. In all cases, the
active sites associated with CO desorption at lower temperatures
were more abundant than those associated with CO desorption
at higher temperatures. This result is quite important because it
implies that by using a relative simple technique as CO desorption
is possible to determine the relative proportion between Type 1
and Type 2 sites.
5. Conclusions
The present study about the effect of Pd on lowering the synthe-
sis temperature of Ni2P shows that both the contact degree
between Pd and NixPyOz particles and the Pd particle size are of
fundamental importance to understand how Pd promotes the low-
ering in synthesis temperature of nickel phosphate. In this respect,
Pd is able to decrease the NixPyOz reduction temperature only if the
noble metal and the phosphate are in close contact. Moreover,
depending on the nature of the Pd salt, a change in Pd particle size
is observed and this parameter directly influences NixPyOz reduc-
tion conditions. The change in Pd particle size has opposite effects
on the temperature lowering for obtaining Ni2P and on the reduced
amount of NixPyOz: while the increase in Pd particle size improves
the amount of NixPyOz reduced by Pd addition, it leads to a lower
shift of the synthesis temperature. The effectiveness in decreasing
reduction temperatures also directly influences the Ni2P particle
size and then the intrinsic catalytic activity of nickel phosphide.
Pd addition also modifies selectivity properties. A strong suppres-
sion of HYD route intermediates (methoxycyclohexanol and cyclo-
hexanol) was observed on Pd-promoted catalysts. Taking into
account CO temperature-programmed desorption and elemental
analysis data, the change in selectivity could be associated with
an increase of Ni(1) sites, which are responsible for the HDO route.
The differences in CO desorption temperature may be inter-
preted as being due to differences in the active site – CO bonding
strength. In fact, metal – CO bonding is complex and involves
charge donations between CO molecular orbitals and metal
orbitals [58,59] and it can be approximated as a charge donation
from the CO 5
r molecular orbital to the metal d orbital and to a
charge backdonation from the metal d orbital to the CO 2p⁄ molec-
ular orbital. Therefore, it would be reasonable to expect that active
sites having more electrons (sites less positively charged) would
promote charge backdonation in a greater extent and, conse-
quently, would bond more strongly with the CO molecule. Taking
this into account, it may be suggested that the first desorption peak
in CO TPD profiles is related to Ni sites more positively charged
(Ni(1)) while second desorption peak is associated with Ni sites
less positively charged (Ni(2)). If Ni(1) type sites are weaker than
Ni(2) type sites, then it may be proposed that they are related to
the first and second desorption peaks of the CO TPD profiles,
respectively.
Oyama and Lee [52] reported that the amount of Ni(1) and Ni(2)
depends on the particle size, with the former being predominant in
bigger particles and the latter prevalent on small particles. The fact
that all of the CO TPD profiles present two desorption peaks sug-
gests that all samples have particles with different sizes, small
and big. In fact, TEM results of the reduced and passivated Ni2P/C
catalyst reveal that there are particles in a wide range of sizes
(please see Fig. A6 of the Supplementary Material). This broad dis-
tribution in particle size is quite common in nickel phosphide cat-
alysts with high loadings [54,60,61].
Fig. 6 shows that Pd addition induces a slight increase in the rel-
ative quantity of Ni(1) sites. This assumption is in accordance with
the small decrease observed in P excess after TPR, once Ni(1) sites
present a lower coordination number. Finally, if Ni(1) sites are
associated with the DDS route in HDS reactions, they could also
be considered by analogy to be related to the DDO route in HDO
reactions. In other words, CO TPD profiles and elemental analysis
data suggest that Pd addition increases the relative amount of Ni
sites associated with the DDO route at the expense of the HYD
route, which is in accordance with the strong suppression of the
hydrogenating character observed for Pd-promoted catalysts. Note,
however, that the significant low signal to noise ratio in Fig. 6 does
not allow to precisely compare changes in the benzene/cyclohex-
ane ratio among the different Pd(s)–Ni2P/C catalysts.
Acknowledgments
The authors thank Conselho Nacional de Desenvolvimento
Científico e Tecnológico (CNPq) for financial support. Thomas
Davies would like to thank the Research Complex at Harwell for
facilities access through the UK Catalysis Hub funded by EPSRC
(portfolio grants EP/K014706/1, EP/K014668/1, EP/K014854/1 and
EP/K014714/1).
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2016.05.016.
References
[1] A.V. Bridgwater, Chem. Eng. J. 91 (2003) 87.
[2] G.W. Huber, S. Iborra, A. Corma, Chem. Rev. 106 (2006) 4044.
[3] A. Corma, S. Iborra, A. Velty, Chem. Rev. 107 (2007) 2411.
[4] P. O’Connor, in: K.S. Triantafyllidis, A.A. Lappas, M. Stöcker (Eds.), The Role of
Catalysis for the Sustainable Production of Bio-Fuels and Bio-Chemicals,
Elsevier, Oxford, 2013, p. 1.

L.F. Feitosa et al. / Journal of Catalysis 340 (2016) 154–165
165
[5] I.M. Lokman, U. Rashid, R. Yunus, Y.H. Taufiq-Yap, Catal. Rev. 56 (2014) 187.
[35] J. Chen, H. Shi, L. Li, K. Li, Appl. Catal. B: Environ. 144 (2014) 870.
[36] S. Yang, R. Prins, Chem. Commun. 33 (2005) 4178–4180.
[37] S. Yang, C. Liang, R. Prins, J. Catal. 241 (2006) 465.
[38] S. Carenco, I. Resa, X. Le Goff, P. Le Floch, N. Mézailles, Chem. Commun. 22
(2008) 2568.
[39] J. Guan, Y. Wang, M. Qin, Y. Yang, X. Li, A. Wang, J. Solid State Chem. 182 (2009)
1550.
[40] Q. Guan, W. Li, M. Zhang, K. Tao, J. Catal. 263 (2009) 1.
[41] V. Teixeira da Silva, L.A. Sousa, R.M. Amorim, L. Andrini, S.J.A. Figueroa, F.G.
Requejo, F.C. Vicentini, J. Catal. 279 (2011) 88.
[42] P. Bui, J.A. Cecilia, S.T. Oyama, A. Takagaki, A. Infantes-Molina, H. Zhao, D. Li, E.
Rodríguez-Castellón, A.J. López, J. Catal. 294 (2012) 184.
[43] V.N. Bui, G. Toussaint, D. Laurenti, C. Mirodatos, C. Geantet, Catal. Today 143
(2009) 172.
[44] V.N. Bui, D. Laurenti, P. Delichère, C. Geantet, Appl. Catal. B: Environ. 101
(2011) 246.
[6] M. Besson, P. Gallezot, C. Pinel, Chem. Rev. 114 (2014) 1827.
[7] S.K. Maity, Renew. Sust. Energy Rev. 43 (2015) 1427.
[8] Y. Shena, J. Wang, X. Gea, M. Chen, Renew. Sust. Energy Rev. 59 (2016) 1246.
[9] S. Czernik, A.V. Bridgwater, Energy Fuels 18 (2004) 590.
[10] S. Xiu, A. Shahbazi, Renew. Sust. Energy Rev. 16 (2012) 4406.
[11] D.A. Ruddy, J.A. Schaidle, J.R. Ferrell III, J. Wang, L. Moens, J.E. Hensley, Green
Chem. 16 (2014) 454.
[12] Z. Yang, A. Kumarn, R.L. Huhnke, Renew. Sust. Energy Rev. 50 (2015) 859.
[13] N. Koike, S. Hosokai, A. Takagaki, S. Nishimura, R. Kikuchi, K. Ebitani, Y. Suzuki,
S.T. Oyama, J. Catal. 333 (2016) 115.
[14] D.C. Elliot, Energy Fuels 21 (2007) 1792.
[15] P.M. de Souza, R.C. Rabelo-Neto, L.E.P. Borges, G. Jacobs, B.H. Davis, U.M.
Graham, D.E. Resasco, F.B. Noronha, ACS Catal. 5 (2015) 7385.
[16] K. Lee, G.H. Gu, C.A. Mullen, A.A. Boateng, D.G. Vlachos, ChemSusChem 8
(2015) 315.
[17] S.M. Schimming, O.D. LaMont, M. König, A.K. Rogers, A.D. D’Amico, M.M. Yung,
C. Sievers, ChemSusChem 8 (2015) 2073.
[45] C. Sepúlveda, N. Escalona, R. García, D. Laurenti, M. Vrinat, Catal. Today 195
(2012) 101.
[18] P.M. Mortensen, J.-D. Grunwaldt, P.A. Jensen, A.D. Jensen, Catal. Today 259
(2016) 277.
[46] C.R. Lee, J.S. Yoon, Y.-W. Suh, J.-W. Choi, J.-M. Ha, D.J. Suh, Y.-K. Park, Catal.
Commun. 17 (2012) 54.
[19] E.G. Baker, D.C. Elliott, in: E.J. Soltes, T.A. Milne (Eds.), Pyrolysis Oils from
Biomass, American Chemical Society, 1988, p. 228.
[20] H.Y. Zhao, D. Li, P. Bui, S.T. Oyama, Appl. Catal. A: Gen. 391 (2011) 305.
[21] E. Furimsky, Catal. Today 217 (2013) 13.
[47] M.V. Bykova, D.Y. Ermakov, V.V. Kaichev, O.A. Bulavchenko, A.A. Saraev, M.Y.
Lebedev, V.A. Yakovlev, Appl. Catal. B: Environ. 113–114 (2012) 296.
[48] J. Sun, A.M. Karim, H. Zhang, L. Kovarik, X.S. Li, A.J. Hensley, J.-S. McEwen, Y.
Wang, J. Catal. 306 (2013) 47.
[22] H. Wang, J. Male, Y. Wang, ACS Catal. 3 (2013) 1047.
[23] W.R.A.M. Robinson, J.N.M. Van Gestel, T.I. Korányi, S. Eijsbouts, A.M. Van Der
Kraan, J.A.R. Van Veen, V.H.J. De Beer, J. Catal. 161 (1996) 539.
[24] W. Li, B. Dhandapani, S.T. Oyama, Chem. Lett. 3 (1998) 207.
[25] X. Wang, P. Clark, S.T. Oyama, J. Catal. 208 (2002) 321.
[26] J.A. Rodriguez, J.Y. Kim, J.C. Hanson, S.J. Sawhill, M.E. Bussell, J. Phys. Chem. B
107 (2003) 6276.
[49] M.V. Bykova, S.G. Zavarukhin, L.I. Trusov, V.A. Yakovlev, Kinetics Catal. 54
(2013) 40.
[50] D. Gao, C. Schweitzer, H.T. Hwang, A. Varma, Ind. Eng. Chem. Res. 53 (2014)
18658.
[51] S.T. Oyama, T. Onkawa, A. Takagaki, R. Kikuchi, S. Hosokai, Y. Suzuki, K.K.
Bando, Top. Catal. 58 (2015) 201.
[52] S.T. Oyama, Y.-K. Lee, J. Catal. 258 (2008) 393.
[53] C. Stinner, Z. Tang, M. Haouas, T. Weber, R. Prins, J. Catal. 208 (2002) 456.
[54] H. Xin, K. Guo, D. Li, H. Yang, C. Hu, Appl. Catal. B: Environ. 187 (2016) 375.
[55] L.-N. Ding, A.-Q. Wang, M.-Y. Zheng, T. Zhang, ChemSusChem 3 (2010) 818–
821.
[27] S.T. Oyama, J. Catal. 216 (2003) 343.
[28] Y. Shu, S.T. Oyama, Carbon 43 (2005) 1517.
[29] S.T. Oyama, T. Gott, H. Zhao, Y.K. Lee, Catal. Today 143 (2009) 94.
[30] A. Infantes-Molina, J.A. Cecilia, B. Paweleca, J.L.G. Fierro, E. Rodríguez-
Castellón, A. Jiménez-López, Appl. Catal. A: Gen. 390 (2010) 253.
[31] R. Prins, M.E. Bussell, Catal. Lett. 142 (12) (2012) 1413.
[32] K. Li, J. Wang, J. Chen, Energy Fuels 25 (2011) 854.
[33] S.K. Wu, P.C. Lai, Y.C. Lin, H.P. Wan, H.T. Lee, Y.H. Chang, ACS Sust. Chem. Eng. 1
(2013) 349.
[56] A. Gutierrez, R.K. Kaila, M.L. Honkela, R. Slioor, A.O.I. Krause, Catal. Today 147
(2009) 239.
[57] D.C. Elliott, T.R. Hart, Energy Fuels 23 (2009) 631.
[58] P. Hu, D.A. King, M.-H. Lee, M.C. Payne, Chem. Phys. Lett. 246 (1995) 73.
[59] B. Hammer, Y. Morikawa, J.K. Nørskov, Phys. Rev. Lett. 76 (1996) 2141.
[60] Y. Pa, Y. Liu, C. Liu, J. Power Sources 285 (2015) 169.
[61] Q. Guan, F. Wan, F. Han, Z. Liu, W. Li, Catal. Today 259 (2016) 467.
[34] J.A. Cecilia, A. Infantes-Molina, E. Rodríguez-Castellón, A. Jiménez-López, S.T.
Oyama, Appl. Catal. B: Environ. 136–137 (2013) 140.