Palladium Nanoparticle Size Effect in Hydrodesulfurization of 4,6-Dimethyldibenzothiophene (4,6-DMDBT)

Article abstract of DOI:10.1002/cctc.201600509

Pd nanoparticle size sensitivity of 4,6-dimethyldibenzothiophene (4,6-DMDBT) hydrodesulfurization was investigated by using 4, 8, 13, and 87 nm particles and was compared with the sulfur-free and sulfur-inhibited hydrogenation of 3,3-dimethylbiphenyl, which is a product of direct desulfurization of 4,6-DMDBT. The smallest 4 nm particles provided unprecedented (for Pd at 5 MPa and 300 °C) direct desulfurization selectivity of 20 % at 40 % conversion because of the reduced contribution of the hydrogenation path. The 4 nm particles were poisoned by the adsorbed sulfur to the greatest extent. The optimal size, providing the highest Pd mass-based yield of the desulfurized products, was found to be 8 nm. The catalyst with 87 nm particles was based on Pd nanocubes with the lowest edge/terrace surface atom ratio and large terraces and this showed the lowest sulfur extraction from both 4,6-DMDBT and sulfurous intermediates as a result of the low availability of edge atoms for a perpendicular sigma-mode adsorption through the lone pair of the sulfur atom.

Full text of DOI:10.1002/cctc.201600509

DOI: 10.1002/cctc.201600509
Full Papers
Palladium Nanoparticle Size Effect in Hydrodesulfurization
of 4,6-Dimethyldibenzothiophene (4,6-DMDBT)
Jing Shen and Natalia Semagina*[a]
Pd nanoparticle size sensitivity of 4,6-dimethyldibenzothio-
phene (4,6-DMDBT) hydrodesulfurization was investigated by
using 4, 8, 13, and 87 nm particles and was compared with the
sulfur-free and sulfur-inhibited hydrogenation of 3,3-dimethyl-
biphenyl, which is a product of direct desulfurization of 4,6-
DMDBT. The smallest 4 nm particles provided unprecedented
(for Pd at 5 MPa and 3008C) direct desulfurization selectivity of
20% at 40% conversion because of the reduced contribution
of the hydrogenation path. The 4 nm particles were poisoned
by the adsorbed sulfur to the greatest extent. The optimal size,
providing the highest Pd mass-based yield of the desulfurized
products, was found to be 8 nm. The catalyst with 87 nm parti-
cles was based on Pd nanocubes with the lowest edge/terrace
surface atom ratio and large terraces and this showed the
lowest sulfur extraction from both 4,6-DMDBT and sulfurous in-
termediates as a result of the low availability of edge atoms
for a perpendicular sigma-mode adsorption through the lone
pair of the sulfur atom.
Introduction
Increasing demand for ultra-low sulfur fuels has reinvigorated
academic and industrial interest in developing active catalysts
and energy-conscious technologies that are able to bring the S
content down to 10 ppm. The enormous body of knowledge
that has accumulated in the field over many decades can deal
with targets of 200–500 ppm. On pre-reduction of the S con-
tent in fuels to 500 ppm, 20 wt% of all S species belong to
4,6-dialkyldibenzothiophenes, which undergo the slowest de-
sulfurization; that is why further S removal becomes such a dif-
ficult task.[1] The reason is steric hindrance: alkyl groups pre-
vent the perpendicular adsorption required for CÀS bond hy-
drogenolysis on a catalyst surface. The 4,6-dimethyldibenzo-
thiophene (4,6-DMDBT) hydrodesulfurization (HDS) is known to
occur through two reaction mechanisms (Scheme 1): perpen-
dicular 4,6-DMDBT s-adsorption through its sulfur atom leads
to the direct desulfurization (DDS); and flat-lying p-adsorption
through the aromatic rings results in hydrogenation (HYD) fol-
lowed by sulfur extraction.[2] The DDS selectivity for dibenzo-
thiophenes with alkyl groups in the 4 and 6 positions is very
low with predominant HYD route for HDS. The saturation of
the aromatic rings changes the spatial configuration of the sul-
furous molecules, providing access of the sulfur atom to the
catalyst surface.[3]
Scheme 1. Reaction pathways for HDS of 4,6-DMDBT at 3008C and 5 MPa on
a metal function (adapted from [6]).
the second reactor of a two-stage hydrotreating process: the
sulfur amounts entering the second reactor may be low
enough for the noble metals to maintain sufficient activity. A
previous study has documented that the catalytic activity of
noble metals decreases in the order PdꢀPtÀPd>Pt>Rh>
RuÀRh@Ru for 4,6-DMDBT hydrodesulfurization.[5]
To promote the hydrogenation of the aromatic rings, highly
active hydrogenation catalysts are required; that is why signifi-
cant efforts have been made to research noble metals known
for exceptional hydrogenation activity.[2,4] They may be used in
Catalyst performance is often affected by the active nano-
particle size. As was shown for a FeNi phosphide catalyst, low
coordination sites are particularly active in the direct desulfuri-
zation of 4,6-DMDBT,[6] which is in line with the DDS mecha-
nisms: perpendicular adsorption onto the nanoparticle edges
should provide better access of the otherwise hindered sulfur
to the metal surface. The proportion of atoms on the edges
and corners increases along with the particle dispersions. Thio-
phene conversion through the DDS pathway induced by the
perpendicular adsorption mode was found to be greater on
[a] Dr. J. Shen, Prof. N. Semagina
Department of Chemical and Materials Engineering
University of Alberta
9211, 116 Street, Edmonton, Alberta, T6G 1H9 (Canada)
E-mail: semagina@ualberta.ca
Supporting information for this article can be found under http://
dx.doi.org/10.1002/cctc.201600509.
ChemCatChem 2016, 8, 1 – 8
1
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
smaller Pt clusters than on larger ones;[7] a similar trend was
also observed with Ru dispersion changes.[8] Thiophene binds
to the metallic active sites by s-bonding through a lone pair
on the sulfur atom (h1) or p-bonding of the aromatic ring
(h4).[7] Coordinatively unsaturated corner and edge atoms,
abundant on small metal clusters, bind sulfur stronger than
the less-unsaturated terrace atoms that are prevalent on the
larger clusters. These interactions control the available metal
vacancies necessary for hydrogen and thiophene adsorption.
For Ru, for example, the size decrease from 6 to 1 nm led to
the four-fold increase in DDS/HYD turnover rates in thiophene
hydrodesulfurization.[8] Higher turnover frequencies in both
pathways during thiophene HDS were observed on Pt nano-
particles than on Ru and were related to the relatively weak
sulfur binding on Pt surfaces.[7,8] A different conclusion on the
available metal centers on sulfided Pt and Pd clusters was
drawn in ref. [9] where the 2 nm particles show the highest
sulfur tolerance, which decreases with increasing size, and the
decrease was more dramatic for Pd than for Pt. Overall, sulfur
poisoning is known to affect product distribution, such as by
enhancing dehydrogenation over hydrogenolysis[10] or boost-
ing selectivity to methanol in syngas conversion.[11] A vigorous
investigation of platinum catalysts supported on a variety of
supports (alumina, silica, and molybdena) revealed that the ob-
served support effect on the Pt sulfur poisoning could be as-
cribed to surface topography and linked to particle size rather
than to a direct support effect.[12] A recent study on the size-
dependent sulfur poisoning of Pt hydrogenation catalysts
showed that the sulfur-free structure-insensitive hydrogenation
becomes size-dependent upon thiophene addition to the
feed; in the presence of S, the turnover frequency increased
approximately four-fold upon 2 to 7 nm particle size in-
crease.[13]
and, for the first time, reports the optimal nanoparticle size for
enhanced sulfur extraction from 4,6-DMDBT.
Results and Discussion
Figure 1 depicts images and size distribution histograms for
as-synthesized Pd nanoparticles of different sizes and shapes
before deposition on the catalyst support. The particles have
a high degree of size and shape uniformity. However, Pd is
known for low sintering resistance, especially at the catalytic
application temperature (3008C). Indeed, significant sintering
occurred during the reaction, accompanied not only by size in-
crease for the spherical particles, but also the loss of the de-
fined shape for some cubic nanoparticles (Figure 2 and Sup-
porting Information).
To estimate the number of available surface palladium after
the high-temperature treatment, CO chemisorption was per-
formed. The results are presented in Table 1, and they are in
line with the TEM images of the used catalysts. Pd“S” and Pd“M”
catalysts were prepared with different polymeric stabilizers; as
nanoparticle sizes estimated by CO chemisorption and by TEM
are nearly identical, efficient surface purification from the sta-
The current study aims to investigate the size effect of palla-
dium nanoparticles on hydrodesulfurization of 4,6-DMDBT and
understand the contribution of hydrogenation sites through
3,3’-dimethylbiphenyl (3,3’-DMBP) hydrogenation with and
without sulfur addition. To verify the hypothesis of the nano-
particle edge contribution to the enhancement of the direct
desulfurization pathway, we also compared nearly spherical
particles with nanocubes that have the lowest edge/terrace
atom ratio among the different structures.[14] To exclude the in-
fluence of extremely small particles that may be co-present
with large agglomerates in traditional catalysts prepared by
support impregnation with a metal precursor, we pre-formed
Pd nanoparticles in a colloidal solution by using polymeric sta-
bilizers, followed by deposition on a support. Such an ap-
proach yields narrower size distribution as opposed to the con-
ventional catalysts and allows us to study anisotropic and
larger particle sizes that otherwise would require very high
metal loading in traditionally prepared catalysts. The feed
amount of 4,6-DMDBT corresponds to approximately
800 ppmw of sulfur (hydrogen-free) to simulate the conditions
of a potential second-stage hydrotreater for refractory sulfur
compound removal. The study demonstrates how the relative
DDS and HYD contributions in HDS of a refractory molecule
may be controlled by the Pd nanoparticle size and structure
Figure 1. TEM images of as-synthesized Pd“S” (a), Pd“M” (b), Pd“L” (c), and Pd“XL”
(d) nanoparticles and corresponding size-distribution histograms.
&
ChemCatChem 2016, 8, 1 – 8
www.chemcatchem.org
2
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Figure 2. TEM images and size-distribution histograms of the catalysts after
18 h on stream.
Table 1. Nanoparticle sizes from CO chemisorption analyses of the
supported catalysts after the high-temperature treatment.
Catalyst
Nanoparticle size[a] [nm]
Dispersion[a] [%]
Pd“S”
Pd“M”
Pd“L”
Pd“XL”
4.1
8.0
12.8
86.9
28.5
15.2
8.7
Figure 3. Catalyst performance in HDS of 4,6-DMDBT at 3008C, 50 bar H2:
(a) 4,6-DMDBT consumption rate per total mass of Pd; (b) the consumption
rates per mole of surface Pd as found in Table 1. The results are at similar
4,6-DMDBT conversions (ca. 40%).
1.3
[a] Based on the molar adsorbed CO/surface Pd ratio of 0.6.[15]
bilizers was achieved by pretreatment in air at 3508C and hy-
drogen at 3008C. The same pretreatment was used for all the
catalytic reactions.
The catalytic behavior of the four synthesized samples in
4,6-DMDBT HDS was dramatically different. Figures 3 and 4
report the catalyst activity and selectivity at the similar 4,6-
DMDBT conversions (ca. 40%). The dispersion and particle sizes
refer to the data of Table 1 (catalysts after the high-tempera-
ture treatment). The DMDBT consumption rate per total
amount of Pd in the reactor (Figure 3a) increases from the
Pd“XL” to Pd“M” sample, which is expected with the metal disper-
sion increase. However, the Pd“S” sample with ~25-fold higher
dispersion compared with the Pd“XL” catalyst shows the same
low activity per total gram of palladium, which results in the
volcano plot. If only surface Pd amount is taken into consider-
ation (Figure 3b), the activity of the largest particles and the
smallest ones differ by an order of magnitude.
The activity behavior trend is in line with product selectivi-
ties (Figure 4a) and product formation rates (Figure 4b): the
largest nanocube-derived particles Pd“XL” have the highest hy-
drogenation activity, resulting in the highest formation of the
hydrogenated intermediates. They provide the highest conver-
sion of the 4,6- DMDBT, but at the same time they are the
least active for sulfur extraction, both in HYD and DDS routes.
The large terraces and very low edge/terrace ratio for the
nanocube-derived large Pd“XL” sample likely facilitate the hydro-
genations, which are often reported as structure-sensitive reac-
Figure 4. Catalyst performance in HDS of 4,6-DMDBT at 3008C, 50 bar H2:
(a) selectivities and (b) product formation rates per gram of total Pd. The dis-
persion refers to the data of Table 1. The results are at similar 4,6-DMDBT
conversions (ca. 40%).
ChemCatChem 2016, 8, 1 – 8
www.chemcatchem.org
3
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
tions,[16] with the preference for larger nanoparticles providing
enough terrace surface for the flat reactant adsorption.[17]
The lack of edges compared with other samples corrobo-
rates with the hypothesis of facilitated direct desulfurization
on low-coordination active sites. As the size decreases from
Pd“XL” to Pd“M”, the DMDBT conversion increases, but the effi-
ciently formed hydrogenated intermediates convert more
easily to the sulfur-free products because of the increased
edge/terrace proportion; the direct desulfurization activity also
increases. The above changes occur for average particle sizes
in the range 87 to 8 nm. The most active particles (on a total
mass basis, Figure 3a and Figure 4b) for the DMDBT conver-
sion and sulfur-free product formation are the 8 nm particles
because of the seemingly optimal terrace/edge proportions
and high metal dispersion.
Al2O3 support (Figure 4a for dispersions of 1%, 9%, and 15%).
Thus, the support effect in terms of its acidity may be consid-
ered negligible for the present size-effect study, although not
completely excluded for the catalyst with the highest disper-
sion.
The observed trend for the selectivity and surface activity
change for Pd“S”, Pd“M”, and Pd“L” catalysts cannot be directly ex-
plained by the change in the number of edge/corner and ter-
race atoms: as known from surface crystal statistics, at such Pd
particle sizes (>4 nm), no significant changes in the relative
proportion of different surface atoms are expected for near-
spherical (cuboctahedral) particles.[14b] The dramatic difference
between the reactivities of the 13 and 87 nm particles could
be a result of their structure: the cuboctahedral 13 nm parti-
cles possess mostly (111) terraces with some contribution from
(100) surfaces, whereas the nanocubes used to prepare the
87 nm particles display only (100) terraces. The cubes also
possess a larger terrace/edge ratio than cuboctahedral parti-
cles.
The smallest 4 nm particles, Pd“S”, were the least active in
the overall DMDBT conversion, both based on the total and
surface Pd amounts, and higher Pd loading in the reactor had
to be used to ensure similar DMDBT conversion for a fair selec-
tivity comparison. The observed ~20% DDS selectivity is un-
precedented, to the best of our knowledge, for 4,6-DMDBT
HDS on monometallic Pd catalysts. However, it does not indi-
cate improved direct desulfurization activity, but rather that
the hydrogenation rates are reduced to a greater extent than
the drop in desulfurization activity with the particle size de-
creasing to Pd“S”. As Figure 4b shows, the DDS rates for Pd“M”
and Pd“S” catalysts per total gram of Pd are practically the
same, but the metal dispersion is twice as high for Pd“S”, indi-
cating that the surface Pd DDS activity on the smaller particles
is twice as low, with hydrogenation rates reduced almost by
an order of magnitude. This gives rise to the observed appar-
ent DDS selectivity improvement. Overall, per total gram of Pd,
the Pd“S” catalyst with 4 nm particles behaves very similarly in
4,6-DMDBT HDS as the largest 87 nm nanocube-derived parti-
cles. The only difference is that the latter shows the lowest
DDS selectivity because of the lowest edge/terrace ratio on
the nanocube-derived sample compared with all other cuboc-
tahedral particles. The smallest particles, thus, possess the
lowest HYD and DDS activity, with HYD suppressed to a greater
extent than the DDS compared with the larger particles.
Thus, there is another size-dependent factor that affects the
number of available atoms for HYD and DDS routes, which is
suggested to be structure-sensitive sulfur poisoning. The
p(H2S)/p(H2) ratio of 10À4 used in this work suggests the ther-
modynamic stability of metallic Pd, which was shown to be
stable at a H2S/H2 ratio below 0.01 at 600 K.[4c] Sulfur com-
pounds at such low concentrations, thus, compete with sulfur-
free molecules for metallic active sites and are referred to as
“labile sulfur” participating in the reaction.[4d] Sulfur association
with Pd can been seen from the elemental mapping per-
formed on the spent Pd“M” catalyst (Figure 5).
Hydrodesulfurization rates and product distribution can also
be affected by the catalyst support acidity, as was shown for
4,6-DMDBT HDS over zeolite- and alumina-supported Pd and
Pt catalysts.[4e] For palladium, with a strong hydrogenation abil-
ity, no products of acid-catalyzed isomerization were observed
even for the strongly acidic zeolite.[4e] The four-fold increased
HDS conversion for Pd/MNZ-5 versus Pd/g-Al2O3 was ascribed
to the surface acidity; however, the Pd particle size changed
from 3.8 to 1.3 nm, which according to our current findings
could also affect the reaction rate. We used g-Al2O3 as a support
for the Pd“M”, Pd“L”, and Pd“XL” catalysts, and MgAl2O4 spinel for
the Pd“S” particles for improved sintering resistance.[18] The
spinel is known to possess similar acidity as Al2O3 with an iso-
electric point value of 8.4.[19] In the current work, no acid-cata-
lyzed isomers were observed, and the most dramatic changes
in the contributions to hydrogenation and sulfur extraction
were found for the three catalysts deposited on the same
Figure 5. Pd and S elemental mapping of the spent Pd“M” catalyst.
To evaluate the hypothesis of structure-sensitive sulfur poi-
soning, we performed hydrogenation of 3,3’-DMBP for the
Pd“S”, Pd“M”, and Pd“L” samples at the same conditions (3008C,
5 MPa) in the presence and absence of 300 ppmw sulfur in the
feed (as CS2 based on the sulfur amount corresponding to
40% 4,6-DMDBT conversion in HDS, i.e., 40% of 800 ppmw).
Conversions below 10% were ensured, which allowed us to
calculate turnover frequencies (TOF) under the assumption of
a differential reactor performance. As seen from Figure 6, in
the sulfur-free feed, the hydrogenation TOF increased with the
particle size increase, which is in line with our above discussion
on the larger particle requirement for the flat reactant adsorp-
tion. A support effect on sulfur poisoning of Pt catalysts was
&
ChemCatChem 2016, 8, 1 – 8
www.chemcatchem.org
4
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
sulfur extraction based on total Pd mass loading in the reactor
was found to be 8 nm. Above this size, in spite of the surface
activity increase, the size increase decreases the atom efficien-
cy of the active metal.
Experimental Section
Materials
Palladium(II) chloride solution (PdCl2, 5% w/v, Acros), sodium tetra-
chloropalladate(II) (Na2PdCl4, 98%, Sigma–Aldrich), poly(vinylpyrro-
lidone) (PVP) (MW=40000, Sigma–Aldrich), generation 4 poly(ami-
doamine)-hydroxyl dendrimer (G4OH, 10 wt% in methanol, Dendri-
tech Inc.), potassium bromide (KBr, >99%, Sigma–Aldrich), ethyl al-
cohol (denatured, Sigma–Aldrich), reagent alcohol (ethanol,
95 vol%, Fisher Scientific), ethylene glycol (EG, 99.8%, Sigma–Al-
drich), ascorbic acid (ꢁ99%, Sigma–Aldrich), sodium borohydride
(NaBH4, >99%, Sigma–Aldrich), aluminum oxide (g-Al2O3,
150 mesh, 58 ꢁ pore size, Sigma–Aldrich), aluminum isopropoxide
(Al(OCH(CH3)2)3, ꢁ98%, Sigma–Aldrich), magnesium nitrate hexa-
hydrate (Mg(NO3)2·6H2O, ꢁ98%, Sigma–Aldrich), and acetone
(99.7%, Fisher Scientific) were used as received. The gases of ultra-
high purity (i.e., 99.999%) were purchased from Praxair. For catalyt-
ic reactions, n-decane (99.4%) from Fisher Scientific, n-dodecane
(ꢁ99%), and 4,6-DMDBT (97%), 3,3’-dimethylbiphenyl (99%), and
carbon disulfide (99.5%) from Sigma–Aldrich were used as re-
ceived. Milli-Q water was used throughout the work.
Figure 6. Turnover frequency of DMBP hydrogenation over Pd“S”, Pd“M”, and
Pd“L” catalysts, with and without 300 ppm of sulfur in the feed (CS2 equiva-
lent corresponding to 40% 4,6-DMDBT conversion in the HDS experiments
with 800 ppm S).
suggested to be negligible, apart from providing different
metal nanoparticle sizes.[12]
However, comparing Figure 3b and Figure 6, the activity per
surface palladium during 4,6-DMDBT conversion changed more
dramatically on going from Pd“S” to Pd“M” catalysts compared
with the sulfur-free hydrogenation. On addition of sulfur (as
CS2) to the 3,3’-DMBP feed, all three catalysts exhibited two
orders of magnitude lower hydrogenation rates, but the small-
est Pd“S” nanoparticles were poisoned to the largest extent and
showed the lowest hydrogenation activity. The drop in DDS
and HYD activity was also reported for thiophene hydrogena-
tion on Pt and Ru for smaller particle sizes and was ascribed to
the stronger metal–sulfur bonds and lower active vacancy con-
centration on corner and edge sites that are prevalent on the
smaller clusters.[7,8] Our findings, thus, are in line with reported
mechanisms and active site requirements for metal-catalyzed
HDS and have established the most desirable Pd nanoparticle
(8 nm) for enhanced sulfur extraction from a refractory 4,6-
DMDBT molecule. Below this size, significant sulfur poisoning
inhibits both hydrogenation and direct desulfurization rates.
Catalyst preparation and characterization
Four samples of stabilized palladium nanoparticles with controlled
size were prepared in colloidal solutions following methods avail-
able in the literature. The smallest nanoparticles (Pd“S”) with <2 nm
mean diameter were prepared by using a dendrimer-templating
strategy, originally developed by Crooks and co-workers,[20] with
some modifications. G4OH solution (10 mL of 0.25 mm) was pre-
pared by diluting G4OH methanol solution (0.357 g of 10 wt%,
0.0025 mmol) in water. PdCl2 methanol solution (100 mL of 1 mm)
was prepared by adding PdCl2 solution (0.356 mL, 5% w/v,
0.1 mmol) in methanol in a 250 mL round-bottom flask. Then, the
G4OH solution was added into the reaction flask containing the
PdCl2 methanol solution (molar ratio of G4OH to Pd=1:40). After
1 h of complexation, NaBH4 solution (2 mL of 1m, 2 mmol; NaBH4
to Pd molar ratio=20:1; prepared and kept at 08C) was added to
the reaction mixture dropwise under vigorous stirring. The color of
the Pd2+–G4OH solution changed from pale yellow to dark brown
indicating the reduction of Pd ions to metallic Pd. The reaction
was continued for 1 h for complete reduction.
Conclusions
Among Pd clusters of average 4, 8, 13, and 87 nm particle size,
the largest particles exhibited the highest surface activity in
4,6-DMDBT conversion, which, however, was a result of the for-
mation of hydrogenated sulfurous intermediates with a low
contribution from sulfur extraction. The smallest particles pro-
vided an unprecedented (for Pd at 5 MPa reaction pressure)
direct desulfurization selectivity of 20% at 40% 4,6-DMDBT
conversion because of the reduced contribution of the hydro-
genation path, which requires a low edge/terrace atom ratio
on the nanoparticle surface. The smallest particles were also
found to be poisoned by adsorbed sulfur to a greater extent,
as per kinetic studies of sulfur-free and sulfur-inhibited hydro-
genation of 3,3’-dimethylbiphenyl. Among the studied cata-
lysts, the optimal Pd nanoparticle size that ensures maximum
Medium-size particles of ~3 nm (Pd“M”) were prepared by using Ter-
anishi and Miyake’s one-step alcohol reduction method in the pres-
ence of PVP,[21] with minor modifications.[22] A mixture containing
PdCl2 (0.2 mmol), ethanol/water solution (170 mL, 41 vol% etha-
nol), and PVP (0.4 mmol, PVP to metal molar ratio=20:1) was
stirred and heated at reflux for 3 h under air. Transparent dark-
brown colloidal solutions of monometallic Pd nanoparticles were
obtained without any precipitate.
Larger nanoparticles of ~7 nm diameter (Pd“L”) were prepared by
the ethylene glycol reduction method in the presence of PVP.[22,23]
PdCl2 (0.2 mmol) and PVP (4 mmol, PVP to Pd molar ratio=20:1)
were well dissolved in ethylene glycol (200 mL) at room tempera-
ture. Then, the PVP–Pd2+–EG solution was stirred and heated at
reflux under air for 3 h for complete reduction.
ChemCatChem 2016, 8, 1 – 8
www.chemcatchem.org
5
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
Pd nanocubes with an average edge length of 16 nm (Pd“XL”) were
prepared by using slow reduction by ascorbic acid in the presence
of PVP and KBr, following the synthetic procedure described by Xia
and co-workers[24] with some modifications. PVP (1 mmol), KBr
(5.2 mmol), and ascorbic acid (0.2 mmol) were dissolved in water
(20 mL) at 808C. Then, an aqueous solution (5 mL) containing
NaPdCl4 (0.2 mmol) was added to the reaction solution while stir-
ring. The molar ratios of PVP, KBr, and ascorbic acid to Pd were 5:1,
2:1, and 26:1, respectively. The color of the reaction solution
changed from yellow to dark brown, indicating the reduction of
Pd2+ to metallic Pd. The reaction was continued for 3 h at 808C for
complete reduction.
stream). The absence of external and internal mass transfer limita-
tions was verified experimentally for the most active Pd“M” catalyst.
The results can be found in the Supporting Information. Liquid
samples were analyzed offline by using a Varian 430 gas chromato-
gram (GC) equipped with a flame ionization detector (FID). The GC
capillary column was a CP-Sil 8 CB column, 50 m lengthꢂ0.25 mm
inner diameter. Initially, the oven temperature was stabilized at
508C for 1 min and then increased to 3008C with a rate of
108CminÀ1. The observed products were identified by GC-MS and
found to be in agreement with the reaction mechanism in
Scheme 1. The mass balance was closed within 94–99%. No acid-
catalyzed isomerization products[4e] were observed. The reported
selectivity is an integral selectivity (reported on molar basis; pro-
duced H2S not included) that was determined as the amount of
each HDS product formed divided by the total amount of HDS
products (except H2S). In the Results section, the selectivity to DDS
is the selectivity to 3,3’-dimethylbiphenyl (DMBP); the selectivity to
S-free products by HYD is the summation of the selectivities to
3,3’-dimethylcyclohexylbenzene (3,3’-DMCHB) and 3,3’-dimethylbi-
cyclohexyl (3,3’-DMBCH); the selectivity to S-containing intermedi-
ates by the HYD route is the summation of selectivities to 4,6-di-
methyltetrahydrodibenzothiophene (4,6-DMTHDBT), 4,6-dimethyl-
hexahydrodibenzothiophene (4,6-DMHHDBT), and 4,6-dimethylper-
hydrodibenzothiophene (4,6-DMPHDBT); the selectivity to the total
HYD route is (100%Àselectivity to DDS), as per Scheme 1.
The colloidal solutions were used to deposit the nanoparticles on
a solid support with a target loading of 0.3 wt%. The freshly pre-
pared Pd“M”, Pd“L”, and Pd“XL” particles were deposited on g-Al2O3,
which was pre-calcined in air at 5008C for 2 h. Pd“M” particles were
deposited on g-Al2O3 by incipient impregnation. Pd“L” and Pd“XL”
particles were precipitated with acetone and deposited on g-Al2O3
by wet impregnation. Pd“S” particles were precipitated with acetone
and deposited on MgAl2O4 spinel support by wet impregnation.
The support was prepared by a known procedure:[18] in a 1000 mL
Pyrex bottle (autoclavable), Al(OCH(CH3)2)3 (0.1 mol) and
Mg(NO3)2·6H2O (0.05 mol) were dissolved in denatured ethanol
(300 mL) under stirring until complete dissolution (at least 1 h).
The reaction vessel was sealed, and the reaction mixture was
heated to 1508C (hot plate digital display) for 12 h. The solvent,
denatured ethanol, was evaporated at 1008C under stirring, and
the resulting gel was then dried in an oven at 908C overnight. To
obtain the MgAl2O4 spinel material, the dry powder was calcined in
a furnace under static air at 8008C for 12 h. The deposited Pd cata-
lysts were calcined in air at 3508C for 2 h, followed by reduction in
a hydrogen flow at 3008C for 1 h for the stabilizing polymer re-
moval. The final Pd loadings were determined by neutron activa-
tion analysis (Becquerel Laboratories, Maxxam Company, Canada)
and found to be 0.200 wt%, 0.240 wt%, 0.240 wt%, and
0.210 wt% for the Pd“S”, Pd“M”, Pd“L”, and Pd“XL” catalysts, respective-
ly. Transmission electron microscopy (TEM) and CO chemisorption
of the nanoparticles and/or supported catalysts were performed as
described previously.[22] For the TEM analysis, 100–200 nanoparti-
cles per sample were counted from TEM images by using ImageJ
software. The reported sizes are based on the linear distribution.
Prior to CO chemisorption, the catalysts were calcined at 3508C for
2 h in air, which was followed by in situ reduction in 10% H2/Ar at
3008C for 1 h.
Hydrogenation of 3,3’-DMBP
DMBP hydrogenation to DMBCH and DMCHB was studied by using
the same HDS setup at 3008C and 50 bar with a hydrogen flow
rate of 100 mLminÀ1, which simulated the experimental conditions
of HDS. The DMBP was fed into the catalytic system by pumping
a solution containing 1 wt% DMBP and 3.5 wt% dodecane (inter-
nal standard) balanced in decane; the corresponding DMBP flow
rate was 1.96ꢂ10À6 molminÀ1. Pd catalyst (0.005 g) was pre-mixed
with g-Al2O3 (0.2 g) to avoid bypassing and then packed into the
reactor according to the procedure for the HDS reactions. The cat-
alytic performances in DMBP hydrogenation were analyzed and re-
ported at 18 h time-on-stream. DMBP hydrogenation was also per-
formed with the presence of 300 ppmw sulfur in the feed solution
(in the form of CS2) to simulate the amount of H2S produced at
about 40% DMDBT conversion during HDS reactions. Owing to the
strong S-poisoning effect, DMBP hydrogenation with sulfur was
tested at a lower space velocity; 0.2, 0.05, and 0.1 g of Pd“S”, Pd“M”
,
and Pd“L” catalysts were packed in the reactor, respectively.
Hydrodesulfurization of 4,6-DMDBT
Acknowledgements
Hydrodesulfurization of 4,6-DMDBT was performed at 3008C and
50 bar with
a
hydrogen flow rate of 100 mLminÀ1 (STP) in
Financial support from the Institute for Oil Sands Innovation at
the University of Alberta is gratefully acknowledged (grant 2011-
01). Dr. L. Wu, L. Dean, W. Boddez, and Dr. H. Ziaei-Azad contrib-
uted to the HDS setup development.
a packed bed reactor (22“ long stainless steel tube with an inner
diameter of 1/2”), according to the pioneering works published by
Prins’s group[2,4] with some modifications by our group.[25] The cal-
cined catalysts (0.09–0.54 g, diluted in 150 mesh SiC, total of 4 g)
were packed in the reactor and then reduced in situ at 3008C and
50 bar for 1 h under a hydrogen flow (100 mLminÀ1). The catalyst
amounts were different to ensure similar conversions for fair selec-
tivity comparison. The 4,6-DMDBT was fed into the catalytic system
by pumping a solution of 0.5 wt% 4,6-DMDBT (750 ppmw of
sulfur) and 3.5 wt% dodecane (internal standard) balanced in
decane (solvent) with a flow rate of 87ꢂ10À6 molminÀ1. The cata-
lytic performances were assessed after 16 h stabilization. Prior to
sampling, the condenser was emptied after the 16-hour stabiliza-
tion; liquid samples were then collected after 2 h (at 18 h time-on-
Keywords: desulfurization · nanoparticles · size control ·
structure sensitivity · supported catalysts
[1] a) R. Prins in Handbook of Heterogeneous Catalysis, Vol. 6 (Eds.: G. Ertl, H.
Knçzinger, F. Schꢃth, J. Weitkamp), Wiley-VCH, Weinheim, 2008,
pp. 2695–2718; b) C. Song, X. Ma, Appl. Catal. B 2003, 41, 207–238;
c) E. Ito, J. A. Rob van Veen, Catal. Today 2006, 116, 446–460; d) X. Ma,
K. Sakanishi, I. Mochida, Ind. Eng. Chem. Res. 1996, 35, 2487–2494; e) X.
Ma, K. Sakanishi, T. Isoda, I. Mochida, Energy Fuels 1995, 9, 33–37.
&
ChemCatChem 2016, 8, 1 – 8
www.chemcatchem.org
6
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
[2] A. Niquille-Rçthlisberger, R. Prins, J. Catal. 2006, 242, 207–216.
[3] a) A. Stanislaus, A. Marafi, M. S. Rana, Catal. Today 2010, 153, 1–68;
b) D. D. Whitehurst, H. Farag, T. Nagamatsu, K. Sakanishi, I. Mochida,
Catal. Today 1998, 45, 299–305.
[4] a) A. Rçthlisberger, R. Prins, J. Catal. 2005, 235, 229–240; b) H. Guo, Y.
Sun, R. Prins, Catal. Today 2008, 130, 249–253; c) E. J. M. Hensen,
H. J. A. Brans, G. M. H. J. Lardinois, V. H. J. de Beer, J. A. R. van Veen, R. A.
van Santen, J. Catal. 2000, 192, 98–107; d) E. W. Qian, K. Otani, L. Li, A.
Ishihara, T. Kabe, J. Catal. 2004, 221, 294–301; e) Y. Sun, R. Prins, Angew.
Chem. Int. Ed. 2008, 47, 8478–8481; Angew. Chem. 2008, 120, 8606–
8609.
[5] A. Ishihara, F. Dumeignil, J. Lee, K. Mitsuhashi, E. W. Qian, T. Kabe, Appl.
Catal. A 2005, 289, 163–173.
[6] S. T. Oyama, H. Zhao, H. J. Freund, K. Asakura, R. Wlodarczyk, M. Sierka,
J. Catal. 2012, 285, 1–5.
[7] H. Wang, E. Iglesia, ChemCatChem 2011, 3, 1166–1175.
[8] H. Wang, E. Iglesia, J. Catal. 2010, 273, 245–256.
[9] T. Matsui, M. Harada, Y. Ichihashi, K. K. Bando, N. Matsubayashi, M. Toba,
Y. Yoshimura, Appl. Catal. A 2005, 286, 249–257.
[14] a) R. Ma, N. Semagina, J. Phys. Chem. C 2010, 114, 15417–15423; b) R.
van Hardeveld, F. Hartog, Surf. Sci. 1969, 15, 189–230.
[15] I. Yuranov, L. Kiwi-Minsker, A. Renken, Appl. Catal. B 2003, 43, 217–227.
[16] V. V. Pushkarev, K. An, S. Alayoglu, S. K. Beaumont, G. A. Somorjai, J.
Catal. 2012, 292, 64–72.
[17] N. Semagina, A. Renken, L. Kiwi-Minsker, J. Phys. Chem. C 2007, 111,
13933–13937.
[18] W. Z. Li, L. Kovarik, D. Mei, J. Liu, Y. Wang, C. H. F. Peden, Nat. Commun.
2013, 4, 2481.
[19] a) J. Guo, H. Zhao, D. Chai, X. Zheng, Appl. Catal. A 2004, 273, 75–82;
b) S. A. Bocanegra, S. R. de Miguel, A. A. Castro, O. A. Scelza, Catal. Lett.
2004, 96, 129–140.
[20] R. W. J. Scott, O. M. Wilson, R. M. Crooks, J. Phys. Chem. B 2005, 109,
692–704.
[21] T. Teranishi, M. Miyake, Chem. Mater. 1998, 10, 594–600.
[22] J. Shen, N. Semagina, ACS Catal. 2014, 4, 268–279.
[23] J. Shen, X. Yin, D. Karpuzov, N. Semagina, Catal. Sci. Technol. 2013, 3,
208–221.
[24] M. Jin, H. Liu, H. Zhang, Z. Xie, J. Liu, Y. Xia, Nano Res. 2011, 4, 83–91.
[25] H. Ziaei-azad, Ph.D. Thesis, University of Alberta, 2014.
[10] S. D. Jackson, P. Leeming, J. Grenfell, J. Catal. 1994, 150, 170–176.
[11] S. D. Jackson, B. J. Brandreth, D. Winstanley, J. Catal. 1991, 129, 540–
543.
[12] S. D. Jackson, P. Leeming, G. Webb, J. Catal. 1996, 160, 235–243.
[13] L. M. Baldyga, S. O. Blavo, C.-H. Kuo, C.-K. Tsung, J. N. Kuhn, ACS Catal.
2012, 2, 2626–2629.
Received: April 28, 2016
Published online on && &&, 0000
ChemCatChem 2016, 8, 1 – 8
www.chemcatchem.org
7
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

FULL PAPERS
J. Shen, N. Semagina*
&& – &&
Palladium Nanoparticle Size Effect in
Hydrodesulfurization of 4,6-
Dimethyldibenzothiophene (4,6-
DMDBT)
Optimal size: 8 nm Pd nanoparticles
provide the deepest hydrodesulfuriza-
tion as a result of balanced hydrogena-
tion, sulfur extraction, and sulfur poison-
ing.
&
ChemCatChem 2016, 8, 1 – 8
www.chemcatchem.org
8
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!