Catalytic deoxygenation of fatty acids: Elucidation of the inhibition process

Article abstract of DOI:10.1002/cctc.201402290

Catalytic deoxygenation of unsaturated fatty acids in the absence of H₂ is known to suffer from significant catalyst inhibition. Thus far, no conclusive results have been reported on the cause of deactivation. Here we show that C-C double bonds present in the feed or the products dramatically reduce the deoxygenation activity of supported palladium catalysts. In the case of stearic acid deoxygenation the addition of 0.1 equivalents of a mono-unsaturated fatty acid or olefin already reduces the catalytic deoxygenation activity by 60%. This effect becomes more pronounced with an increasing number of double bonds. The inhibition is shown to be reversible in H₂ atmosphere, indicating no significant contribution from irreversibly deposited hard coke. Furthermore, the type of support material has no apparent effect on catalyst inhibition. Hence we propose that initial catalyst inhibition proceeds through reversible adsorption of C-C double bonds on the palladium active sites.

Full text of DOI:10.1002/cctc.201402290

CHEMCATCHEM
FULL PAPERS
DOI: 10.1002/cctc.201402290
Catalytic Deoxygenation of Fatty Acids: Elucidation of the
Inhibition Process
[a]
[b]
[a]
Stefan A. W. Hollak, Krijn P. de Jong, and Daan S. van Es*
Catalytic deoxygenation of unsaturated fatty acids in the ab-
becomes more pronounced with an increasing number of
double bonds. The inhibition is shown to be reversible in H2
atmosphere, indicating no significant contribution from irrever-
sibly deposited hard coke. Furthermore, the type of support
material has no apparent effect on catalyst inhibition. Hence
we propose that initial catalyst inhibition proceeds through re-
versible adsorption of C=C double bonds on the palladium
active sites.
sence of H is known to suffer from significant catalyst inhibi-
2
tion. Thus far, no conclusive results have been reported on the
cause of deactivation. Here we show that C=C double bonds
present in the feed or the products dramatically reduce the de-
oxygenation activity of supported palladium catalysts. In the
case of stearic acid deoxygenation the addition of 0.1 equiva-
lents of a mono-unsaturated fatty acid or olefin already re-
duces the catalytic deoxygenation activity by 60%. This effect
Introduction
Fossil resources such as coal, oil, and gas have contributed
greatly to our modern lifestyle and will continue to do so in
the coming decades. To ensure a lasting supply of energy for
the longer term, we will increasingly depend on renewable re-
sources such as biomass. Science and industry are working to-
gether to turn biomass into, for instance, biofuels that are
compatible with the existing distribution infrastructure and
engine technologies.
tions were reported with the use of more practical feed-
[2]
stocks. Catalyst deactivation appeared more severe during
H -free deoxygenation reactions and was regularly related to
2
the presence of unsaturated compounds. Decreasing H pres-
2
sure during the reaction resulted in an increasing amount of
unsaturated compounds and more severe catalyst deactiva-
[
1a,b,3]
tion.
faster deactivation if higher selectivity towards unsaturated
Snꢀre et al. screened several catalysts and reported
[4]
Vegetable oil-based feedstocks have gained significant inter-
est as a sustainable resource for the production of fuels and
chemicals, mainly owing to their high energy density and
structural similarity to petroleum-based fuels. As such, the im-
plementation of these compounds would require minimal
changes in the present infrastructure and the internal combus-
tion engine, which is highly beneficial for the introduction of
new technologies. Especially the use of non-edible or waste
fats and oils are of interest as this will not interfere with the
food production.
products was obtained (e.g., if using Ru/C and Rh/C).
Thus far, no conclusive results have been reported on this
process of catalyst inhibition and/or deactivation during H2-
free deoxygenation. Several explanations for these phenom-
ena, however, have been proposed in the literature. The forma-
tion of unsaturated products often also gives rise to CO in the
[
1e,4]
gas stream as a direct decarbonylation product,
which
[5]
could deactivate the metal phase through poisoning. Howev-
er, CO poisoning could not explain observations of decreased
[6]
activity after re-reduction of the spent catalysts.
The use of (nonrenewable) H2 is to be avoided for the
design of a sustainable process, because this has impact on
The formation of irreversibly deposited coke (referred to as
hard coke, for example, graphitic carbon or polycyclic aromatic
[7]
the environmental footprint. H -free deoxygenation has been
compounds ) from unsaturated compounds is also mentioned
in the literature as a reason for catalyst deactivation. Lestari
et al. reported the presence of trace amounts of aromatic com-
pounds during the deoxygenation of stearic acid over Pd/C,
which was related to coke formation and a decreasing surface
2
investigated frequently, and although demonstrating much po-
[
1]
tential in the deoxygenation of saturated feeds, complica-
[a] S. A. W. Hollak, Dr. D. S. van Es
Food and Biobased Research
[1c]
area of the spent catalyst. The authors suggested that unsa-
Wageningen University and Research Centre
P.O. Box 17, Wageningen, AA, 6700 (The Netherlands)
Fax: (+31)317475347
turated products were the origin of aromatic formation and
subsequent coke formation. The formation of alkenylbenzenes
as side products (e.g., undecylbenzene) was indeed confirmed
E-mail: daan.vanes@wur.nl
[1d,8]
[b] Prof. Dr. Ir. K. P. de Jong
during the deoxygenation of fatty acids.
Simakova et al. re-
Inorganic Chemistry and Catalysis
Utrecht University
Sorbonnelaan 16, Utrecht, CA, 3584 (The Netherlands)
ported decreasing access to palladium metal clusters after cat-
alytic deoxygenation of several fatty acids and related this to
[9]
coke formation as well.
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201402290.
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 2648 – 2655 2648

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Reduced accessibility of active metal clusters could
also be explained by the adsorption of unsaturated
reactants or (side) products on the catalyst by the
C=C double bond in the alkyl chain, as was speculat-
Table 1. Catalyst properties.
[
a]
[c]
[d]
Pd loading BETsurface
Pore volume Dispersion
Average particle
size [nm]
[
b]
2
À1
3
À1
[e]
[
wt%]
area [m g
]
[cm g
]
[%]
[
1e]
ed by Immer et al. Unlike in the case of hard-coke
formation, the inhibition by adsorption of unsaturat-
ed reactants could be a reversible process. No further
investigation was, however, performed to test and
prove this hypothesis.
Pd/SiO2 5.07
218
106
n.d.
0.012
0.003
n.d.
6.94
9.64
6.88
16.1
11.6
16.3
Pd/Al
Pd/C
2
O
3
4.98
4.67
[
f]
[a] Elemental analysis; [b] N
volume; [d] H chemisorption; [e] average particle size calculated from the dispersion,
as described by Scholten et al.; [f] not determined
2 2
physisorption; [c] N physisorption, t-plot micropore
2
[
10]
Ping et al. claimed that adsorption of the feed
(
stearic acid), solvent (dodecane), and product (hep-
[
6]
tadecane) diminished the deoxygenation activity.
These authors do not relate the inhibition to the presence of
unsaturations but to the presence of saturated compounds in-
stead. These conclusions are puzzling because the reduced de-
oxygenation activity is only reported for recycle runs, but the
catalyst is also in contact with saturated compounds during
the initial run. To the best of our knowledge, no other reports
confirm the inhibition in deoxygenation activity by the strong
adsorption of saturated compounds.
Catalyst properties are summarized in Table 1. Catalytic de-
oxygenation reactions were performed for 20 h at 2508C
under 7 bar N pressure (4 bar at room temperature) unless
2
noted otherwise. Pd/SiO was selected as the main catalyst for
2
this inhibition study because SiO is considered to be chemical-
2
ly inert towards fatty acids under the reaction conditions, and
also allows for spectroscopic analysis.
A stepwise investigation is essential to study the effect of
potential inhibition agents and to elucidate the process of cat-
alyst inhibition and/or deactivation. The influence of the fatty
acid feed on the deoxygenation activity and product selectivity
was initially investigated to confirm the relationship between
unsaturations and deoxygenation activity, and to investigate
the different reactions that occur.
In summary, several explanations are proposed to explain
the catalyst deactivation during H -free deoxygenation: CO
2
poisoning, hard-coke formation, and adsorption of saturated
and of unsaturated reactants/products on the catalyst surface
[
7]
(
i.e., soft coke ). Possibly several of these processes are occur-
ring simultaneously. No conclusive results have, however, been
reported thus far that thoroughly elucidate the inhibition pro-
cess and investigate the effect of feed, products, unsaturations,
and CO formation on the inhibition process.
Catalytic deoxygenation of stearic acid (SA, 97% pure) over
Pd/SiO results in deoxygenation to heptadecane (C17sat.) and
2
heptadecenes (C17unsat.) by decarboxylation and decarbonyla-
tion, respectively (Scheme 1). In contrast, the catalytic deoxy-
In this work we report on the inhibition process during the
deoxygenation of fatty acids over several Pd-based catalysts.
We show that inhibition is caused by strong but reversible ad-
sorption of unsaturated molecules by the C=C double bond on
the palladium active sites. Inhibition occurs in presence of un-
saturated feeds or products as well as with the presence of
CO. The inhibition agents are shown to have dramatic effects
on the deoxygenation activity even if present in very small
quantities. However, the inhibition is demonstrated to be re-
Scheme 1. Deoxygenation pathways of fatty acids (SA shown here) in the
absence of H yielding a) a paraffinic product after decarboxylation and
2
b) an olefinic product in the case of decarbonylation.
versible and the presence of H is essential to effectively regen-
2
erate the deoxygenation activity.
genation activity drastically decreases if oleic acid (OA, 90%) is
used as a feedstock (Figure 1). Note that the oleic acid feed
contains stearic acid (ꢀ8%) and some linoleic acid (ꢀ2%) as
impurities. No deoxygenation products are observed if using li-
noleic acid (LL, 99% pure) or linolenic acid (LLN, 71% pure,
ꢀ20% LL, ꢀ9% OA) as a feed. If linoleic acid is used as feed,
the reaction composition also contains 13 mol% oleic acid iso-
mers and 14 mol% of other oxygenates (including mainly lino-
lenic isomers). These are most probably formed by transfer hy-
drogenation of the feed.
Results and Discussion
For our study into catalyst inhibition, supported palladium cat-
alysts are chosen, as palladium is reported as the most active
metal in H -free fatty acid deoxygenation and has been investi-
2
[
1a,2a,b,3a,4,11]
gated most extensively.
Furthermore, we used com-
mercially available catalysts for this study to have a constant
quality of catalyst over a broad range of experiments.
Pd/SiO and Pd/g-Al O were selected because of the differ-
2
2
3
ent deoxygenation pathways reported for these catalysts (de-
carboxylation for Pd/Al O and combined decarbonylation and
Double-bond isomerization under the applied reaction con-
ditions poses serious challenges to analyze the products quan-
titatively by gas chromatography. As an example, peak overlap
occurs for linolenic acid isomers with several oxygenates like
alcohols, aldehydes, and cyclization products (clustered as
“other oxygenates”). The accuracy of the calculated mass bal-
2
3
[
12]
decarboxylation for Pd/SiO2 ), in combination with the differ-
ent support properties (e.g., the difference in isoelectric point
of 1.7–3.5 for SiO and 7–8 for g-Al O ). Furthermore, Pd/C was
2
2
3
also used because it is the most active reported catalyst for
[
4]
the H -free deoxygenation of stearic acid.
2
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 2648 – 2655 2649

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
6
% of the original TON with the addition of 0.4 equivalents
oleic acid.
If inhibition is caused solely by the presence of the C=C
double bond, comparable inhibition is expected if an olefin is
used as inhibition agent. As shown in Figure 2b, pentadecene
(
C15:1) indeed gives a comparable effect on the TON during
the catalytic deoxygenation of stearic acid. These results show
that the presence of unsaturated feed and products have a sim-
ilar inhibiting effect on the deoxygenation activity. This inhibi-
tory effect by unsaturated compounds suggests inhibition initi-
ated by adsorption of the C=C double bond on the catalyst
surface and possibly subsequent side reactions or coke forma-
tion. In the case of inhibition by competitive adsorption be-
tween the carboxyl group and the C=C double bond, the ad-
sorption of the C=C double bond must be irreversible or much
stronger than adsorption of the carboxyl group to explain the
substantial inhibition at relatively low concentrations.
Figure 1. Reaction mixture composition after catalytic reaction of different
feeds over Pd/SiO .
2
Except for product inhibition in the presence of olefins, CO
poisoning could also occur through adsorption on the palladi-
um metal surface, because Pd/SiO2 was shown to be also
ance for that reason decreases with increasing amounts of iso-
mers and other oxygenates.
[1e,4–5]
active in fatty acid decarbonylation.
To investigate the in-
Small amounts of heavies (e.g., stearyl stearate, stearone,
and other dimers) are also present in the reaction mixture after
reactions with unsaturated feeds, as analyzed by gas–liquid
chromatography. The amount of heavies is increasing with the
amount of unsaturations in the feed, as shown in Figure 1. Hy-
drocarbon cracking products (C13–C16 alkanes) were not
found in the product mixture.
hibition by CO formation we used hexadecanal as an inhibition
agent, which is selectively decarbonylated to form CO and
pentadecane (Scheme 2). Comparable mechanisms are report-
ed if using octadecanal or ethanal as the feed with heptade-
cane and methane as decarbonylation products, respective-
[2d,13]
ly.
As shown in Figure 2c, a decrease in TON is observed if
These results confirm that the presence of C=C double
bonds in the feedstock has severe detrimental effects on the
deoxygenation activity. A more systematic approach is essen-
tial to elucidate the inhibition process, because several poten-
tial inhibition products are pres-
hexadecanal was added to the reaction mixture. As GC analysis
of the reaction mixture reveals full conversion to pentadecane
within 2 h, and as pentadecane itself does not inhibit the reac-
tion (Figure 2d), CO poisoning is indeed inhibiting the deoxy-
ent in the reaction mixture, such
as unsaturated feed and prod-
ucts, heavies and possibly other
side products.
The impact of potential inhibi-
tion agents was investigated by
monitoring the effect of small
amounts (0.05–0.6 equiv. relative
to stearic acid) of inhibition
agent on the turnover number
(TON) during the catalytic deoxy-
genation of stearic acid. In
Figure 2, the TONs are shown as
a function of the molar equiva-
lents of inhibition agent. Note
that TON is defined as conver-
sion to C17 hydrocarbons
[
[
mmol] per surface palladium
mmol]. As shown in Figure 2a,
the addition of only 0.05 equiva-
lents of oleic acid to the reaction
mixture is enough to significant-
ly decrease the TON (>30%),
Figure 2. The effect of a) OA, b) pentadecene, c) hexadecanal, d) pentadecane, e) LL, and f) 1,7-octadiene on the
TON during SA deoxygenation over Pd/SiO
2
. Error bars are shown to illustrate the deviation when multiple experi-
which further deceases to only ments are performed.
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 2648 – 2655 2650

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
catalyst. Theoretically, an amount of linoleic acid of only
À1
2
mggcat. would already be enough for a monolayer of linole-
[16]
ic acid on the palladium metal surface.
Scheme 2. Decarbonylation of hexadecanal to CO and pentadecane.
As the carboxylic acid group of the fatty acid was shown to
have no significant effect on the inhibition process (Figure 2a,
b), comparable inhibition is expected for linoleic acid and a di-
[17]
genation activity under inert atmosphere. It should be noted
olefin such as 1,7-octadiene. In Figure 2 f, it is shown that
the addition of 0.05 equivalents of 1,7-octadiene as inhibition
agent almost entirely suppresses the deoxygenation activity.
The overlay of the inhibition curves of the fatty acids and
olefins clearly demonstrates the correlation between the
amount of unsaturations and the impact on the deoxygenation
activity (Figure 3). However, these results do not indicate
whether the inhibition is initiated by reversible adsorption on
the catalyst surface (i.e., soft coke) or by irreversible adsorption
and conversion of the adsorbed species to, for instance, poly-
cyclic aromatic compounds or graphitic carbon (i.e., hard
coke).
that H O is also formed during decarbonylation of fatty acids
2
(Scheme 1), which could decrease CO poisoning, because
[4,14]
water-gas-shift reactions would yield H and CO instead.
2
2
Linoleic acid is used as an inhibition agent to study the
effect of multiple unsaturations on the TON. As shown in Fig-
ure 2e, the inhibition of linoleic acid on the deoxygenation ac-
tivity is much stronger than that observed with the mono-un-
saturated compounds: the addition of only 0.05 equivalents of
linoleic acid already decreases the TON with 95%. The de-
crease in activity is an order of magnitude higher than with
oleic acid or pentadecene. As it is known that linoleic acid ad-
sorbs more strongly on metal surfaces because of the multiple
unsaturation (i.e., chelating effect), it is possible that it monop-
To investigate whether the inhibition is reversible, catalytic
deoxygenation reactions of stearic acid with inhibition agent
[15]
olizes the catalyst surface already at lower concentration.
were performed in N atmosphere (7 bar), in which the reac-
2
The addition of only 0.05 equivalents of linoleic acid would
indeed be enough to cover the total palladium surface of the
tion atmosphere was switched from N to H (7 bar) after 20 h
2
2
at 2508C. The reactions were subsequently continued for an-
other 20 h under H atmosphere. H is known to adsorb on the
2
2
palladium surface and Pd–H is considered as very efficient in
[18]
the hydrogenation of unsaturated species. In the case of re-
versible inhibition, the deoxygenation activity is expected to
restore after switching to H atmosphere as a result of hydro-
2
genation of the unsaturated inhibition agents. The deoxygena-
tion reactions were performed with stearic acid and 0.4 equiva-
lents of inhibition agent present, because this amount of inhib-
ition agent was shown to considerably inhibit the deoxygena-
tion (Figure 2).
In Figure 4 the results for the described N /H sequence re-
2
2
actions are compared with those for the catalytic reactions
performed under either N or H atmosphere. No deoxygena-
2
2
tion is observed under N2 atmosphere if 0.4 equivalents of
pentadecene was present (Figure 4a). Under H2 atmosphere
however, the pentadecene was fully hydrogenated to pentade-
cane (not shown in the figure) and 38 mol% of the stearic acid
is converted to heptadecane. The N /H sequence reaction re-
2
2
Figure 3. The effect of unsaturations on the TON during SA deoxygenation
over Pd/SiO . u=Number of unsaturations in the inhibition agent.
2
sulted in reaction composition comparable to that for the cata-
lytic reaction of 20 h under H atmosphere (Figure 4a). These
2
Figure 4. Reaction composition after SA deoxygenation over Pd/SiO
as the inhibition agent. Catalytic experiments were performed under N
/H sequence, 2ꢂ20 h).
2
in the presence of the amount of 1.3 mmol of a) pentadecene, b) OA, c) LL, and d) LLN
2
2 2 2
atmosphere (20 h), H atmosphere (20 h), and switching from N to H atmosphere
(N
2
2
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 2648 – 2655 2651

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
results demonstrate that the inhibition is reversible in H at-
2
mosphere and that deoxygenation activity is restored by hy-
drogenation of the pentadecene.
If oleic acid is used as the inhibition agent (Figure 4b), full
hydrogenation of the inhibition agent resulted in restored de-
oxygenation activity and comparable reaction compositions
relative to pentadecene. As the deoxygenation activity and
product selectivity are highly comparable to reactions per-
formed under H atmosphere, it can be concluded that the in-
2
hibition by 0.4 equivalents of mono-unsaturated compounds is
fully reversible with the introduction of 7 bar H at 2508C. De-
2
oxygenation activity was also restored if linoleic acid or linolen-
ic acid were used as inhibition agent (Figure 4c, d). However,
the heptadecane yields were slightly lower for the N /H se-
2
2
quence reactions than in the H2 reactions. This can be ex-
plained by the fact that two or three times as many double
bonds, respectively, are present in the inhibition agents.
The reversibility of the deoxygenation activity indicates that
initial inhibition is caused by strong but reversible adsorption
of the unsaturated compounds. H was shown to “clean” the
2
catalyst surface, most likely by adsorption on the palladium
surface, which subsequently hydrogenates the adsorbed unsa-
turated species followed by desorption. These results show
that the formation of irreversibly deposited hard coke does
not significantly contribute to the initial inhibition process.
Note that the mass balances are not complete if either lino-
leic or linolenic acid are added to the reaction mixture (Fig-
ure 4c, d). As the mass balances appear independent of the re-
action atmosphere (N or H ) and reaction time (20 or 40 h),
2
2
the loss in mass is presumably caused by adsorption on the
catalyst surface. A monolayer of fatty acids on the SiO surface
2
[
16]
would already contain up to 12 mol% feed. Chemisorption
and physisorption on the catalyst surface could thus possibly
account for the observed apparent mass loss.
2
Figure 5. Details of the IR spectra of the fresh Pd/SiO catalyst (“before reac-
tion”) and the spent catalysts using SA, OA, LL, and LLN as the feed, respec-
tively.
Although Soxhlet extraction of the spent Pd/SiO catalyst re-
2
moved part of the adsorbed species, chemisorbed organic de-
posits remained present on the catalyst surface, shown by
thermogravimetric analysis (TGA). The amount of organic dep-
osition increased if more unsaturations were present in the
sorbed H O on the catalyst surface (Figure 5). After reaction
2
À1
feed; from 27 mgg
in the case of stearic acid up to
with pentadecene or stearic acid as the feed, also CÀH stretch
cat.
À1
À1
64 mggcat. if using linolenic acid as feed (see the Supporting
vibrations at 3005 cm (from C=CÀH) and vibrations at ap-
À1
Information, Table S1). Comparable amounts of organic deposi-
tions were present after reaction with oleic acid or pentade-
cene as the feed, confirming the correlation between organic
deposition and the presence of C=C double bonds. The organ-
ic deposition is clearly not solely present on the palladium
active sites because the maximum amount of chemisorbed
proximately 2924 and 2854 cm (CH asymmetric and CH
2
2
symmetric stretch, respectively) are visible. The adsorbed unsa-
turated species could be product related if stearic acid was
used as the feed because olefins are produced during the reac-
tion (Figure 1) and are shown to inhibit deoxygenation activity
by adsorption on the catalyst surface (Figure 2b). The presence
À1
feed on the palladium metal surface would be between 0.2
of a small vibration at approximately 1710 cm (C=O stretch
À1
and 2 mgg
, depending on the coordination to the palladi-
vibrations) could also indicate feed-related species on the cata-
cat.
[
16]
[19]
À1
um surface. However, as mentioned earlier, a monolayer of
lyst surface.
The C=O stretch vibrations at 1710 cm ap-
À1
fatty acid on the SiO surface contains up to 240 mg feedg
,
peared more clearly in the spectra after the reaction with unsa-
turated fatty acids, confirming the presence of feed-related ad-
sorbates. Combined with the absence of symmetric and asym-
2
cat.
which could be the case for these spent catalysts (Table S1).
IR spectroscopy was performed to investigate the nature of
the chemisorbed organic deposits on the spent catalysts. The
À
À1
metric COO bands (1410 and 1556 cm ), this could indicate
that fatty acids are not chemisorbed as a carboxylate onto the
catalyst surface but instead mainly interact with the surface by
fresh catalyst (before reaction) exhibits H O bending vibrations
2
À1
at 1638 cm (OÀH scissors) and a broad OÀH stretch band at
À1
[19,20]
approximately 3300 cm , which can both be assigned to ad-
means of the carboxylic acid group.
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 2648 – 2655 2652

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
The IR data confirm the adsorption of olefinic adsorbates
after the reaction with stearic acid. Feed-related unsaturated
adsorbates appear to be present after reaction with unsaturat-
ed feeds. It should, however, be noted that these spectroscop-
ic data indicate the presence of chemisorbed molecules on the
support surface and not on the palladium active sites, as was
also concluded from TGA.
In summary, we reported the significant influence of small
amounts of unsaturated species on the deoxygenation activity
in the absence of added H . Although in the literature the for-
2
mation of hard coke was suggested by several authors as
a reason for the catalyst deactivation, we showed that the re-
versible adsorption of unsaturated species (soft coke) on palla-
dium is the initial and main reason for the inhibition of deoxy-
genation activity.
Although the palladium surface is active in the deoxygena-
tion of fatty acids, the support material is known to greatly
In comparison of our results with the work presented by
Ping et al., possibly not the abundantly present saturated com-
pounds caused the catalyst deactivation, but instead small
amounts of unsaturated compounds that could be formed
[
4,12]
affect the deoxygenation activity.
The influence of the sup-
port material on the deoxygenation activity was indeed con-
firmed by measuring the deoxygenation activities for several
supported palladium catalysts, shown in the Supporting Infor-
mation (Table S2). For that reason, it is expected that the inhib-
ition process is also affected by the type of support material.
To test this hypothesis, catalytic reactions were performed
over Pd/SiO , Pd/Al O , and Pd/C by using pentadecene and
[6]
during the reaction. Ping et al. reported the removal of the
organic deposits with more severe washing steps and showed
subsequent restored catalyst activity, which confirms our re-
sults in the present work demonstrating the reversibility of the
inhibition.
2
2
3
oleic acid as inhibition agents. The TON without inhibition
agent was set at unity to correct for the differences in deoxy-
genation activity.
Conclusions
From Figure 6 it is seen that comparable inhibition occurs
for the different catalysts if plotting the relative TON as a func-
The inhibition during H -free deoxygenation of fatty acids over
2
supported palladium catalysts is highly dependent on the pres-
ence of unsaturated compounds. Significant feed and product
inhibition occurs if unsaturated compounds are present, even
in low concentrations. The reversibility of the inhibition under
H atmosphere indicates that hard-coke formation does not
2
significantly contribute to the initial deactivation process. In-
stead, adsorption on the support by the C=C double bond was
inferred. The nature of the support material does not contrib-
ute to the inhibition process. This could be explained by the
adsorption of the unsaturations on the palladium surface as
main inhibition process.
Overall, we demonstrated that the reversible adsorption of
unsaturated molecules with the C=C double bond on the pal-
ladium active sites significantly contributes to catalyst inhibi-
tion. The presence of H is therefore essential to retain catalytic
2
deoxygenation activity.
Experimental Section
Chemicals
Figure 6. Relative TONs as a function of the amount of inhibition agent if
(^/~), Pd/Al (^/~), and Pd/C (^/~) as catalyst. The TON
using Pd/SiO O
2 2 3
without inhibition agent was set at unity. Pentadecene (C15:1, ^) and
OA (~) were used as inhibition agents.
The following catalysts were used for this investigation: Palladium
on silica powder (5 wt% Pd/SiO ) reduced dry (Strem Chemicals
2
Inc., Escat 1351); Palladium on alumina powder (5 wt% Pd/Al O )
2
3
reduced (Strem Chemicals Inc., Escat1241); Palladium on activated
carbon (5 wt% Pd/C) reduced dry powder (Strem Chemicals Inc.).
The following chemicals were used without further purification
treatments: Methanol, HPLC gradient 99.9% (GC) (Actu-All Chemi-
cals); Chloroform, HPLC (stabilized with ethanol) 99.9% (Actu-All
Chemicals); trimethylsulfonium hydroxide (TMSH), purum, ꢀ0.25m
in methanol (Sigma–Aldrich); dodecane, ReagentPlus ꢁ99%
tion of the amount of inhibition agent. Thus, although deoxy-
genation activity is dependent on the catalyst support, there is
no correlation between support material characteristics and
catalyst inhibition. Differences in, for instance, the coordination
of the feed and products to the support material or differences
in metal–support interaction apparently do not significantly
contribute to the inhibition process. Instead, these results indi-
cate that inhibition is predominantly caused by the adsorption
of unsaturated compounds on the active palladium surface, in-
dependent of the support material used.
(
Sigma–Aldrich); Tetradecane, olefin-free ꢁ99% (GC) (Sigma–Al-
drich); n-hexane, PEC Grade 97.0% (Actu-All Chemicals); 1,7-octa-
diene, 98% (Sigma–Aldrich); Pentadecane, ꢁ99% (Sigma–Aldrich);
1
-pentadecene, 98% (Sigma—Aldrich); heptadecane 99%, (Sigma–
Aldrich); hexadecanal, >97% (TCI Europe N.V.); 1-heptadecanol,
98% (Sigma–Aldrich); stearyl alcohol, >99% (GC) (TCI Europe N.V.);
stearic acid, 97% (Acros Organics), oleic acid, tech., 90% (Sigma–
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 2648 – 2655 2653

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Aldrich); oleic acid, reagent grade ꢀ99% (Sigma–Aldrich); linoleic
acid, ꢁ99% (SAFC, Sigma–Aldrich); linolenic acid, ꢀ70% (GC)
parameters and settings were used: Varian Select biodiesel for
glycerides (15 mꢂ0.32 mmꢂ0.45 mm), 3708C injection port tem-
perature, 1 mL injection volume, splitless injection mode,
(
(
Sigma–Aldrich); 1-phenylundecane, 99% (undecylbenzene)
Sigma–Aldrich); pyrene 98% (Sigma–Aldrich); 2,3-benzanthracene,
À1
À1
60 mLmin split flow, ignition threshold 0.5 pA, 350 mLmin air
À1
À1
9
8% (tetracene) (Sigma–Aldrich); 18-pentatriacontanone (stearone)
flow, 35 mLmin hydrogen flow, 40 mLmin makeup gas flow,
(
Alfa Aesar); Stearic anhydride (TCI Europe N.V.).
and flame ionization detector at 3808C.
Catalyst analysis
Catalytic testing
Nitrogen physisorption was performed at À1968C by using a Micro-
meritics Tristar 3000V, 6.04 ꢃ. The obtained data were used to cal-
culate the BET surface area. Prior to the physisorption measure-
ments, the samples were dried at 2008C for approximately 14 h
under nitrogen flow. Micropore volumes and external surface areas
were determined by using t-plot analysis.
The catalyst (typically 0.5 g) was activated prior to catalytic reac-
tions in a plug-flow reactor in which the catalyst was heated to
À1
À1
1
508C (58Cmin ) in nitrogen (3.0 Linde, 200 mLmin ) atmos-
phere, dried for 30 min and subsequently reduced at 1508C for 2 h
À1
in hydrogen (5.0 Linde, 200 mLmin ). After reduction, the catalyst
was flushed with nitrogen at 1508C to remove all hydrogen from
the reactor and catalyst surface and cooled to RT in nitrogen
Total palladium surface area and palladium dispersion was deter-
mined by hydrogen chemisorption measurements using a Micro-
meritics ASAP2020 apparatus. The samples were dried at 1208C in
À1
(
200 mLmin ). The plug-flow reactor was subsequently transferred
to a glove box to add solvent, reactant and internal standard prior
to catalytic reactions.
He flow for 30 min and subsequently reduced in 50% H /He flow
2
À1
at 2508C for 2 h (58Cmin ). The H2 adsorption isotherms were
Catalytic experiments were performed in a multiple reactor system
measured at 1508C, and calculations for the percentage Pd disper-
sions were performed assuming a complete Pd reduction with the
stoichiometry of one hydrogen atom adsorbed per Pd surface
atom.
(
Parr, model MRS 5000) containing six 75 mL vessels with magnetic
stirring. Pressure and temperature were monitored continuously
during catalytic tests using this setup. The reaction mixture was
stirred (1100 rpm) to avoid external mass transfer limitations. It was
determined that pore diffusion limitations are negligible according
to the Weisz–Prater criterion (the liquid-phase effective diffusivities
were estimated according to Wilke and Chang ). In a typical reac-
tion, amounts of 1 g feed and 18 g dodecane (25 mL) as a solvent
were used. Tetradecane (0.5 g, 2.5 mmol) was used as an internal
standard for quantitative analysis by GC.
Elemental analysis was performed at the Mikroanalytisches Labora-
torium Kolbe to determine the exact palladium loading of the
dried catalysts.
[21]
Organic deposits on the catalyst surface were analyzed by TGA
with a PerkinElmer STA6000. Samples were heated with a heating
À1
rate of 108Cmin from 308C to 8008C by using nitrogen as
À1
a purge gas with a flow rate of 30 mLmin . 180 mL ceramic pans
were used to hold the sample. The total organic content was esti-
mated from the weight loss between 150 and 8008C.
Product analysis
Reaction-mixture work-up after reaction and cooling was per-
formed with a chloroform methanol mixture (2:1) that was heated
to 408C to ensure complete dissolution of the fatty acid. A sample
was subsequently taken from the mixture and methylated by
mixing with trimethylsulfonium hydroxide, an in situ esterification
reagent to derivatize carboxylic acids for accurate gas–liquid chro-
matography quantification. Soxhlet extraction of the spent catalyst
was performed (by using chloroform, methanol, and hexane) to
remove physisorbed organic species from the catalyst surface
before further analysis.
FTIR spectra of the fresh and spent catalysts were obtained on
a Varian Scimitar 1000 FT-IR spectrometer equipped with a DTSG
À1
detector. The spectra were collected in the range 4000–650 cm
À1
(
measurement resolution at 4 cm ) by using attenuated total re-
flectance with a diamond w/ZnSe lens single reflection plate with
12 co-added scans. The sample chamber was purged with N gas
5
2
for at least 10 min before scanning.
Acknowledgements
Analysis of the reaction mixture (products with molecular weights
À1
between ꢀ180–350 gmol ) was performed on a Thermo Focus
The authors would like to thank Herman de Beukelaer BSc. and
Marina van Leeuwen BSc. for the TGA measurements, Arjan
gas chromatograph equipped with an automatic injection system
(
2
AS3000 autosampler). GC program: Hold 1 min at 508C, ramp
den Otter MSc. for the H -chemisorption and Nazila Masoud MSc.
À1
À1
À1
2
08Cmin to 1708C, ramp 58Cmin to 1958C, ramp 158Cmin
À1
for the N -physisorption measurements. This research was per-
2
to 2258C, then ramp 58Cmin to 2508C and hold at 2508C for
formed within the framework of the CatchBio program. The au-
thors gratefully acknowledge support of the Smart Mix Program
of the Netherlands Ministry of Economic Affairs and the Nether-
lands Ministry of Education, Culture and Science.
1
5 min. The following parameters and settings were used: Varian
CP-FFAP (free fatty acids) GC column (25 mꢂ0.32 mmꢂ0.30 mm),
908C injection port temperature, 1 mL injection volume, split in-
2
À1
À1
jection mode, 20 mLmin split flow, column flow of 1.1 mLmin
,
hydrogen and air flow control, ignition threshold 0.5 pA, and flame
ionization detector at 2708C.
Keywords: adsorption · alkenes · fatty acids · hydrogen ·
Analysis of high-boiling-point substrates (e.g., stearyl stearate and
stearone) was performed on an Interscience Trace 1300 gas chro-
matograph equipped with an automatic injection system (AS1310
palladium
[
1] a) P. Mꢄki-Arvela, I. Kubickova, M. Snꢀre, K. Erꢄnen, D. Y. Murzin, Energy
Fuels 2007, 21, 30–41; b) I. Kubi cˇ kovꢅ, M. Snꢀre, K. Erꢄnen, P. Mꢄki-
Arvela, D. Y. Murzin, Catal. Today 2005, 106, 197; c) S. Lestari, P. Mꢄki-
Arvela, H. Bernas, O. Simakova, R. Sjçholm, J. Beltramini, G. Q. M. Lu, J.
À1
autosampler). GC program: Hold 1 min at 508C, ramp 158Cmin
À1
À1
to 1508C, ramp 78Cmin to 1808C, ramp 158Cmin to 2808C,
À1
ramp 78Cmin to 3508C and hold 10 min at 3508C. The following
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 2648 – 2655 2654

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Myllyoja, I. Simakova, D. Y. Murzin, Energy Fuels 2009, 23, 3842–3845;
d) A. T. Madsen, B. Rozmysłowicz, I. L. Simakova, T. Kilpiç, A.-R. Leino, K.
Kordꢅs, K. Erꢄnen, P. Mꢄki-Arvela, D. Y. Murzin, Ind. Eng. Chem. Res. 2011,
[10] J. J. F. Scholten, A. P. Pijpers, M. L. Hustings, Catal. Rev. Sci. Eng. 1985, 27,
151–206.
[11] a) A. T. Madsen, E. H. Ahmed, C. H. Christensen, R. Fehrmann, A. Riisager,
Fuel 2011, 90, 3433–3438; b) L. Boda, G. Onyestyꢅk, H. Solt, F. Lꢇnyi, J.
Valyon, A. Thernesz, Appl. Catal. A 2010, 374, 158–169; c) M. Snꢀre, I.
Kubi cˇ kovꢅ, P. Mꢄki-Arvela, K. Erꢄnen, J. Wꢄrnꢀ, D. Yu. Murzin, Chem. Eng.
J. 2007, 134, 29–34; d) J. Han, J. Duan, P. Chen, H. Lou, X. Zheng, Adv.
Synth. Catal. 2011, 353, 2577–2583; e) S. Lestari, P. Mꢄki-Arvela, I. Sima-
kova, J. Beltramini, G. Lu, D. Murzin, Catal. Lett. 2009, 130, 48–51; f) A. S.
Berenblyum, R. S. Shamsiev, T. A. Podoplelova, V. Y. Danyushevsky, Russ.
J. Phys. Chem. 2012, 86, 1199–1203; g) A. S. Berenblyum, T. A. Podople-
lova, R. S. Shamsiev, E. A. Katsman, V. Y. Danyushevsky, Petrol. Chem.
2011, 51, 336–341.
5
0, 11049–11058; e) J. G. Immer, M. J. Kelly, H. H. Lamb, Appl. Catal. A
010, 375, 134–139.
2] a) T. Morgan, D. Grubb, E. Santillan-Jimenez, M. Crocker, Top. Catal.
010, 53, 820–829; b) E. Santillan-Jimenez, T. Morgan, J. Lacny, S. Moha-
2
[
2
patra, M. Crocker, Fuel 2013, 103, 1010–1017; c) P. Do, M. Chiappero, L.
Lobban, D. Resasco, Catal. Lett. 2009, 130, 9–18; d) S. A. W. Hollak, J. H.
Bitter, J. van Haveren, K. P. de Jong, D. S. van Es, RSC Adv. 2012, 2,
9387–9391; e) B. Rozmysłowicz, P. Mꢄki-Arvela, A. Tokarev, A.-R. Leino,
K. Erꢄnen, D. Y. Murzin, Ind. Eng. Chem. Res. 2012, 51, 8922–8927; f) T.
Morgan, E. Santillan-Jimenez, A. E. Harman-Ware, Y. Ji, D. Grubb, M.
Crocker, Chem. Eng. J. 2012, 189–190, 346–355; g) J. A. Botas, D. P. Ser-
rano, A. Garcꢆa, J. de Vicente, R. Ramos, Catal. Today 2012, 195, 59–70;
h) J. Han, H. Sun, Y. Ding, H. Lou, X. Zheng, Green Chem. 2010, 12, 463–
[12] J. P. Ford, J. G. Immer, H. H. Lamb, Top. Catal. 2012, 55, 175–184.
[13] J. M. Davidson, C. M. McGregor, L. K. Doraiswamy, Ind. Eng. Chem. Res.
2001, 40, 101–107.
4
67; i) R. W. Gosselink, S. A. W. Hollak, S.-W. Chang, J. van Haveren, K. P.
[14] A. Vonortas, D. Kubi cˇ ka, N. Papayannakos, Fuel 2014, 116, 49–55.
[15] J. W. E. Coenen in Margarine Today, Brill, Leiden, 1970, pp. 62–91.
[16] A. Bailey, D. Mitcham, A. French, G. Sumrell, J. Am. Oil Chem. Soc. 1975,
52, 196–197.
[17] G. C. Bond, Discuss. Faraday Soc. 1966, 41, 200–214.
[18] a) W. Dong, V. Ledentu, P. Sautet, A. Eichler, J. Hafner, Surf. Sci. 1998,
411, 123–136; b) H. Conrad, G. Ertl, E. E. Latta, Surf. Sci. 1974, 41, 435–
446.
de Jong, J. H. Bitter, D. S. van Es, ChemSusChem 2013, 6, 1576–1594.
3] a) M. Snꢀre, I. Kubickovꢅ, P. Mꢄki-Arvela, D. Chichova, K. Erꢄnen, D. Y.
Murzin, Fuel 2008, 87, 933–945; b) B. Rozmysłowicz, P. Mꢄki-Arvela, S.
Lestari, O. Simakova, K. Erꢄnen, I. Simakova, D. Murzin, T. Salmi, Top.
Catal. 2010, 53, 1274–1277.
4] M. Snꢀre, I. Kubi cˇ kovꢅ, P. Mꢄki-Arvela, K. Erꢄnen, D. Y. Murzin, Ind. Eng.
Chem. Res. 2006, 45, 5708–5715.
5] a) P. Mꢄki-Arvela, M. Snꢀre, K. Erꢄnen, J. Myllyoja, D. Y. Murzin, Fuel
[
[
[
[19] a) T. K. Phung, A. A. Casazza, B. Aliakbarian, E. Finocchio, P. Perego, G.
Busca, Chem. Eng. J. 2013, 215, 838–848; b) M. Hasegawa, M. Low, J.
Colloid Interface Sci. 1969, 30, 378–386.
2
5
008, 87, 3543–3549; b) J. G. Immer, H. H. Lamb, Energy Fuels 2010, 24,
291–5299.
[
6] E. W. Ping, J. Pierson, R. Wallace, J. T. Miller, T. F. Fuller, C. W. Jones, Appl.
Catal. A 2011, 396, 85–90.
[20] L. Zhang, R. He, H.-C. Gu, Appl. Surf. Sci. 2006, 253, 2611–2617.
[21] C. R. Wilke, P. Chang, AIChE J. 1955, 1, 264–270.
[
[
7] A. P. Borole, Biocatalysis in Oil Refining, Elsevier, Amsterdam, 2011.
8] H. Bernas, K. Erꢄnen, I. Simakova, A.-R. Leino, K. Kordꢅs, J. Myllyoja, P.
Mꢄki-Arvela, T. Salmi, D. Y. Murzin, Fuel 2010, 89, 2033–2039.
9] I. Simakova, O. Simakova, P. Mꢄki-Arvela, D. Y. Murzin, Catal. Today 2010,
[
Received: May 2, 2014
150, 28–31.
Published online on July 30, 2014
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 2648 – 2655 2655