Hydrodeoxygenation of cellulose pyrolysis model compounds using molybdenum oxide and low pressure hydrogen

Article abstract of DOI:10.1039/c7gc01477e

A molybdenum oxide catalyst in a low pressure hydrogen atmosphere was used for the hydrodeoxygenation (HDO) of pulsed injections of cellulose pyrolysis model compounds to examine reaction products. Higher catalyst loadings (≥20: 1 catalyst: cellulose injection) in the HDO reactor were found to preferentially produce alkanes, while at lower loadings (≤10: 1 catalyst: cellulose injection) alkene selectivity was increased. However, as the amount of catalyst was decreased, the pyrolysis vapors were not completely deoxygenated. The HDO of monofunctional oxygenated C₄ compounds found hydroxyl groups to be the most readily reacted and ether linkages to be the most recalcitrant. In general, the reactivity towards deoxygenation of the tested oxygen-containing functional groups was observed to be C-OH > C=O > C-OC. Several cellulose pyrolysis model compounds were also tested, including methyl glyoxal, glycolaldehyde, furfural, 5-hydroxymethylfurfural, and levoglucosan, and found the same general trend to occur, except for levoglucosan, which was totally reacted and did not yield any oxygenated low molecular weight compounds despite containing two ether linkages. Across the compounds, the general reaction pathway was observed to include carbonyl/alcohol hydrogenation/dehydrogenation, deoxygenation, and alkene isomerization and hydrogenation.

Full text of DOI:10.1039/c7gc01477e

View Article Online
View Journal
Green
Chemistry
Accepted Manuscript
This article can be cited before page numbers have been issued, to do this please use: M. W. Nolte, A.
Saraeian and B. H. Shanks, Green Chem., 2017, DOI: 10.1039/C7GC01477E.
Volume 18 Number
7
7
April 2016 Pages 1821–2242
This is an Accepted Manuscript, which has been through the
Royal Society of Chemistry peer review process and has been
accepted for publication.
Green
Chemistry
Cutting-edge research for a greener sustainable future
www.rsc.org/greenchem
Accepted Manuscripts are published online shortly after
acceptance, before technical editing, formatting and proof reading.
Using this free service, authors can make their results available
to the community, in citable form, before we publish the edited
article. We will replace this Accepted Manuscript with the edited
and formatted Advance Article as soon as it is available.
You can find more information about Accepted Manuscripts in the
author guidelines.
Please note that technical editing may introduce minor changes
to the text and/or graphics, which may alter content. The journal’s
standard Terms & Conditions and the ethical guidelines, outlined
in our author and reviewer resource centre, still apply. In no
event shall the Royal Society of Chemistry be held responsible
for any errors or omissions in this Accepted Manuscript or any
consequences arising from the use of any information it contains.
ISSN 1463-9262
CRITICAL REVIEW
G. Chatel et al.
Heterogeneous catalytic oxidation for lignin valorization into valuable
chemicals: what results? What limitations? What trends?
rsc.li/green-chem

Page 1 of 11
Green Chemistry
View Article Online
DOI: 10.1039/C7GC01477E

Green Chemistry
Page 2 of 11
View Article Online
DOI: 10.1039/C7GC01477E
to use industrially, and transition metal sulfides, which may re-
ides. Using monooxygenated pyrolysis model compounds having
ketone, furan, or aromatic oxygen functionality, Román-Leshkov
and coworkers found MoO3 to be an effective catalyst in their di-
rect hydrodeoxygenation to yield alkenes and aromatics at high
quire a H S co-feed to maintain catalytic activity.
2
A potentially more advantageous upgrading strategy would be
to upgrade the pyrolysis vapors leading to a stabilized, deoxy-
genated bio-oil obtained upon condensation. Vapor phase deoxy-
genation could be achieved by co-feeding biomass and catalyst
into the pyrolysis reactor (in situ) or placing the catalyst in a sep-
arate reactor downstream from the pyrolyzer (ex situ). In situ
upgrading has the benefit of requiring the use of a single reactor
and being more process intensive, while ex situ upgrading allows
for the use of different operating conditions in each reaction step.
Common petroleum refinery catalysts, such as zeolites, have
been tested for the deoxygenation of biomass pyrolysis vapors.
HZSM-5 has been found to be able to generate aromatics in the
selectivity28–30
.
The HDO was performed at relatively mild
◦
temperature of ≤400 C and atmospheric hydrogen pressure.
The dominant hydrocarbon products were alkenes or aromatics,
which generally contained the same number of carbon atoms as
the starting reactant. From the non-aromatic model compounds,
the selectivity to hydrocarbons was >98%. Under the reaction
conditions, the aromatic model compounds could be partially de-
oxygenated to form mainly phenol or anisole, or totally deoxy-
genated to benzene or toluene. Further, supporting MoO3 on TiO2
or ZrO2 gave an increased reaction rate for m-cresol deoxygena-
tion, while maintaining a high selectivity to hydrocarbons. The
use of a support was found to prolong catalyst activity compared
with bulk MoO3. The support was hypothesized to stabilize the
catalytic fast pyrolysis of cellulose, hemicellulose, and lignin17
.
However, the hydrocarbon yields were low (<30 C%), while coke
yields were as high as 80 C%. Carlson et al. reported a sim-
ilar trend in the catalytic fast pyrolysis of cellulose and biomass
5+
Mo oxidation state, which was the proposed active species. For
all of the tested model compounds and catalysts, the amount of
carbon that was converted to coke was minimal and CO and CO2
were not detected as reaction products. In addition, Mo-based
catalysts have been found to be effective in the condensed phase
where aromatic yields were <30 C% and CO, CO , and coke yields
2
were significant18. Unfortunately, zeolites and other silica alu-
mina catalysts have been found to give low to moderate hydro-
carbon yields and also to rapidly deactivate due to loss of acid
sites or coking and had poor hydrothermal stability 19–22. In ad-
dition, by performing the catalytic fast pyrolysis in an inert atmo-
sphere a noticeable amount of deoxygenation occurred through
deoxygenation of lignocellulosic biomass31–33
.
Examining the full suite of pyrolysis vapors from cellulose,
lignin, or corn stover feedstocks, we have previously shown that
decarbonylation (CO) and decarboxylation (CO ) leading to low
2
MoO was found to be able to effectively catalyze the formation
3
hydrocarbon yields.
of alkanes and aromatics34. As shown, MoO is a promising cat-
3
An alternative deoxygenation route would be to perform the
catalysis in a hydrogen atmosphere to facilitate the hydrodeoxy-
genation (HDO) of the pyrolysis vapors. In a hydrogen atmo-
alyst for biomass vapor phase hydrodeoxygenation, however, in
our initial work high catalyst loadings were employed in the HDO
reactor, which resulted in the saturation of all alkenes to alkanes.
Therefore, in this work we looked at the HDO of cellulose pyroly-
sis vapors at lower MoO3 loadings as a means of increasing alkene
yields. Further, simplified model compounds were investigated to
gain a more fundamental understanding of general reaction path-
ways and relative reactivities of different oxygen moieties for hy-
drodeoxygenation. Herein, model compounds containing a single
or multiple oxygen atoms and functional groups were used as the
feedstock. These model compounds, representative of low molec-
ular weight pyrolysis compounds, furanic, and anhydrosugar cel-
lulose pyrolysis products, were hydrodeoxygenated over a MoO3
catalyst under a low pressure H2 atmosphere. The model com-
pounds included simpler C4 monooxygenates as well as major
cellulose pyrolysis products, such as methyl glyoxal (MG), gly-
colaldehyde (GA), 5-hydroxymethylfurfural (HMF), and levoglu-
cosan (LG).
sphere, dehydration (H O) would be the primary method of oxy-
2
gen removal, which in turn would maximize the product car-
bon yield. A diverse array of catalysts have been investigated
in the literature for HDO efficacy including transition metal car-
bide, nitride, phosphide, and sulfide, and noble metal catalysts,
which have been reviewed in a recent publication23. Bhan and
coworkers have carried out deoxygenation studies of pyrolysis
compounds using Mo C24–27. The catalyst was found to be
2
much more selective towards C=O bond cleavage than C-C bond
cleavage, as evidenced by the higher conversion of furfural to
2
-methylfuran (55-70% selectivity) versus furan (<1%)25. Us-
ing anisole as the feed, the catalyst was found to be selective to-
wards cleavage of the Caromatic-O bond over the Cmethoxy-O bond.
The selectivity towards benzene was >90% while phenol selec-
tivity was <1%26. However, the Mo C was found to be active
2
for hydrogenation leading to saturated products. From anisole,
the selectivity to cyclohexane and methylcyclopentane was up to
2
Experimental Methods
1
0%26. Alkene hydrogenation was more dramatic in acetone hy-
Cellulose (Sigmacell 50), levoglucosan (99%), furfural (99%),
5-hydroxymethylfurfural (99%), glycolaldehyde, methyl glyoxal
(40% in water), and butanal (>99%) were purchased from Sigma
Aldrich and 1-butanol, 2-butanol, methyl ethyl ketone, furfural,
furan, and tetrahydrofuran were purchased from Fisher Scien-
tific as Certified ACS grade. All reactant chemicals were used as
received. n-Butane (2% in He) and 1-butene (2% in He) were
obtained from Airgas.
drodeoxygenation, where propane was obtained at a selectivity
of over 80%27. These model compound studies show Mo C is a
2
promising biomass HDO catalyst, however further studies will be
needed with biomass pyrolysis vapors as the feedstock. Also, it
would be more advantageous economically if hydrogen efficiency
was maximized by minimizing hydrogenation reactions and max-
imizing alkene and aromatic yields.
Another interesting class of catalysts are transition metal ox-
The molybdenum(VI) oxide (>99%, Sigma Aldrich) was pel-
2
|
1–10

Page 3 of 11
Green Chemistry
View Article Online
DOI: 10.1039/C7GC01477E

Green Chemistry
Page 4 of 11
View Article Online
DOI: 10.1039/C7GC01477E
reduction in the catalyst amount was investigated at four differ-
From the alcohol-containing model compounds, 92-96% of the
hydrocarbons were C products. Among the remaining C model
ent MoO :cellulose loadings (2:1, 10:1, 20:1, 200:1) to find if an
3
4
4
increase in alkene yield could be obtained (Figure 1). The highest
loading of 200:1 yielded similar results to previous work in that
alkanes and aromatics were the dominant products34. Decreas-
compounds, a lower selectivity to butanes and butenes was ob-
served, although C s were still the major hydrocarbon product.
4
Carbon-carbon bond coupling reactions were found to be min-
imal, as the yield of pentanes and larger aliphatics was generally
ing the MoO mass by an order of magnitude to a 20:1 loading
3
resulted in a higher yield of alkenes of 36 C%, but alkanes were
still a major product at 32 C%. The highest alkene yield and se-
lectivity (among hydrocarbons) of 46 C% and 64%, respectively,
were achieved from a loading of 10:1. Although, the pyrolysis
vapors were not fully deoxygenated, as a 9 C% oxygenate yield
was obtained in the product stream. The lowest tested loading of
<
1 C% and aromatic products were not detected. Although, some
C8 condensation products were detected, with the highest yield
being obtained from butanal at about 5 C%.
In terms of carbon-carbon bond cleavage, 1-butanol and 2-
butanol were found to have the lowest degree of chain cleavage,
giving ∼85-90 C% yield of C products. The carbonyl compounds,
4
2:1 saw the lowest degree of deoxygenation with an oxygenate
butanal and 2-butanone, and ether compound, THF, showed a
yield of 59 C%. However, the selectivity to alkenes among the
hydrocarbon products was 63%, which was similar to the alkene
selectivity at a 10:1 loading. The oxygenates produced at a 2:1
and 10:1 loading included various furanics, anhydrosugars, and
low molecular weight pyrolysis compounds.
higher degree of C-C cleavage as the C selectivity was 70-90%.
4
C-C bond cleavage was the most dramatic from butyric acid where
the yield of C -C hydrocarbons was 9.3 C%. Correspondingly,
1
3
butyric acid gave the highest observed yields of CO and CO at 0.8
2
C%, indicating decarbonylation and decarboxylation may have
been a more significant deoxygenation pathway. However, ex-
amination of the model compounds found no discernible trend
between CO/CO and C -C yields, meaning other reactions in-
The cellulose pyrolysis product distribution has been shown
to contain products with a diverse array of oxygen functionali-
ties, including carboxylic acids, alcohols, aldehydes, ethers, ke-
tones, furans, and others9,35,36. This diversified product distribu-
tion could be entirely deoxygenated provided the catalyst loading
was sufficiently high (Figure 1). However, deoxygenating the full
product slate from cellulose pyrolysis did not allow for exami-
nation of the general reactivities and reaction pathways of the
various pyrolysis compounds. Therefore, the use of model com-
pounds were used to facilitate a better understanding of the HDO
reaction and possibly aid in engineering increased alkene yields
and hydrogen efficiency.
2
1
3
volving C-C bond cleavage may have occurred. There was a min-
imal loss of carbon to coke, as coke yields were <1.1 C% for all
tested compounds.
Hydrogenation and dehydrogenation products were observed
for carbonyl and alcohol reactants, respectively. For example, 1-
butanol was a product in butanal HDO, and butanal was identified
as a product in 1-butanol HDO. The same trend was observed
with 2-butanol and 2-butanone. In butyric acid HDO, the partially
deoxygenated butanal was a major product. Conversely, butanal
and 1-butanol were not observed to be oxidized to butyric acid by
the catalyst.
3
.2 HDO of C Monooxygenates
4
Previously, MoO has been used in the HDO of ketonic (acetone,
3
2
2
-hexanone, and cyclohexanone) and furanic (2-methylfuran and
,5-dimethylfuran) pyrolysis model compounds28. In this study,
3.3 HDO of Cellulose Pyrolysis Model Compounds
In cellulose pyrolysis, furanic compounds can account for ∼6
wt.% of the products37. In this study, three of the more promi-
the model compounds have been expanded to include C4 com-
pounds having alcohol (1-butanol and 2-butanol), carbonyl (bu-
tanal and 2-butanone), carboxylic acid (butyric acid), or ether
nent furanic pyrolysis products, furan, furfural, and HMF, were
investigated (Table 2). Among the furanic compounds, furan
exhibited the lowest conversion at 22.3%. Furfural and HMF
had higher conversions at 83.7% and 100%, respectively, al-
though, they yielded a greater amount of oxygenates (40-50 C%).
These oxygenates included 2-methylfuran from furfural and 2,5-
furandicarboxaldehyde, 2,5-dimethylfuran, and 5-methylfurfural
from HMF. In fact, all of the oxygenated products contained a fu-
ran ring, while no aliphatic oxygenates were detected. As a result
of the higher oxygenate yields, the overall selectivities to hydro-
carbons from furfural and HMF were lower than from furan. For
furfural and HMF, hydrocarbon selectivity was 25-35 mol%, and
for furan the selectivity was 83.2%. As with the C4 monooxy-
genates, the selectivity from the furanics towards alkenes was
similar or greater than the selectivity towards alkanes.
(
tetrahydrofuran (THF)) functional groups (Table 1).
The conversions of the alcohol-containing model compounds
were near 100%, while the carbonyl and carboxylic acid-
containing compounds had conversions around 50-60%. The
ether model compound, THF, had the lowest conversion at ∼28%.
The oxygenated reactants were found to be effectively deoxy-
genated to primarily aliphatic hydrocarbons, as the selectivity to
alkenes and alkanes was generally >90%. In all cases, the se-
lectivity towards alkenes was about equal to or greater than to
alkanes. The conversion of 2-butanol and 2-butanone were the
most selective to alkenes with values of ∼63-69 mol%. Conver-
sion of butyric acid gave the lowest selectivity to hydrocarbons
(
55%), which was due to the higher yields of oxygenated prod-
ucts, CO/CO , and self-condensation products (C8 ketones and
2
Furan and THF can be compared to see the effect of having un-
saturated C-C bonds. THF gave a slightly higher conversion, but
both were low, <30 mol%. A higher selectivity towards alkenes
was achieved with furan, likely due to the fact that it contains
ester).
The product hydrocarbons were found to generally contain at
least the same number of carbon atoms as the original reactant.
4
|
1–10

Page 5 of 11
Green Chemistry
View Article Online
DOI: 10.1039/C7GC01477E
Table 1 The selectivities of alkanes and alkenes (C%) and yields (C%) of products from the HDO of C4 model compounds. Reaction conditions: 120
◦
mL/min H2, 0.2 µL feed injection, 0.8 mg MoO3, 400 C catalyst bed, catalyst pre-reduced 1 hr.
Feed
1-Butanol
99.7
47.8
Butanal
52.4
46.5
Butyric Acid
62.6
25.6
2-Butanol
99.9
35.2
2-Butanone
55.5
27.4
THF
28.2
48.8
49.9
Conversion (C%)
Selectivity to Alkanes (C%)
Selectivity to Alkenes (C%)
48.6
46.0
30.6
63.1
68.5
Yield (C%)
Hydrocarbons
C1-C3
7.3
85.0
0.1
4.5
37.4
4.6
9.3
25.8
0.3
3.5
91.3
0.1
2.6
42.5
2.7
5.1
17.8
0.1
Butanes/Butenes
C5+
Oxygenates
1-Butanol
Butanal
-
2.6
-
0.5
17.4
0.0
0.0
0.0
-
0.0
0.0
0.5
-
0.0
0.0
0.0
0.0
0.1
0.2
95.0
2.8
0.0
0.0
0.2
0.4
96.2
2
2
-Butanol
-Butanone
0.0
0.0
0.2
0.1
97.9
0.0
0.8
0.4
0.6
0.6
96.6
CO/CO2
Coke
0.3
1.1
94.5
0.7
Total Carbon Balance (C%)
93.0
unsaturated carbon-carbon bonds while THF does not. A higher
cluded acetone (58.8 C%), propanal (9.4 C%), propane (4.2 C%)
and propylene (5.6 C%). In contrast with the C4 model com-
pounds, no carbonyl hydrogenation products were detected (e.g.,
yield of C products was obtained from THF (17.8 C%) as com-
4
pared with furan (11.2 C%), demonstrating carbon-carbon bond
cleavage was more prevalent in furan HDO. One possible means
of reducing the carbon chain length in furan HDO may have been
through the production of CO/CO2 where the yield was higher
no propanol or hydroxyacetone was detected). CO/CO forma-
2
tion was also significant at a yield of 4.1 C%. In conjunction with
C-C bond cleavage to form CO and CO2 was the formation of
from furan than the CO/CO yield from THF. The coke yields for
2
acetaldehyde (5.2 C%) as well as a low yield of C1 and C hy-
2
furan and THF were minor at <1 C%.
drocarbons. C-C bond coupling was observed to have occurred to
form C -C hydrocarbons at a combined yield of 3.3 C%.
Furfural and furan were compared to see the effect of adding
a carbaldehyde group to the furan ring. The carbonyl group was
found to be more reactive than the ring oxygen, however the se-
lectivity to hydrocarbons was 24 mol% due to the lower degree of
4
6
GA, a 2-carbon molecule with hydroxy and aldehyde function-
ality, was fully converted over the MoO3 catalyst. The hydro-
carbon yield was 51.8 C% with the remaining carbon being con-
furan ring deoxygenation and higher oxygenate yield. The C car-
4
verted to oxygenates (29.7 C%), CO and CO (4.3 C%), and coke
2
bonyl compounds were found to be hydrogenated to a minor de-
gree to alcohols, however the furfural hydrogenation product, 2-
furanmethanol, was not observed. Instead, the oxygenated prod-
uct was mainly 2-methylfuran (45.9 C%) and a minor amount of
furan (3.4 C%) and larger condensation products (0.4 C%). The
(11.1 C%). Ethane and ethylene were the dominant hydrocarbon
products, although larger chain C -C hydrocarbons were pro-
3
4
duced in elevated yields. The yield of ethane was 20.2 C% and of
ethylene was 9.7 C%. Among the oxygenates, acetaldehyde was
the dominant product at 23 C%. Acetone and furan were also
formed at about 3.4 C% each. Ethylene glycol or glyoxal were
not detected. The yield of coke was also fairly high at 11.1 C%.
yield of CO and CO was slightly higher at 0.9 C%.
2
HMF is a doubly substituted furan, containing a carbaldehyde
bonded to the 2-carbon, same as furfural, and a hydroxymethyl
bonded to the 5-carbon. The HDO of HMF was found to proceed
similarly to that of furfural, in that the ring-substituents were
more readily converted or deoxygenated as compared with the
furan ring oxygen atom. In the case of HMF, 2,5-dimethylfuran
was the major product (33.9 C%), though 2-methylfuran (5.0 C%
Levoglucosan is the major product from cellulose pyrolysis with
its yield typically being 50-60 wt.%. It was found to be fully
converted over the MoO3. The HDO products were found to
evolve off the catalyst bed at a slower rate leading to broad chro-
matogram peaks and some significant peak overlapping. As such,
it was not possible to quantify accurately the yields of the prod-
ucts that eluted at the early retention times, namely CO/CO2,
methane, and ethane. The ethylene peak was also broad, but
did not overlap with other peaks so its yield could more easily
be quantified. These unquantified products were likely the ma-
jority of the 59.1 C% of unaccounted carbon in the total carbon
balance.
)
and 5-methylfurfural (1.4 C%) were also detected. As with fur-
fural, no alcohol-containing oxygenates were observed from HMF
HDO.
Two low molecular weight cellulose pyrolysis products were
investigated, methyl glyoxal (MG) and glycolaldehyde (GA). MG,
a 3-carbon molecule containing ketone and aldehyde functional
groups, was nearly fully converted over the catalyst bed and
found to afford partially deoxygenated compounds and fully de-
oxygenated compounds as major products. These products in-
Levoglucosan, along with GA, gave a higher selectivity towards
alkanes over alkenes, though LG was the only model compound
to generate aromatic products. Of these alkanes, n-pentane, n-
1
–10 | 5

Green Chemistry
Page 6 of 11
View Article Online
DOI: 10.1039/C7GC01477E

Page 7 of 11
Green Chemistry
View Article Online
DOI: 10.1039/C7GC01477E
produce cis- and trans-2-butene in about equal amounts with the
cis/trans ratio being 0.9 (Table 3), indicating the isomerization
detected.
The relative concentration of the Mo + phase was lower in this
work when compared with the results from Román-Leshkov and
coworkers29. Though, Román-Leshkov et al. only reported XPS
results for post-reaction catalysts and hypothesized that the for-
mation of an oxycarbohydride (MoOxCyHz) phase could improve
catalyst stability and lead to a higher percentage of Mo5+ atoms.
In this work, even after four cellulose injections, the relative con-
centration of Mo + was not more than 2.3%. The lower concen-
tration of Mo + may have been due to the fact that cellulose was
injected batch-wise in pulses and not continuously. Between the
injections, the H2 flow was maintained, which would have contin-
uously reduced the catalyst, possibly leading to a higher concen-
tration of Mo +. Also, the catalyst was passivated to prepare the
sample for XPS analysis, which may have resulted in oxidation of
5
was likely acid-catalyzed rather than base-catalyzed.
A qualitative analysis of the acidity of fresh and reduced MoO3
was performed using pyridine TPD (Figure 2). Fresh MoO was
3
found to only contain a single pyridine desorption peak at a tem-
◦
perature about 200 C, indicating the presence of weak acid sites.
◦
Another sample of fresh MoO was then reduced at 400 C un-
3
5
der 120 mL/min of H flow for 1 hr. The reduced catalyst was
2
5
found to contain a dramatically increased amount of acidity, both
in terms of the number of acid sites as well as the strength of
◦
acid sites. A wide peak was observed that ranged from 175 C to
◦
00 C with apparent peak maxima around 400 and 600 C. No
◦
◦
7
4
pyridine was detected when a blank catalyst bed containing sand
and quartz wool was used.
Mo to Mo6+. There was a noticeable increase in the formation
of lower valence Mo oxidation states, which would be consistent
with the hypothesized HDO reaction pathways presented in the
5+
!6#ꢀ1+2,%0&
<
6#'+*3)+*#%0&
<
0?E
%0?E
%
5
:99D6
5><4:D6
literature29,39
.
0=E
==4@D6
4 Discussion
%
5
The hydrodeoxygenation of acetone has been performed over
MoO and Mo C catalysts, although the reactions conditions were
somewhat different between the two studies27,28. Both reactions
3
2
%
5
0>E
;4;D6
2
40
236
232
228
224
244
240
236
232
228
224
were performed at near atmospheric pressure, but the partial
pressure of H2 was varied. With MoO3, the reaction atmosphere
was about 20% H and with Mo C it was about 80% H , although
Binding Energy(eV)
Binding Energy(eV)
ꢁ6#(2+*#%0&
<
%0?E
5=<4<D6
"6#(2+*#%0&
<
=#ꢁ+..3.02+#$/-4
%
0?E
:
#ꢁ+..3.02+#$/-4
5
=A4;D6
2
2
2
the combined propylene and propane formation rate was found
to be invariant with H partial pressure over Mo C. Reaction tem-
2
2
%
0=E
>949D6
%0=E
5>=4=D6
5
perature and catalyst loading were markedly different. The HDO
◦
with MoO was performed at 400 C and a contact time of 1.2
3
−
1 −1
)
gcat(gacetone h
, while the reaction with Mo C was carried out
2
0>E
:4AD6
%0>E
5;4<D6
240
%
5
◦
at a lower temperature of 96 C and longer contact times of 1.29-
2
44
240
236
232
228
224
244
236
232
228
224
−1 −1
)
Binding Energy(eV)
Binding Energy(eV)
126 gcat(gacetone
h
.
Nearly full conversion of acetone (96.8%) was achieved with
the MoO catalyst28. The major product was propylene at 67.2
Fig. 3 XPS analysis of the Mo 3d binding energies of A) fresh, B)
reduced MoO3, and used MoO3 samples after C) one cellulose injection
and D) four cellulose injections. The reduced and used samples were
3
C-mol% with hexenes (14.0 C-mol%) and butenes (7.9 C-mol%)
also being produced among other hydrocarbons. The only oxy-
◦
passivated at 30 C for 2 hr under 60 mL/min of 1.7% air in He before
removal from the reactor.
genate detected was hexanone at 0.3 C-mol%. The Mo C catalyst
2
also achieved a high conversion of acetone27. However, the hy-
drogenation product, isopropanol, was observed to be the major
product at lower catalyst contact times while at higher contact
times, propane became the major product. Propylene was de-
termined to be the intermediate product derived from the dehy-
dration of isopropanol, and was subsequently hydrogenated to
propane.
XPS measurements were performed on fresh, reduced, and
used catalyst samples to examine the distribution of Mo oxidation
states on the surface (Figure 3). The peaks of the Mo 3d dou-
blet were positioned at Mo 3d5/2 of 232.7 eV for Mo +, 231.3 for
6
Mo5+, and 229.7 for Mo4+. Fresh MoO was found to have been
3
fully oxidized to the Mo + oxidation state, as no Mo5+ or Mo4+
6
was detected. Reducing the catalyst under the same conditions as
In the present study, 2-butanone was the closest analog of ace-
tone. A low yield of 2-butanol was obtained from 2-butanone,
along with both butene and butane, more similar to the results
mentioned previously resulted in a distribution of Mo +, Mo5+
6
,
and Mo + phases on the catalyst surface, although the Mo6+ and
4
4+
Mo oxidation states were found to dominate. More specifically,
the concentrations of Mo +, Mo5+, and Mo4+ were 53.1, 2.2, and
of Mo C-catalyzed HDO of acetone. It was postulated by Sullivan
2
6
et al. that the reaction over Mo C proceeded through a sequen-
2
4
4.7%, respectively. Similar concentrations of the three Mo oxi-
tial hydrogenation of acetone to isopropanol over metallic sites,
followed by dehydration to propylene over Brønsted acid sites,
and finally hydrogenation of propylene to propane over metal-
lic sites27. Unfortunately, under the current reaction system, it
was not possible to conclude that 2-butanone followed the same
dation states were observed after one and four cellulose pulses.
Although, with reaction time and continued exposure to hydro-
gen, the concentration of Mo + was observed to decrease, which
6
was coupled with increased Mo +. No Mo2+ or Mo metal were
4
0
1
–10 | 7

Green Chemistry
Page 8 of 11
View Article Online
DOI: 10.1039/C7GC01477E
sequential reaction pathway to form 2-butanol, butene, and bu-
tane, only that the results here were consistent with that type of
reaction pathway.
rolysis sample cup, undergo cellulose chain depolymerization and
fragmentation reactions, and then vaporization of pyrolysis prod-
ucts. The slower release of volatiles from a cellulose pulse could
lead to a lower concentration of oxygenates passing over the cat-
alyst bed across a longer time span. Effectively, even though cat-
alyst loadings are similar, the space velocity or residence time of
oxygenated reactants may be different among the different feed-
stocks. This difference in residence time would likely lead to
differing rates of conversion between the C4 model compounds
and cellulose pyrolysis model compounds, despite containing the
same oxygen functionalities.
Román-Leshkov and coworkers have proposed the HDO to oc-
cur through a redox cycle reverse Mars-van Krevelen mecha-
nism28–30. In the mechanism, H interacts with surface oxygen
2
atoms to form water and an oxygen vacant site on the catalyst
surface that can bind the reactant oxygenate. After chemisorp-
tion, the oxygen could be transferred from the reactant to the
catalyst thereby forming the product alkene and an oxidized cat-
alyst. The necessity of an initial induction time to reduce the
catalyst and form oxygen vacant sites in addition to XPS mea-
surements showing the increased concentration of reduced Mo +
species, the proposed active oxidation state, lend credence to the
reverse Mars-van Krevelen mechanism. In addition, the presence
of alkene hydrogenation reactions was not reported.
Interestingly, among the model compounds only levoglucosan
was converted to aromatic products at a 4 C% yield, whereas with
a cellulose feed, the aromatic yield was ∼15 C%. In non-catalytic
cellulose pyrolysis, the levoglucosan yield is 50-60 wt%41. There-
fore, in the HDO of cellulose pyrolysis vapors, the aromatic yield
would be anticipated to be ∼2 C%. An actual yield of ∼15 C%
suggested aromatic products may not have been solely produced
from levoglucosan and that interactions between products may
have occurred. It has been shown in the literature that ethy-
lene and furanic compounds can undergo Diels-Alder reactions to
form aromatics42. A similar interaction between product alkenes
and furanic compounds may have occurred in the current system
as well, but further work will be needed to confirm. The dif-
ferences between the model compounds and cellulose pyrolysis
species demonstrates the need to use real biomass as a feedstock
when trying to capture interaction effects.
5
More complex model compounds, such as furfural, have been
investigated in the literature. Using an acid-base bifunctional cat-
alyst, for example, furfural was mainly decarbonylated to form
furan and CO40
.
However, over Mo C, deoxygenation to 2-
2
methylfuran became a dominant reaction pathway, while decar-
bonylation selecitivity to furan was <1%25. The selectivity to
2-methylfuran was 50-60%, but further deoxygenation to hydro-
carbons was not reported. Similarly, in this study the selectivity
to 2-methylfuran was 58%, while decarbonylation selectivity to
furan was much lower at 4.3%. In addition, hydrocarbons were
produced at a noticeable selectivity of 24%.
Examining the degrees of conversion of the C4 model com-
pounds, it can be seen that the alcohol groups were most readily
reacted over the catalyst, followed by carbonyl groups, and then
ether linkages. This trend could also be discerned with the three
tested furanic compounds, where the hydroxyl group in HMF was
completely reacted, the carbonyl group was partially converted,
but to a greater degree than the furan ring oxygen. Similar to this
work, Román-Leshkov and coworkers also found the furan ring
deoxygenation to be more refractory as compared with carbonyl
groups 28. Interestingly, this order of reactivity of C-OH > C=O
Haber and Lalik have reported on the isomerization of 1-butene
over MoO3 43. The isomerization was performed at a higher tem-
◦
perature (440 C) and in an inert atmosphere. Initially cis- and
trans-2-butene were the major products, although the cis/trans
ratio was not reported. With successive pulses of 1-butene, the
catalyst was reduced through the production of CO and CO re-
2
sulting in a decreased selectivity to 2-butene and acid site density.
As the catalyst was nearly completely reduced to MoO , 2-butene
2
became the major product again. In the current study, CO and
CO , butadiene, maleic anhydride, or any other product were not
2
>
C-O-C does not follow the trend in bond dissociation energies
where C=O (754 kJ/mol) > C-OH (385 kJ/mol) > C-OC (339
kJ/mol)28
detected during 1-butene conversion, in contrast with Haber and
Lalik. The absence of CO and CO may be the result of perform-
2
.
ing the reaction at a lower temperature and in a reducing atmo-
sphere rather than an inert atmosphere. Haber and Lalik have
also shown MoO3 to contain Brønsted acid sites through the iso-
merization of 3,3-dimethyl-1-butene as a probe reaction43. In an
inert atmosphere, the degree of Brønsted acidity was found to de-
crease as the catalyst was reduced from MoO3 to MoO2 by the
reaction.
Cellulose pyrolysis products are known to contain all of the oxy-
gen functionalities present in the tested model compounds. While
some of the compounds were seen to give conversions less than
100% at the reaction conditions used, cellulose pyrolysis prod-
ucts at the same conditions were totally deoxygenated. The ma-
jor cellulose pyrolysis product, levoglucosan, was also completely
deoxygenated, despite containing five oxygen moieties (two ether
linkages and three hydroxyl groups). The apparent disparity be-
tween the C4 model compounds and cellulose pyrolysis model
compounds may be a result of the inherent transience of the mi-
Performing pyridine TPD on the fresh catalyst found that
weaker acid sites were present (Figure 2). However, upon reduc-
ing the catalyst, the number and strength of acid sites was found
to greatly increase. Treating MoO with hydrogen will lead to the
3
cropyrolyzer system. Upon injection, the more volatile C model
4
formation of hydrogen molybdenum bronzes, HxMoO3 (0.21 <
x ≤ 2)39. Hydrogen bronzes of Mo and W have been hypothesized
to be the active phases in acrolein hydrogenation, 2-propanol de-
hydration, and allyl alcohol hydrodeoxygenation44–46. On the
surface, hydrogen can be present in hydroxyl groups bound to
either terminal (Mo=O) or bridging (Mo-O-Mo) lattice oxygen
compounds would be readily vaporized and swept onto the cat-
alyst bed. As a result, the pulse of reactant would lead to a
higher concentration of the model compound passing over the
catalyst bed within a shorter time span. Whereas for cellulose,
the volatilization rate may be slower due to having to heat the py-
8
|
1–10

Page 9 of 11
Green Chemistry
View Article Online
DOI: 10.1039/C7GC01477E
atoms. These surface hydroxyls will determine the degree of
and then hydrogenation to alkanes (Figure 4). With the more
complex cellulose pyrolysis compounds, the same general reac-
tion pathway was partly apparent. For HMF, the hydroxyl group
was the most readily reacted. The hydroxyl group could be de-
hydrogenated to a carbonyl forming 2,5-furandicarboxaldehyde,
or deoxygenated to a methyl group to form 5-methylfurfural.
The carbonyl oxygen in HMF or furfural was not found to be
hydrogenated to form 2,5-furandiol or furfuryl alcohol. How-
ever, it could be deoxygenated to a methyl group to form 2,5-
dimethylfuran from HMF or 2-methylfuran from HMF or fur-
fural. Therefore, in the overall reaction scheme the hydrogena-
tion/dehydrogenation interconversion between hydroxyl groups
and carbonyl groups appeared to be minor compared with the de-
oxygenation pathway. The main apparent HDO pathway involved
oxygenate deoxygenation to alkenes, isomerization of alkenes,
and hydrogenation of alkenes to alkanes (Figure 4).
Brønsted acidity of the catalyst. The removal of oxygen from
the catalyst surface through the formation of a water molecule
will produce an oxygen vacant site. These oxygen vacant sites
have been proposed to be an uncoordinated Mo atom (Mo5+)
and would act as a Lewis acid and be the active site for oxygenate
adsorption30,39
.
The injection of 1-butene into the reactor containing only sand
(no Mo ) resulted in a minor amount of isomerization and hy-
3
drogenation products to be generated, revealing the sand to not
be entirely inert as it exhibited a low level of background activity.
The hydrogenation of 1-butene to n-butane was much more ap-
parent with MoO present (Table 3). Given the reaction temper-
3
ature, the dehydrogenation of butane to butene, was not appre-
ciably catalyzed by reduced MoO . With a Mo C catalyst, a simi-
3
2
lar phenomenon was observed, whereby in the deoxygenation of
acetone to propylene, the propylene could be hydrogenated over
metallic sites to yield propane27. In that work, as the catalyst
contact time was increased, the selectivity to propane increased
to nearly 100% at the highest tested contact time. Previously,
5 Conclusion
The hydrodeoxygenation of cellulose and cellulose pyrolysis
model compounds was performed over a molybdenum oxide cat-
alyst and low pressure hydrogen. From monofunctional C4 com-
pounds, the observed reactions included hydrogenation and de-
hydrogenation of carbonyl and hydroxyl groups, respectively, as
well as deoxygenation to hydrocarbons. The cellulose pyroly-
sis model compounds, particularly the furanic and low molecu-
lar weight model compounds, underwent the same general re-
actions during HDO. For example, HMF was converted into 2,5-
furandicarboxaldehyde, 5-methylfurfural, 2,5-dimethylfuran, and
aliphatic hydrocarbons. Levoglucosan, on the other hand, did not
produce any partially deoxygenated oxygenate products and was
found to be totally deoxygenated to aliphatic and aromatic hy-
drocarbons. The general reactivity towards deoxygenation of the
oxygen functional groups was found to be C-OH > C=O > C-O-C.
our group has found that high loadings of MoO also produced
3
alkanes at the expense of alkenes34. For the adsorbed alkenes,
there may be a competition between desorption and hydrogena-
tion to alkanes. It can be seen that butene could adsorb and be
hydrogenated, while butane was not readily dehydrogenated to
butene (Table 3). Therefore, using lower catalyst loadings fa-
cilitated higher alkene yields, while higher catalyst loadings al-
lowed the alkenes to be hydrogenated resulting in higher alkane
yields. This result was in contrast with earlier published work
that reported only alkenes as products from the HDO of ketone
and furanic model compounds over MoO 28. It is possible that a
3
portion of the disparity may be related to the GC analytical meth-
ods. In this study, a GasPro separation column was used, which
could effectively separate alkenes and alkanes of the same carbon
number and their isomers. In the previous study, a non-polar DB-
The HDO of C model compounds generally yielded a mix of
4
butene isomers and butane. Injecting 1-butene over the catalyst
produced the same butene isomers as well as butane, showing
alkene isomerization and hydrogenation to be catalyzed by re-
5
column was used, which would be less effective in separating
light gases. Therefore, if propane had been formed in acetone
HDO, it may not have been effectively separated from propylene
and identified with the DB-5 column.
duced MoO . While the inert sand used in the catalyst beds was
3
found to catalyze the isomerization and hydrogenation, it was to
a much lesser degree.
O
When applied to the HDO of cellulose pyrolysis vapors, a trade-
off existed between achieving high levels of deoxygenation and
high selectivities towards unsaturated hydrocarbons. At higher
catalyst loadings, total deoxygenation can be reached, however
the hydrocarbon products were nearly all saturated. Conversely
at lower catalyst loadings, a higher selectivity to alkenes was ob-
tained, but the product stream contained oxygenated molecules.
Therefore, further catalyst development and process development
are required wherein the pyrolysis vapor feed stream can be to-
tally deoxygenated, but the final reaction hydrogenation step to
alkanes is minimized or prevented. In addition,
R'
R'
R
R
R'
R
Alkane
OH
R
R'
R
R'
Pyrolysis
Oxygenate
Alkene
Fig. 4 Proposed Reaction Pathway for HDO over MoO3
Among the model compounds, only levoglucosan was con-
verted to aromatic products, and at a much lower yield than from
a cellulose feed (4 C% vs. ∼15 C%). These results could conveiv-
ably indicate the presence of interactions between compounds,
such as Diels-Alder reactions between alkenes and furanics. This
The simpler C model compounds were found to undergo hy-
4
drogenation/dehydrogenation of carbonyl and alcohol functional
groups, deoxygenation to alkenes and subsequent isomerization,
1
–10 | 9

Green Chemistry
Page 10 of 11
View Article Online
DOI: 10.1039/C7GC01477E
stresses the need to use real biomass as a feed in catalyst devel-
opment to capture any interaction effects that may not be present
when only model compounds are used.
Energy & Fuels, 2004, 18, 1640–1647.
23 D. A. Ruddy, J. A. Schaidle, J. R. Ferrell III, J. Wang, L. Moens
and J. E. Hensley, Green Chemistry, 2014, 16, 454–490.
2
2
2
2
2
2
4 C.-J. Chen, W.-S. Lee and A. Bhan, Applied Catalysis A: Gen-
Acknowledgements
eral, 2016, 510, 42–48.
This work was funded by ExxonMobil. The authors would like to
thank Dapeng Jing for performing the XPS measurements.
5 W.-S. Lee, Z. Wang, W. Zheng, D. G. Vlachos and A. Bhan,
Catalysis Science & Technology, 2014, 4, 2340–2352.
6 W.-S. Lee, Z. Wang, R. J. Wu and A. Bhan, Journal of Catalysis,
References
2
014, 319, 44–53.
7 M. M. Sullivan and A. Bhan, ACS Catalysis, 2016, 6, 1145–
152.
1
2
3
4
5
6
7
M. Balat and M. Balat, International Journal of Hydrogen En-
ergy, 2009, 34, 3589–3603.
1
A. B. Chhetri and M. R. Islam, Energy Sources, Part A: Recovery,
Utilization, and Environmental Effects, 2008, 30, 754–764.
H. L. Chum and R. P. Overend, Fuel Processing Technology,
8 T. Prasomsri, T. Nimmanwudipong and Y. Roman-Leshkov,
Energy & Environmental Science, 2013, 6, 1732–1738.
9 T. Prasomsri, M. Shetty, K. Murugappan and Y. Roman-
Leshkov, Energy & Environmental Science, 2014, 7, 2660–
2
001, 71, 187–195.
J. H. Clark, Journal of Chemical Technology & Biotechnology,
007, 82, 603–609.
2
669.
2
3
3
3
3
3
3
3
3
0 M. Shetty, K. Murugappan, T. Prasomsri, W. H. Green and
Y. Román-Leshkov, Journal of Catalysis, 2015, 331, 86–97.
1 M. Grilc, B. Likozar and J. Levec, Applied Catalysis B: Environ-
mental, 2014, 150, 275–287.
C. R. Coronado, J. A. de Carvalho Jr. and J. L. Silveira, Fuel
Processing Technology, 2009, 90, 204–211.
A. Demirbas, Progress In Energy and Combustion Science,
2
007, 33, 1–18.
S. Czernik and A. V. Bridgwater, Energy & Fuels, 2004, 18,
90–598.
2 M. Grilc, G. Veryasov, B. Likozar, A. Jesih and J. Levec, Applied
Catalysis B: Environmental, 2015, 163, 467–477.
5
3 G. Veryasov, M. Grilc, B. Likozar and A. Jesih, Catalysis Com-
munications, 2014, 46, 183–186.
8
9
A. V. Bridgwater, Biomass & Bioenergy, 2012, 38, 68–94.
D. Mohan, C. U. Pittman and P. H. Steele, Energy & Fuels,
4 M. W. Nolte, J. Zhang and B. H. Shanks, Green Chemistry,
2
006, 20, 848–889.
0 J. P. Diebold and S. Czernik, Energy & Fuels, 1997, 11, 1081–
091.
1 S. Czernik, D. K. Johnson and S. Black, Biomass & Bioenergy,
994, 7, 187–192.
2 M. W. Nolte and M. W. Liberatore, Energy & Fuels, 2011, 25,
314–3317.
3 J. Zhang, Y. S. Choi and B. H. Shanks, ChemSusChem, 2015,
, 4256–5631.
4 M. W. Nolte and B. H. Shanks, Energy Technology, 2016, 5,
–18.
2
016, 18, 134–138.
5 S. Czernik and A. V. Bridgwater, Energy & Fuels, 2004, 18,
90–598.
1
1
1
1
1
1
5
6 J. Piskorz, D. S. Scott and D. Radlein, Pyrolysis Oils from
Biomass, 1988, ch. 17, pp. 167–178.
1
7 P. R. Patwardhan, J. A. Satrio, R. C. Brown and B. H. Shanks,
Journal of Analytical and Applied Pyrolysis, 2009, 86, 323–
3
3
30.
8
3
3
4
4
4
8 H. Hattori and Y. Ono, Solid Acid Catalysis: From Fundamen-
tals to Applications, Pan Stanford Publishing, 2015.
7
9 D. R. Moberg, T. J. Thibodeau, F. G. Amar and B. G. Frederick,
Journal of Physical Chemistry C, 2010, 114, 13782–13795.
0 J. Zhang, K. Wang, M. W. Nolte, Y. S. Choi, R. C. Brown and
B. H. Shanks, ACS Catalysis, 2016, 6, 2608–2621.
1
5 D. C. Elliot, Energy & Fuels, 2007, 21, 1792–1815.
6 A. H. Zacher, M. V. Olarte, D. M. Santosa, D. C. Elliot and S. B.
Jones, Green Chemistry, 2014, 16, 491–515.
1
1
1
1
7 K. Wang, K. H. Kim and R. C. Brown, Green Chemistry, 2014,
1 J. Zhang, M. W. Nolte and B. H. Shanks, ACS Sustainable
Chemistry & Engineering, 2014, 2, 2820–2830.
1
6, 727–735.
8 T. R. Carlson, G. A. Tompsett, W. C. Conner and G. W. Huber,
Topics in Catalysis, 2009, 52, 241–252.
2 C. L. Williams, C.-C. Chang, P. Do, N. Nikbin, S. Caratzoulas,
D. G. Vlachos, R. F. Lobo, W. Fan and P. J. Dauenhauer, ACS
Catalysis, 2012, 2, 935–939.
9 J. Jae, G. A. Tompsett, A. J. Foster, K. D. Hammond, S. A.
Auerbach, R. F. Lobo and G. W. Huber, Journal of Catalysis,
4
3 J. Haber and E. Lalik, Catalysis Today, 1997, 33, 119–137.
4 C. Hoang-Van and O. Zegaoui, Applied Catalysis A: General,
2011, 279, 257–268.
4
2
2
2
0 A. T. Aguayo, A. G. Gayubo, A. Atutxa, M. Olazar and J. Bil-
1
995, 130, 89–103.
bao, Industrial & Engineering Chemistry Research, 2002, 41,
4
4
5 T. Matsuda, Y. Hirata, S. Suga, H. Sakagami and N. Takahashi,
Applied Catalysis A: General, 2000, 193, 185–193.
4216–4224.
1 A. G. Gayubo, A. T. Aguayo, A. Atutxa, R. Prieto and J. Bilbao,
6 T. J. Thibodeau, A. S. Canney, W. J. De Sisto, M. C. Wheeler,
F. G. Amar and B. G. Frederick, Applied Catalysis A: General,
Industrial & Engineering Chemistry Research, 2004, 43, 5042–
5048.
2
010, 388, 86–95.
2 A. G. Gayubo, A. T. Aguayo, A. Atutxa, R. Prieto and J. Bilbao,
1
0 |
1–10

Page 11 of 11
Green Chemistry
View Article Online
DOI: 10.1039/C7GC01477E
Low pressure hydrodeoxygenation of cellulose pyrolysis model compounds over MoO3 to form
hydrocarbons