(496g) Reaction Pathways and Microkinetic Modeling of Levulinic Acid Hydrodeoxygenation over Sulfided NiMo/Al2O3

Miha Grilc and Blaž Likozar, Laboratory
of Catalysis and Chemical Reaction Engineering, National Institute of
Chemistry, Ljubljana, Slovenia

Abstract text:

Interest in the
bio-chemicals and bio-fuels production from renewable resources has increased
in the last decade, mainly as a result of the limited availability of petroleum
reserves, uncertain prices of crude oil derivatives and environmental
concerns. There is an ongoing challenge to develop economically-efficient
and environmentally-friendly technologies to transform lignocellulosic
biomass to fuels (BTF) and chemicals.

Levulinic acid
is a representative of important biomass-derived building blocks for added-value
chemicals production and can be produced from cellulose or hemicellulose.
Production from cellulose is conducted by an acidic hydrolysis or solvolysis
into 5-hydroxymethylfurfural (HMF), that is further derfomylated
to levulinic acid. On the other hand, it can also be produced from
hemicellulose, although the pathway is much more challenging and includes its
acidic hydrolysis to pentose units, subsequent dehydration to furfural,
hydrogenation to furfuryl alcohol and its hydrolysis
to levulinic acid. Levulinic acid is therefore the main product of
lignocellulosic biomass liquefaction by acidic solvolysis or hydrolysis, with
up to 83 mol% yields reported if cellulose is used as
a feedstock. Levulinic acid production from biomass shows higher sustainability
as well as economic feasibility in comparison to the traditionally used
synthesis method involving the petrochemical maleic anhydride conversion. As a
member of gamma-keto acids it is recognized as an ideal platform chemical to
produce various targeted added-value bio-chemicals or simply a liquid fuel.

A promising
method to convert levulinic acid into added-value chemicals is a catalytic
hydrodeoxygenation. Hydrodeoxygenation takes place at temperatures above 200 °C
in the presence of gaseous hydrogen (or other hydrogen source) and a
heterogeneous hydrodeoxygenation catalysts either based on transition metals
(Mo, W, Ni, Co) in their sulfide, metal, oxide nitride or carbide form or noble
metals (Pt, Ru, Pd) on different supports. Process
parameters required for hydrodeoxygenation can also induce competitive (and
often undesired) decarbonylation or decarboxylation
reactions that can significantly decrease the selectivity of desired catalytic transformation.
For that reason the optimization of process parameters and appropriate catalyst
design or selection is crucial to maximize the rate of desired transformation
and still preserving the permissible selectivity.

Commercially
available NiMo/γ-Al₂O₃ catalyst,
developed for hydrodesulphurization (HDS) of crude oil, was used in this work.
Prior to testing, the catalyst (received in its oxide form) was sulphided using DMDS under 2.5 MPa hydrogen pressure at
temperature of 350 °C for 60 minutes in a stainless steel autoclave. Morphology
of the catalyst was characterized by scanning electron microscopy (SEM),
specific surface area was determined by Brunauer–Emmett–Teller
method (179 m² g^–1), while the concentration of
accessible NiMoO₄/Al₂O₃ sites was determined
according to the hydrogen consumption during the second TPR–TPO–TPR step which
served as an estimation of active sites (200 ± 40 μmol
g^–1) after the sulfidation. Additional
characterization techniques, namely TPR, TPD_NH3, DRIFT, XRD and EDX
were performed as well for additional insights into the characteristics of the
catalyst.

Hydrodeoxygenation
was investigated in the same 300 mL autoclave equipped as used for the
activation, equipped with a magnetically driven turbine stirrer and electric
heating jacket. Hydrodeoxygenation took place at solventless
conditions, therefore only solid catalyst, liquid levulinic acid and gaseous
hydrogen were in contact in the initial stages of the experiment. Experiment
was performed under a constant hydrogen purge (flow rate of 1.0 L_N
min^–1) that allowed continuous gas phase analysis, high availability
of H₂ and constant pressure throughout the experiment.

The influence of
temperature (the lowest value 225 °C, the highest value 300 °C), pressure (the
lowest value 2.5 MPa, the highest value 7.5 MPa), stirring speed (the lowest
value 200 rpm, the highest value 1400 rpm), catalyst particle size (the
smallest fraction < 40 μm, the largest
fraction: pellet), mass of catalyst (the lowest mass 0 wt%,
the highest value 4 wt%) and gas type (hydrogen or
nitrogen) was tested in the process-conditions screening tests.

The analysis of
gaseous (GC, FTIR) and liquid phase (GC–MS/FID) reveals the presence of two
deoxygenation mechanisms: hydrodeoxygenation (by a combination of dehydration
and hydrogenation in various order) and decarboxylation as shown in Figure 1.
Hydrodeoxygenation reaction only took place in the coexistent presence of a catalyst
and elevated hydrogen pressure. In case that nitrogen gas was introduced
instead of hydrogen, the hydrodeoxygenation reaction was negligible at all
temperatures, while the increase of the latter resulted in substantial
decarboxylation. However, the experiment in absence of hydrogen showed that
catalytic dehydration of levulinic acid to angelica lactone is relatively slow,
which reveals that in presence of a catalyst and hydrogen γ-valerolactone is obviously not being formed via angelica
lactone route (kc3, kc4), but rather by hydrogenation to unstable hydroxyvaleric acid intermediate and subsequent dehydration
(kc1, kc2).

Figure 1: Proposed reaction pathway network for
levulinic acid hydrotreatment over NiMo/Al₂O₃.

Experiment with
hydrogen and without a catalyst showed similar results, however; decarboxylation
proceeded in two different ways, via direct decarboxylation of levulinic acid
to ketone or by decarboxylation with subsequent C–C coupling with angelica
lactone. It is important to mention that oxobutyl
derivate of γ-valerolactone is the main product
obtained during the experiment without a catalyst, although its formation rate
is low. Figure 2 shows the results for the experiment at 275 °C under 5 MPa of
hydrogen pressure and by using sulphided NiMo/Al₂O₃ in pelletized form.
Competition between decarboxylation and catalytic hydrodeoxygenation is clearly
seen. Main decarboxylation products in the liquid phase (Butanone and Butanol)
were detected, but could not be representatively quantified, due to their high
volatility. Hydrodeoxygenation reaction rate is obviously governed by
hydrogenation of levulinic acid to hydroxyvaleric acid, followed by rapid dehydration to γ-valerolactone,
since the concentration of the intermediate in the liquid phase was below
detection limit. Involvement of hydroxylpentanoic
acid (although not identified in the liquid phase) in the reaction mechanism
can be confirmed with valeric (pentanoic)
acid formation. Its low (but constant) reaction rate shows that the dehydration
of hydroxypentanoic acid is much faster than
hydrodeoxygenation. The main product of catalytic hydrotreatment
within the range of tested reaction conditions is therefore γ-valerolactone, since the products of further
reactions were only detected in traces (kc5 extremely low).

Figure 2: Experimental (points) and calculated
(lines) results of levulinic hydrotreatment (T = 275 °C, P = 5 MPa H₂, N
= 1000 rpm, NiMo/Al₂O₃
pellets) for liquid (a) and gas (b) phase.

Kinetic model
was developed according to the reaction pathway network. With the regression
analysis (by taking all experimental conditions and results into account) the
influence of the following phenomena was quantitatively addressed: a)
hydrodynamic conditions in the reactor, b) transport phenomena, c) homogeneous
(non-catalytic) chemical reaction, d) adsorption rate of components from
solid–liquid interphase to the active sites on the catalyst surface and
desorption in the opposite direction, e) the rate of chemical reactions on the
active sites of the catalyst. Modelling results allowed evaluating the
contribution of above-mentioned phenomena to overall rate of components’
transformations in gaseous and liquid phase. With evaluation of parameters
obtained by the modelling, the bottlenecks and rate-limiting steps of the
process were identified. Detailed information about the modeling methodology
and results will be available at the conference.

Acknowledgment: The authors gratefully
acknowledge the financial support of the Slovenian Research Agency (ARRS)
through the programme P2-0152.