Kinetics and reaction engineering of levulinic acid production from aqueous glucose solutions

Article abstract of DOI:10.1002/cssc.201100717

We have developed a kinetic model for aqueous-phase production of levulinic acid from glucose using a homogeneous acid catalyst. The proposed model shows a good fit with experimental data collected in this study in a batch reactor. The model was also fitted to steady-state data obtained in a plug flow reactor (PFR) and a continuously stirred tank reactor (CSTR). The kinetic model consists of four key steps: (1) glucose dehydration to form 5-hydroxymethylfurfural (HMF); (2) glucose reversion/degradation reactions to produce humins (highly polymerized insoluble carbonaceous species); (3) HMF rehydration to form levulinic acid and formic acid; and (4) HMF degradation to form humins. We use our model to predict the optimal reactor design and operating conditions for HMF and levulinic acid production in a continuous reactor system. Higher temperatures (180-200 °C) and shorter reaction times (less than 1 min) are essential to maximize the HMF content. In contrast, relatively low temperatures (140-160 °C) and longer residence times (above 100 min) are essential for maximum levulinic acid yield. We estimate that a maximum HMF carbon yield of 14 % can be obtained in a PFR at 200 °C and a reaction time of 10 s. Levulinic acid can be produced at 57 % carbon yield (68 % of the theoretical yield) in a PFR at 149 °C and a residence time of 500 min. A system of two consecutive PFR reactors shows a higher performance than a PFR and CSTR combination. However, compared to a single PFR, there is no distinct advantage to implement a system of two consecutive reactors. Copyright

Full text of DOI:10.1002/cssc.201100717

DOI: 10.1002/cssc.201100717
Kinetics and Reaction Engineering of Levulinic Acid
Production from Aqueous Glucose Solutions
Ronen Weingarten, Joungmo Cho, Rong Xing, William Curtis Conner, Jr, and
[a]
George W. Huber*
We have developed a kinetic model for aqueous-phase pro-
duction of levulinic acid from glucose using a homogeneous
acid catalyst. The proposed model shows a good fit with ex-
perimental data collected in this study in a batch reactor. The
model was also fitted to steady-state data obtained in a plug
flow reactor (PFR) and a continuously stirred tank reactor
levulinic acid production in a continuous reactor system.
Higher temperatures (180–2008C) and shorter reaction times
(less than 1 min) are essential to maximize the HMF content. In
contrast, relatively low temperatures (140–1608C) and longer
residence times (above 100 min) are essential for maximum
levulinic acid yield. We estimate that a maximum HMF carbon
yield of 14% can be obtained in a PFR at 2008C and a reaction
time of 10 s. Levulinic acid can be produced at 57% carbon
yield (68% of the theoretical yield) in a PFR at 1498C and a resi-
dence time of 500 min. A system of two consecutive PFR reac-
tors shows a higher performance than a PFR and CSTR combi-
nation. However, compared to a single PFR, there is no distinct
advantage to implement a system of two consecutive reactors.
(CSTR). The kinetic model consists of four key steps: (1) glucose
dehydration to form 5-hydroxymethylfurfural (HMF); (2) glu-
cose reversion/degradation reactions to produce humins
(highly polymerized insoluble carbonaceous species); (3) HMF
rehydration to form levulinic acid and formic acid; and (4) HMF
degradation to form humins. We use our model to predict the
optimal reactor design and operating conditions for HMF and
Introduction
Levulinic acid (LA) is a versatile building block that for decades
has been considered a basic raw chemical material owing to
by decarboxylation of GVL. The products are then fed to an oli-
gomerization reactor where butene monomers are coupled to
form condensable alkenes. A comprehensive review of levulinic
[
1]
its high chemical reactivity. This unique feature is attributed
to its two highly reactive keto and carboxyl groups. This re-
newable biochemical can be used as a platform for the pro-
duction of various high-volume organic chemicals with numer-
[12]
acid applications is given by Alonso et al.
The formation of levulinic acid from carbohydrates consists
of a series of consecutive reactions, which includes a hexose
triple dehydration step to produce 5-hydroxymethylfurfural
(HMF) and the rehydration of HMF with two molecules of
water to produce levulinic acid and formic acid. Furfuryl alco-
hol, a product of hemicellulose depolymerization and hydroge-
nation, can also serve as an alternative source of levulinic
[
2]
ous potential industrial applications. For example, levulinic
acid can serve as a feedstock for the production of transporta-
tion fuels (gasoline and diesel). Esterification of levulinic acid
with C –C2 alcohols produces levulinic esters, which can be
1
[
3]
used as diesel additives. Elliott and Frye have shown that lev-
ulinic acid can be hydrogenated in the presence of a bifunc-
tional catalyst to produce methyl tetrahydrofuran (MTHF) in
[13]
acid. Levulinic acid production of above 80% can be ach-
ieved from conversion of aqueous solutions of furfuryl alcohol
[
4]
[14]
one step in relatively high yields. MTHF can directly serve as
with hydrochloric acid. Extensive studies have been reported
[
5]
a gasoline blend-stock and the US Department of Energy has
on the conversion of biomass feedstock to levulinic acid using
homogeneous catalysts, including mineral acids and metal
[6]
approved MTHF as a component of P-series-type fuels.
[
15–19]
Levulinic acid can also be converted into g-valerolactone
chlorides.
An overview of levulinic acid synthesis using
[20]
(
GVL) via hydrogenation with molecular hydrogen or formic
various feedstocks and acid catalysts is given by Girisuta, as
[
7,8]
[21]
acid (FA).
GVL has been shown to be a sustainable liquid
well as by Rackemann and Doherty.
transportation fuel suitable of replacing ethanol in gasoline–
[
9]
ethanol blends. Lange et al. have shown that continued hy-
drogenation of GVL produces valeric acid, which can be esteri-
fied with alcohols to produce a new class of cellulosic trans-
[
a] R. Weingarten, Dr. J. Cho, Dr. R. Xing, Prof. W. C. Conner, Jr,
Prof. G. W. Huber
Department of Chemical Engineering
University of Massachusetts Amherst
[
10]
portation fuels—valeric biofuels.
Blends of these valeric
6
86 North Pleasant St. 159 Goessmann Lab
esters with gasoline have shown promising results in engine
testing. Dumesic and co-workers have recently developed an
integrated catalytic process to convert GVL to liquid alkenes
Amherst, MA 01003 (USA)
Fax: (+1)413-545-1647
E-mail: huber@ecs.umass.edu
(
ranging from C to C ), which could be blended with gasoline
8 24
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cssc.201100717.
[
11]
or jet fuels. Butene and carbon dioxide are initially produced
ChemSusChem 0000, 00, 1 – 12
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
1
&
ÞÞ
These are not the final page numbers!

G. W. Huber et al.
One commercial process for
the production of levulinic acid
was developed by Biofine Incor-
porated (presently Biofine Re-
newables). The Biofine process
claims to produce levulinic acid
at yields higher than 70% of the
theoretical (58% carbon yield),
based on the hexose content of
the cellulosic feedstock, in
[
22,23]
a two-reactor system.
carbohydrate-containing
The
feed-
stock is initially hydrolyzed in
the first reactor, a plug flow re-
actor (PFR), at 210–2308C for
1
5
3–25 s in the presence of 1– Scheme 1. The Biofine process. Adapted from Ref. [2].
wt% sulfuric acid. The product,
HMF, is then continuously re-
moved and supplied to a second
reactor, a continuously stirred tank reactor (CSTR), where it is
further hydrolyzed at 195–2158C for 15–30 min to produce lev-
ulinic acid. A schematic of the Biofine process is shown in
Scheme 1.
Vast interest in levulinic acid applications has led to numer-
ous kinetic studies on the decomposition of carbohydrates to
produce HMF and levulinic acid. Some studies have also incor-
porated a hydrolysis step into their models to produce glucose
from cellulose or woody biomass. The formation of undesired
highly polymerized carbonaceous species (for example,
humins) has been reported in the literature since the early
[
24–26]
stages of this research.
It has also been postulated that
discoloration of sugar solutions is attributed to the polymeri-
zation of HMF to yield colored products of varying degrees of
[
27]
solubility. Nonetheless, early kinetic studies only obtained ki-
netic parameters for the dehydration and rehydration steps
leading to levulinic acid. More recent kinetic studies have in-
corporated undesired byproduct formation steps to enhance
[
28–33]
the accuracy of their models.
The objective of this study is to develop a mechanistically
based kinetic model for the conversion of glucose to levulinic
acid in aqueous media with hydrochloric acid by fitting kinetic
data collected in a batch reactor, a PFR, and a CSTR. The kinetic
model will then be used to estimate the optimal reactor
design and operating conditions in a continuous reactor
system to maximize HMF and levulinic acid yields with a homo-
geneous acid catalyst.
Figure 1. Aqueous-phase acid-catalyzed HMF rehydration in a stirred batch
reactor. Effect of initial HMF concentration on (a) HMF conversion; (b) LA
carbon yield; and (c) FA carbon yield at 1308C and 0.1m HCl. [HMF]
wt%)=4(&), 8(~), 16(N).
0
(
Results
Kinetic model for HMF rehydration
comparable HMF conversions and levulinic acid yields were
obtained for the range of initial HMF concentrations.
HMF rehydration consists of two irreversible parallel reac-
tions as shown in Equations (1) and (2).
Aqueous-phase HMF rehydration experiments were carried out
in a batch reactor at 120–1508C in acidic media (0.1m HCl)
with an initial concentration of 4 wt% HMF. Similarly, additional
experiments were carried out at 1308C with initial HMF con-
centrations ranging between 4–16 wt% to study the effect of
both feedstock and water concentration. Figure 1 shows that
k3
HMF þ 2 H Oꢀ!LA þ FA
ð1Þ
2
&2
&
www.chemsuschem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 12
ÝÝ
These are not the final page numbers!

Levulinic Acid Production
k4
HMFꢀ!D
ð2Þ
where D is the decomposition products (humins). The first re-
action [Eq. (1)] is the rehydration of HMF with two molecules
of water to produce levulinic acid and formic acid. The second
reaction [Eq. (2)] is the degradation of HMF to produce
humins. These proposed reactions are consistent with the
mechanism proposed by Horvat et al. for levulinic acid forma-
[
34,35]
tion from HMF.
The authors claimed that the addition of
water to the 2,3-carbon positions on HMF resulted in unde-
sired polymerization reactions, whereas the addition of water
to the 4,5-carbon positions gave way to levulinic-acid forma-
tion via decarboxylation to produce formic acid. Both reactions
fit equations that are pseudo first order with respect to HMF.
This is in agreement with previous kinetic studies on HMF de-
[
26,36]
composition.
Furthermore, within our range of concentra-
tions, water is considered to be in excess (zero order) in the re-
hydration step (see Figure 1). Some kinetic studies have also
proposed an additional reaction pathway to produce humins
Figure 2. Aqueous-phase acid-catalyzed HMF rehydration to levulinic acid in
a stirred batch reactor. Kinetic model fit for (a) HMF rehydration and (b) levu-
linic acid formation for 4 wt% HMF and 0.1m HCl. T (8C)=120(&), 130(*),
[
32,33]
from levulinic acid.
equimolar concentrations of levulinic acid and formic acid at
808C with 0.1m HCl concluded that levulinic acid did not
degrade after 120 min. This is in agreement with other
However, our separate studies with
140(~), 150(^); Model prediction (c).
1
[
26,29]
studies.
formation of furfural from HMF proceeds via loss of formalde-
[38–41]
hyde.
dration and levulinic-acid production with the fitted model at
.1m HCl.
Figure 2 shows the experimental data for HMF rehy-
Table 1. Estimated kinetic parameters for aqueous-phase HMF con-
[
a]
version.
0
[
b]
ꢀ1
ꢀ1
Rate constants
log10 [A/min
]
E [kJmol ]
[
c]
k
k
3
10.31ꢁ0.71
15.69ꢁ3.22
94.72ꢁ5.54
4
141.94ꢁ25.72
Kinetic model for glucose dehydration
[
4
a] Kinetic parameters fit experimental data at: T=120–1508C; [HMF]
–16 wt%; [H ]=0.1m. [b] 1 order rate parameters are lumped with
0
=
+
st
Glucose dehydration experiments were carried out in a batch
reactor at 140–1808C in 0.1m HCl with an initial glucose con-
centration of 10 wt%. The overall reaction scheme for the
aqueous-phase levulinic acid production from glucose is
shown in Scheme 2.
acid concentration. [c] 95% confidence interval in parameter estimation.
All experimental data for HMF rehydration were fitted to the
proposed kinetic model to estimate the rate parameters. Our
model assumed a first order dependence with respect to the
acid concentration, and the activation energies were deter-
mined for experiments carried out at a constant acid concen-
tration of 0.1m HCl. The best correlated values with their stan-
dard errors appear in Table 1. The estimated values of the acti-
vation energies were assumed to be independent of the acid
concentration and were used for the remainder of the model
fitting. On this note, it has been
Glucose undergoes two irreversible parallel reactions, which
consist of a triple dehydration step to produce HMF (denoted
as Reaction 1), or a degradation reaction to form humins (de-
noted as Reaction 2). Both reactions have been reported to be
[30]
pseudo first order with respect to glucose.
This was also
confirmed with reactions that were carried out at various initial
glucose concentrations ranging from 2–20 wt% at 1608C in
0.1m HCl. Glucose conversions, HMF yields, and levulinic acid
reported that the activation
energy can be a function of acid
concentration. However, this
claim is valid when a reaction is
governed by
a slow proton
[
37]
transfer step. It is also notable
to mention that small amounts
of 2-furaldehyde (furfural) were
detected as a byproduct of this
reaction at less than 0.04% yield.
Scheme 2. Overall reaction scheme for the aqueous-phase acid-catalyzed production of levulinic acid from glu-
It has been reported that the cose.
ChemSusChem 0000, 00, 1 – 12
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
3
&
ÞÞ
These are not the final page numbers!

G. W. Huber et al.
Figure 3. Aqueous-phase acid-catalyzed glucose dehydration in a stirred
batch reactor. Effect of initial glucose concentration on (a) glucose conver-
sion; (b) HMF carbon yield; and (c) LA carbon yield at 1608C and 0.1m HCl.
Figure 4. Aqueous-phase acid-catalyzed glucose dehydration in a stirred
batch reactor. Kinetic model fit for (a) glucose dehydration; (b) HMF forma-
tion; and (c) levulinic acid formation for 10 wt% glucose and 0.1m HCl. T
0
[G] (wt%)=2(&), 10(~), 20(N).
(
8C)=140(&), 150(*), 160(~), 170(^), 180(N); Model prediction (c).
yields were found to be predominantly independent of the ini-
tial glucose concentration, as illustrated in Figure 3.
The overall rate equations for glucose conversion are shown
in Equations (3)–(5).
The rate parameters obtained for glucose dehydration
appear in Table 2. Reactions 3 and 4 (HMF rehydration and de-
composition, respectively) were assumed independent of Reac-
tions 1 and 2 (glucose dehydration and glucose decomposi-
d½Gꢂ
¼
ꢀðk þ k Þ½Gꢂ
ð3Þ
ð4Þ
ð5Þ
1
2
dt
d½HMFꢂ
dt
¼
k ½Gꢂ ꢀ ðk þ k Þ½HMFꢂ
1
3
4
d½LAꢂ
Table 2. Estimated kinetic parameters for aqueous-phase glucose con-
¼ k ½HMFꢂ
3
dt
[
a]
version.
[
b]
ꢀ1
ꢀ1
As with the HMF rehydration study, the activation energies
for glucose dehydration (Reactions 1 and 2) were determined
for experiments carried out at a constant acid concentration of
Rate constants
log10 [A/min
]
E [kJmol ]
[
c]
k
k
1
17.12ꢁ0.62
3.33ꢁ0.29
160.16ꢁ5.15
50.68ꢁ2.38
2
0.1m HCl. Likewise, it was initially assumed that a first order
[
2
a] Kinetic parameters fit experimental data at: T=140–1808C; [G]
0 wt%; [H ]=0.1m. [b] 1 order rate parameters are lumped with acid
concentration. [c] 95% confidence interval in parameter estimation.
0
=2–
dependence exists with respect to the acid concentration.
Glucose can also undergo reversion and epimerization reac-
tions to produce oligosaccharides and anhydrosugars and fruc-
+
st
[25,30,42]
tose, respectively.
The reversion products are mainly dis-
tion, respectively). Therefore, the same rate parameters ob-
tained for HMF rehydration (refer to Table 1) were combined
with those derived for glucose dehydration to fit the experi-
mental data to our proposed model for levulinic acid produc-
tion from glucose. The experimental data for glucose dehydra-
tion, HMF, and levulinic acid production are shown in Figure 4
along with the fitted kinetic model.
accharides, which are formed by way of a coupling reaction of
the anomeric hydroxyl group of one glucose molecule with
[43]
any hydroxyl group of a second molecule. It has also been
suggested that the products from reversion reactions can be
subject to degradation reactions, in which the disaccharides
[44]
react further to form oligosaccharides.
Consequently, the
presence of cellobiose confirmed the occurrence of reversion
&4
&
www.chemsuschem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 12
ÝÝ
These are not the final page numbers!

Levulinic Acid Production
reactions in this study. Levoglucosan (1,6-anhydro-b-d-gluco-
pyranose) and fructose were also detected in the reaction sam-
ples in qualitative amounts. Fructose was present at less than
0.1% yield. Previous studies have shown that the formation of
fructose from glucose in acidic solution proceeds by way of
[
45]
a C -to-C intramolecular hydrogen transfer. The role of the
2
1
acid catalyst is to protonate the carbonyl oxygen atom to facil-
[
46]
itate a hydride shift mechanism. Similar conclusions were de-
duced by Davis and co-workers in their study of aqueous-
phase isomerization of glucose to fructose with a solid Lewis
[
47–49]
acid catalyst.
Hence, one theory suggests that the formation of HMF from
[
43]
glucose proceeds via fructose and that the near-nil presence
of fructose can be attributed to its high reactivity compared to
[
45,50]
glucose.
Conversely, other authors claim that glucose can
Figure 5. Aqueous-phase formic acid production from glucose in a stirred
batch reactor. Comparison between experimental and kinetic model data.
Feed was 10 wt% glucose and 0.1m HCl. T (8C)=140(&), 150(*), 160(~),
170(^), 180(N); Model prediction (c).
be converted directly to HMF through cyclization of a 3-deoxy-
glucosone intermediate formed from the open-ring form of
[
51,52]
glucose.
In this respect, the relatively low conversion of
glucose to HMF is caused by its low affinity to exist in the
open-ring form due to stabilization of the glucose pyranose
S2). As mentioned previously, activation energies were calculat-
ed for reactions at 0.1m HCl and assumed to be constant for
all other acid concentrations. The pre-exponential factors were
calculated for each reaction step x (refer to Scheme 2) at each
acid concentration, and a power law function was derived, as
shown in Equation (6). This approach was similar to those used
[
53]
forms in aqueous solution. Overall, there are two schools of
thought with regard to the mechanism of HMF formation from
C carbohydrates. One theory postulates that the reaction pro-
6
[45,54]
ceeds by way of the acyclic 1,2-enediol intermediate.
The
other takes into account a fructofuranosyl cyclic intermediate
[
51,55]
[24]
[50]
in the formation of HMF from fructose.
Recent computa-
by Saeman and Kuster and Temmink. The best kinetics fit
parameters with their standard errors are tabulated in Table 3.
tional studies have also reported the use of fructofuranosyl in-
termediates in the formation of levulinic acid and HMF from
ꢁ
ꢂ
[
53,56]
Ax ¼ Ax;0 ꢃ H þ ½Hþꢂ
nx
glucose and fructose, respectively.
Caratzoulas and Vlachos
ð6Þ
x;0
studied the energetics of the acid-catalyzed dehydration of
[
57]
fructose to HMF via the closed-ring mechanism. They found
that the reaction proceeds by way of intramolecular proton
and hydride transfers.
A , H , and n are correlating parameters to describe depend-
x,0
x,0
x
ence of the pre-exponential factor on acid concentration. Non-
catalyzed reactions were also performed at various tempera-
tures to study the effect of glucose decomposition without
HCl in the aqueous solutions. As shown in Figure 6a, the rate
of glucose disappearance increased with temperature, and
nearly full disappearance was reached at 1808C after 150 min.
Figure 6b reveals that the maximum attainable carbon yield of
HMF is 7.5% at 1808C after 60 min. Levulinic acid was detected
only at 1808C after 60 min, and a maximum carbon yield of
Furfural was also detected as a final byproduct of the dehy-
dration reaction of glucose, at less than 0.61% yield. As men-
tioned previously, it is possible that HMF is the precursor for
furfural accumulation. However, some have also postulated al-
ternative pathways to produce furfural from hexoses via a pen-
tose unit with formaldehyde or formic acid as byprod-
[
40,58,59]
ucts.
Incidentally, formic acid can also be produced di-
[
59]
[60,61]
rectly from C monosaccharides, as well as from furfural.
6
1
2% was obtained after 150 min. The total organic carbon
TOC) analysis confirmed that at high temperatures only 50%
of the water-soluble organic carbon was accounted for. Up to
0% of the overall organic carbon went into forming insoluble
Regardless, it is reasonable to assume that formic acid was
produced via multiple routes in addition to the conventional
pathway from HMF rehydration. Accordingly, an excess of
formic acid was detected throughout the entire study relative
to the kinetic model data, as depicted in Figure 5. Consequent-
ly, the LA-to-FA carbon molar ratio was lower than its stoichio-
metric value of five in this study.
(
5
humic species. This finding agrees with Figure 6c, which plots
Table 3. Estimated kinetic parameters for aqueous-phase glucose dehy-
[
a]
dration to levulinic acid with dependence on acid concentration.
ꢃ
ꢀ1Þꢄ
ꢀ
nx
ꢀ1
x
log10
A
x;0=ðM min
E
x
[kJmol
]
H
x,0
n
x
Effect of acid concentration
[
b]
1
2
3
4
18.44ꢁ0.98
3.86ꢁ0.52
11.50ꢁ0.83
16.83ꢁ3.43
160.16ꢁ5.15
50.68ꢁ2.38
94.72ꢁ5.54
141.94ꢁ25.72
0
1.290ꢁ0.062
Further glucose dehydration experiments were carried out at
acid concentrations with 0.5 and 1.0m HCl and temperatures
between 140–1808C. The initial glucose concentration was
kept constant at 10 wt%. The kinetic model fitted to the ex-
perimental data of glucose dehydration at 0.5 and 1.0m HCl
appear in the Supporting Information (refer to Figures S1 and
0.29ꢁ0.01 2.764ꢁ0.213
0
0
1.176ꢁ0.103
1.176ꢁ0.114
[a] Kinetic parameters fit experimental data at: T=140–1808C; [G]
0
=2–
0 wt%; 0m <[H ]ꢄ1.0m. [b] 95% confidence interval in parameter esti-
mation.
+
2
ChemSusChem 0000, 00, 1 – 12
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
5
&
ÞÞ
These are not the final page numbers!

G. W. Huber et al.
ulinic acid. Experimental studies were also carried out with
a PFR and a CSTR at temperatures between 160–1808C with
0.5m HCl. Steady-state conditions were attained after a period
corresponding to 4–5 times the residence time of the reactor.
This was confirmed by sampling multiple times at each steady-
state condition and taking the average value along with its
standard deviation. The experimental data along with the
fitted models are shown in Figure 7.
Figure 6. Aqueous-phase non-catalyzed thermal decomposition of glucose
in a stirred batch reactor. T (8C)=140(&), 160(~), 180(N). Feed was
10 wt% glucose. (a) Glucose disappearance; (b) HMF carbon yield;
(c) humins carbon yield.
the total humin carbon yield. For this purpose, humins were
considered non-detectable soluble and insoluble compounds.
Continuous reactor systems
Figure 7. Aqueous-phase acid-catalyzed glucose dehydration in continuous
reactors. Kinetic model fit with a single PFR (plots a, b, c) and single CSTR
(plots d, e, f) for glucose conversion, HMF carbon yield, and LA carbon yield.
Experimental reaction conditions: 3–5 wt% glucose and 0.5m HCl. T
The kinetic model derived above can be further used to model
data from continuous reactor systems. These can be divided
into two reactor types based on the mixing of the reactant—
PFR and CSTR. Under steady-state operating conditions, their
design equations are described as Equations (7) and (8):
(
8C)=160(*), 180(&); Model prediction for 1608C (a); Model prediction
for 1808C (c).
The kinetic model shows that the rate of glucose conversion
is higher in a PFR compared to a CSTR. Likewise, maximum
HMF carbon yields of 10% are obtained in a PFR at 1808C and
short residence times (less than 1 min). A maximum levulinic
acid yield of 55% is obtained in a PFR at 1608C and a residence
time of 100 min. This is compared to 46% LA carbon yield ob-
tained in a CSTR at the same temperature and residence time.
As shown, there is some inconsistency between the experi-
mental data and the theoretical model. This could be due to
a number of factors, such as non-ideal mixing patterns in the
reactors. From an operational point of view, challenges arise
due to the formation of solid humic species in the reactors.
This is predominantly encountered in the PFR, which ultimately
results in high pressure drops across the reactor and reduction
of the reactor volume. Therefore, only experiments at relatively
Z
Xi
dXi
PFR : t ¼ Ci;0
ð7Þ
ð8Þ
ꢀ
ri
0
C X
i;0
ri
i
CSTR : t ¼
ꢀ
where t is the residence time, Ci,0 the initial concentration of
species i, r the corresponding reaction rate, and X the conver-
i
i
sion of species i. The above equations were combined with
the kinetic model [Eqs. (3)–(6)] to simulate continuous produc-
tion of levulinic acid. The key process variables include initial
concentration of glucose, acid concentration, temperature, and
residence time. The variables can be manipulated to maximize
throughput, conversion of glucose, and yields of HMF and lev-
&6
&
www.chemsuschem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 12
ÝÝ
These are not the final page numbers!

Levulinic Acid Production
low temperatures and short residence times are currently feasi-
ble in the PFR. Conversely, the formation of solid humins has
a negligible effect on the operational aspect of the CSTR and
consequently harsher reaction conditions can be employed.
Table 4. Proposed kinetic models for aqueous-phase acid catalyzed glu-
cose conversion to levulinic acid.
[
a]
[c]
Proposed model
Reaction
conditions
Ea
[kJmol
Ref.
[
b]
ꢀ1
]
T=100–1508C
[HCl]=0.35m
Discussion
Ea
Ea
1
=133
=95
[
[
25]
26]
[
0
G] =1 wt%
2
Comparison with previous kinetic models
T=140–2508C
SO ]=0.0125–
.4m
The kinetic parameters derived from our model (Table 3) are in
fairly good agreement with those obtained in previous studies.
The literature reports activation energies for acid-catalyzed glu-
[
H
2
4
0
Ea
Ea
1
=137
=97
0
[G] =5–17 wt%
2
ꢀ1
[HMF] =1–2 wt%
cose dehydration (Reaction 1) in the range 121–152 kJmol
0
(
see Table 4 for references). Our study reported a value of
T=180–2248C
ꢀ1
1
60 kJmol with a 95% confidence interval of ꢁ5, which is in
[
H
2
SO
0.4m
G] =0.4–6 wt%
4
]=0.05–
the range of previous studies. The majority of the literature
values for acid-catalyzed HMF rehydration (Reaction 3) range
Ea =128
[42]
[36]
[44]
[28]
1
[
0
ꢀ
1
from 95 to 111 kJmol . This is in agreement with our reported
T=170–2308C
ꢀ
1
value of 95ꢁ6 kJmol . Girisuta et al. claimed activation ener-
gies of 165 and 111 kJmol for the formation of humins from
[
[
H
G]
3
PO
=0.6–6 wt%
=0.3 wt%
4
]: pH 1–4
ꢀ
1
1
Ea =121
0
Ea
2
=56
[
30]
glucose and HMF, respectively. For the formation of humins
from glucose (Reaction 2) and HMF (Reaction 4), we obtained
[HMF]
0
T=200–2308C
SO ]=0.005–
.02m
[G] =2 wt%
ꢀ
1
values of 51ꢁ2 and 142ꢁ26 kJmol , respectively. In their
[
H
2
4
study of non-catalyzed glucose decomposition, Jing and Lꢁ re-
0
Ea
1
=139
ꢀ
1
ported values of 136 and 109 kJmol for the formation of
0
[
32]
humins from glucose and HMF, respectively.
Wyman and
T=170–1908C
SO ]=0.1–
.5m
ꢀ
1
Shen reported a value of 147 kJmol for the formation of
[
H
2
4
Ea
1
Ea
2
Ea
3
=86
=210
=57
[
33]
humins from HMF. The latter value agrees well with our cal-
0
ꢀ
1
culated value of 142 kJmol . On the other hand, our derived
0
[G] =5 wt%
ꢀ
1
value of 51 kJmol for the formation of humins from glucose
does not coincide with those reported above.
T=98–2008C
[
[
H
G]
2
SO
=2–15 wt%
=1–
1 wt%
4
]=0.05–1m Ea
1
2
3
4
=152
=165
=111
=111
A power-law approach was used to derive the reaction
orders with respect to acid concentration in the rate equations
0
Ea
Ea
Ea
[29,30]
[HMF]
1
0
(
nx in Table 3). All reactions were found to demonstrate a near
first order dependence to the acid concentration, with the ex-
ception of Reaction 2 (glucose to humins), which exhibits
a near third order dependence. This differs from the values re-
T=180–2808C
non-catalyzed
Ea
Ea
Ea
Ea
Ea
1
2
3
4
5
=108
=136
=89
=109
=31
[G]
0
=1 wt%
[32]
[
30]
[HMF] =0.75 wt%
0
ported by Girisuta et al.
[
0
LA] =0.5 wt%
A plausible cause for these observed deviations lies in the
methods used to develop the kinetic models. For example,
T=140–1808C
Ea
1
Ea
2
Ea
3
Ea
4
=160ꢁ5
=51ꢁ2
[
29,30]
unlike our model, Girisuta et al.
used a modified Arrhenius
0m<[HCl]ꢄ1.0m
this
study
[
0
G] =2–20 wt%
equation to determine temperature dependence of the rate
constants and a rate selectivity parameter to maximize the rate
of the desired reactions. They also took into account the disso-
ciation constant of their catalyst, as they used sulfuric acid.
Our model prediction has been made for a wide range of acid
concentrations extended to zero acid concentration to assure
systematic dependence of rate constants on the concentration
of the catalyst, that is, infinite dilution, where some of the reac-
tion rates become negligible. Table 4 summarizes the kinetic
studies that appear in the literature for glucose conversion to
levulinic acid.
=95ꢁ6
[
HMF]
0
=4–
=142ꢁ26
16 wt%
[
a] G=glucose; HMF=5-hydroxymethylfurfural; LA=levulinic acid; FA=
formic acid; I=intermediate; D=decomposition products (humins).
b] Units of feedstock and acid concentrations were converted for consis-
[
tency. Values were rounded to the nearest unit. [c] Activation energy.
of reactor and its operating conditions can be modified to
maximize the yields of these desired products. The kinetic
model fit in Figure 7 shows that at a constant residence time
higher glucose conversions and levulinic acid yields can be ob-
tained in a PFR compared to a CSTR. Likewise, higher tempera-
tures and short residence times are essential to maximize the
HMF yield. These conditions are favored in a PFR-type reactor.
Reactor design for production of HMF and levulinic acid
The apparent rate parameters introduced here allow for theo-
retical calculations of HMF and levulinic acid yields. The type
ChemSusChem 0000, 00, 1 – 12
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
7
&
ÞÞ
These are not the final page numbers!

G. W. Huber et al.
This finding also agrees with our previous work on furfural pro-
also favorable to maximize levulinic acid production, as these
conditions minimize the formation of humins due to the rela-
tively higher activation energy associated with Reaction 4
(humins from HMF) compared with the rehydration reaction to
produce levulinic acid (Reaction 3). Longer residence times are
required in a CSTR compared to a PFR to obtain equivalent
glucose conversions. As shown in Figure 8b, the longer resi-
dence times achieved in a CSTR are favorable in the case of
levulinic acid production, as we have shown that it does not
undergo degradation reactions under the reaction conditions
in this study.
[
62]
duction from xylose. In contrast, lower temperatures are nec-
essary to obtain optimal levulinic acid yields. This is due to the
induced HMF degradation reaction (Reaction 4) that occurs at
relatively higher temperatures (Ea >Ea ).
4
3
The calculated HMF and levulinic acid yields are shown in
Figure 8 as a function of glucose conversion for an ideal PFR
and CSTR. On a conversion basis, the HMF yield is maximized
[30]
This behavior agrees with those reported by Girisuta et al.
However, their kinetic model calculations show a bigger devia-
tion between a CSTR and PFR, with increased glucose conver-
sion. They report a LA carbon yield of 70% in a CSTR at 1408C
and complete glucose conversion. Our model predicts an LA
carbon yield of 54% at the same temperature and conversion
for both reactor types, as shown in Figure 8b. This result is
comparable to that reported by Girisuta et al. in a PFR; howev-
er, there is an inconsistency with the levulinic acid yield report-
ed in a CSTR at complete glucose conversion. To further deter-
mine the credibility of our model, we calculated the projected
glucose conversion and levulinic acid yield obtained in a CSTR
[30]
with the kinetic parameters reported by Girisuta et al. These
calculations resulted in a glucose conversion of 89.8% and an
LA carbon yield of 50.4% at 1608C, 0.5n acid concentration,
and retention time of 200 min. These results are quite similar
to those reported in this study, as shown in Figure 7d and f.
Experimentally, at the same reaction conditions, we obtained
a glucose conversion of 95.1% and an LA carbon yield of
51.3%. Likewise, our kinetic model projected a glucose conver-
sion of 92.0% and an LA carbon yield of 50.4%.
The calculated LA carbon yield, plotted as a function of tem-
perature and residence time in a PFR, is shown in Figure 9.
In a PFR, 56% carbon yield can be achieved at 1538C after
200 min. The carbon yield rises to 57% with an increase in resi-
Figure 8. Continuous reactor modeling for acid-catalyzed glucose dehydra-
tion in a single continuous reactor. Calculated values for (a) HMF carbon
yield and (b) LA carbon yield as a function of glucose conversion and tem-
perature for 10 wt% glucose and 0.5m HCl. PFR (c); CSTR (g). Symbols
dence time to 500 min at 1498C. However, a further increase
to 1000 min shows no improvement in the results. If the resi-
dence time in the PFR decreases, then a higher temperature is
required to maximize levulinic acid production. This in turn re-
sults in lower yields of levulinic acid due to the induced forma-
tion of humins from HMF (Reaction 4). The levulinic acid yield
has an optimum with regard to temperature and residence
time. On the one hand, HMF production from glucose (Reac-
tion 1) is maximized at elevated temperatures (Ea >Ea ). Con-
(
N) and (I) represent maximum HMF yields in a PFR and CSTR, respectively.
in a PFR at high temperatures. This is because the dehydration
step to form HMF (Reaction 1) is favored at increased tempera-
tures due to its higher activation energy compared to the deg-
radation step to form humins from glucose (Reaction 2).
At similar feedstock conversions shorter residence times are
obtained in a PFR compared to a CSTR. This shorter residence
time minimizes the further decomposition of HMF and maxi-
mizes the HMF production. Calculations show that a HMF yield
of 14% can be obtained at 2008C in a PFR at 34% glucose
conversion. This conversion corresponds to a residence time of
1
2
versely, high temperatures are unfavorable for levulinic acid
production from HMF (Reaction 3) due to a parallel degrada-
tion step with a higher activation energy (Ea >Ea ).
4
3
In addition to operating parameters, a variety of reactor con-
figurations can be examined, including a combination of two
reactors in series. Figure 10 plots the overall calculated levulin-
ic acid yield for two reactors in series as a function of the tem-
perature and residence time of the second reactor (T and t ,
10s.
2
2
At equal glucose conversions, a slightly higher yield of levu-
respectively). The first reactor is a PFR, as we have shown that
this reactor type is favorable to maximize levulinic acid yield.
Its operational conditions were set to maximize the levulinic
acid yield in the first reactor. The second reactor depicts a PFR
linic acid can be obtained in a CSTR compared to a PFR (11.6
vs. 10.5% respectively, at 25% glucose conversion and 1608C),
as shown in Figure 8b. Lower temperatures (140–1608C) are
&8
&
www.chemsuschem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 12
ÝÝ
These are not the final page numbers!

Levulinic Acid Production
As mentioned previously, this commercial process consists of
two acid-catalyzed steps: (1) hydrolysis of the lignocellulosic
biomass feedstock to monosaccharides and subsequent dehy-
dration to produce HMF; (2) rehydration of HMF to produce
levulinic and formic acid. Furfural is also produced during the
process. The first stage takes place in a PFR reactor between
210–2308C within 13–25 s. The second stage consists of
a CSTR reactor at temperatures between 195–2158C and resi-
dence times of 15–30 min.
The general operating trend realized by the Biofine process
agrees with our model, in that initially higher temperatures are
necessary to maximize the hydrolysis/dehydration step and
then lower temperatures should be employed to maximize the
levulinic acid yield. However, a discrepancy arises with respect
to the optimal reactor configuration. The Biofine process uses
a PFR followed by a CSTR, whereas our model predicts a single
PFR as the most favorable. This inconsistency could be due to
the fact that, unlike the Biofine process, a biomass hydrolysis
step was not included in our proposed model. This preliminary
depolymerization step has a high energy barrier and requires
Figure 9. Continuous reactor modeling for acid-catalyzed glucose dehydra-
tion in a PFR. Calculated LA carbon yield as a function of temperature for
1
1
0 wt% glucose and 0.5m HCl at varying residence times. t (min)=25(g),
00(a), 200(c), 500(d), 1000(ꢀ··ꢀ··). Symbol (N) represents maxi-
mum LA carbon yield.
[24,63]
or CSTR as shown in Figure 10a and b, respectively. As can be
seen, a combination of two PFR reactors in series is preferred
relative to a PFR and CSTR combination. The former gives rise
to a higher levulinic acid yield (52.5 vs. 47.5%) at shorter resi-
dence times in the second reactor (55 vs. 93 min). The contour
plots demonstrate that a decrease in the temperature of the
second reactor subsequently requires an increase in its resi-
dence time to maintain equivalent levulinic acid yields. The
combination of two PFRs in series shows nearly the same re-
sults as a single PFR. According to Figure 7, a single PFR at
relatively high temperatures,
which could necessitate
[21]
a multi-stage process.
Hayes et al. have suggested that
higher yields are obtained in a CSTR in the Biofine process be-
cause this reactor minimizes the “higher-order” degradation re-
actions, compared to the first order rehydration of HMF to pro-
[64]
duce levulinic acid. Their claim differs from our proposed ki-
netic model, which suggests that all four of the reactions are
first order with respect to the reactants [refer to Eqs. (3)–(5)].
A CSTR may also be beneficial compared to a PFR because of
the operational issues associated with the formation of solid
humins during the reaction. As observed in this study, it may
well be easier to operate a CSTR amid the formation of solid
humins compared to a PFR.
1608C and residence time of 50 min can yield 51.7% levulinic
acid, which is similar to the maximum yield of 52.5% obtained
with two PFRs in series at 1608C and residence time of 60 min.
Our proposed model for levulinic acid production from glu-
cose differs from that employed by the Biofine process.
According to the Biofine process, levulinic acid production
can reach carbon yields higher than 58% (70% of the theoreti-
Figure 10. Continuous reactor modeling for acid-catalyzed glucose dehydration in a system of two reactors in series. Calculated total LA carbon yield as
a function of the residence time (t ) and temperature (T ) of the second reactor for 10 wt% glucose. For both cases the first reactor is a PFR at: T =2008C,
=5 s, and 0.5m HCl. The second reactor is (a) PFR or (b) CSTR, both at 0.5m HCl.
2
2
1
t
1
ChemSusChem 0000, 00, 1 – 12
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
9
&
ÞÞ
These are not the final page numbers!

G. W. Huber et al.
[
22,23]
cal yield) from cellulosic biomass.
Similarly, in this study
formance than a PFR and CSTR combination. Compared to
a single PFR, for the aqueous-phase levulinic acid production
from glucose, there is no distinct advantage to implementing
a system of two consecutive reactors.
we project a maximum carbon yield of 57% (68% of the theo-
retical yield) levulinic acid from glucose, as shown in Figure 9.
Likewise, in their kinetic study of levulinic acid production
from cellulose, Wyman and Shen report a maximum LA carbon
yield of 50% (60% of the theoretical) at an initial 99.6mm cel-
lulose concentration, a 0.927m acid concentration, and tem-
Experimental Section
[
33]
peratures between 180–2008C.
Reaction kinetics measurements
Batch reactions were carried out in a 100 mL reactor vessel provid-
ed by Parr Instrument Company, series 4560. The feedstock solu-
tions were prepared with deionized (DI) water at the specified con-
centrations. Acidic solutions were prepared with HCl (Fisher Scien-
tific). Glucose was provided by Fisher Scientific. HMF (99%) was
provided by Sigma Aldrich. Levulinic acid (98%) and formic acid
(98%) were provided by Acros Organics. Temperatures in the reac-
tor were measured by a thermocouple in the solution. All reaction
solutions were mixed at a maximum constant rate of 600 rpm
using an internal stirrer. The temperature and stirring were con-
trolled by a 4848 Controller provided by Parr. The reaction vessel
was initially pressurized to 5.5 MPa with industrial grade helium
Conclusions
In this paper we have developed a kinetic model for aqueous-
phase glucose dehydration to produce 5-hydroxymethylfurfural
(HMF) and levulinic acid. Our model involves four reactions.
Glucose first undergoes a dehydration reaction in which three
molecules of water are removed to produce HMF (Reaction 1).
In a parallel step, glucose can undergo reversion and decom-
position reactions to form humins, which are highly polymer-
ized insoluble carbonaceous species (Reaction 2). Once HMF is
formed, it can also undergo parallel reactions. In the presence
of water, a rehydration reaction takes place with two mole-
cules of water to produce levulinic acid and formic acid (Reac-
tion 3). Likewise, HMF can also decompose to form humic spe-
cies (Reaction 4).
(Airgas). Samples were taken periodically by way of a sampling
port. The samples were immediately quenched in an ice water
bath and filtered with a 0.2 mm syringe filter prior to analysis. The
reactor was repressurized with helium after each sampling.
Continuous reactions were carried out in both a PFR and a CSTR.
The PFR reactor was prepared by using a stainless steel tube of
6.35 mm outer diameter (OD). A Varian HPLC pump (Prostar 210)
was used to introduce the feedstock into the reactor at flow rates
The proposed kinetic model is consistent with the experi-
mental data for batch reactions within the conditions of this
study. Some inconsistency was observed with regard to the
continuous reactors, predominantly with the PFR. Formation of
solid humins during the reaction is the probable cause of this
discrepancy. Minimizing the occurrence of humins would con-
sequently improve the operational aspect of the continuous
reactors.
ꢀ1
ranging from 0.065–1.293 mLmin . The reactor was heated by
means of a heating tape (McMaster–Carr), and insulation tape was
used to minimize heat losses. A thermocouple was placed adjacent
to the reactor wall to measure the temperature and was connected
to a temperature controller. The reactor system was initially pres-
surized to 4.1 MPa with UHP grade helium (Airgas). Liquid products
were recovered in a sample vessel at room temperature. The sam-
ples were filtered with a 0.2 mm syringe filter prior to analysis.
Theoretical calculations have allowed us to recognize the
optimal reactor configuration and operating conditions to ach-
ieve maximum HMF and levulinic acid yields. In general, higher
temperatures (i.e., 180–2008C) and short reaction times of less
than 1 min are essential to maximize the HMF content. On the
other hand, low temperatures between 140–1608C and long
residence times of greater than 100 min are essential for maxi-
mum levulinic acid yield.
The CSTR with a 100 mL reactor vessel was modified from a Parr
batch reactor, allowing continuous liquid flow in and out of the re-
actor. A Varian HPLC pump (Prostar 210) was used to introduce the
feedstock into the reactor through a port on the reactor cap. The
ꢀ1
flow rates ranged from 0.300–2.400 mLmin . A dip tube was used
as the outlet of the reactor. The temperature and stirring were con-
trolled as described for the batch reactions. The reaction vessel
was initially pressurized to 5.5 MPa with industrial grade helium
A plug flow-type reactor (PFR)-type reactor is favorable for
the aqueous-phase production of HMF and levulinic acid from
glucose, as compared with a continuously stirred tank reactor
(Airgas). A back-pressure regulator was used to monitor the pres-
sure of the system. Liquid products were recovered in a sample
vessel at room temperature. The samples were filtered with
a 0.2 mm syringe filter prior to analysis.
(CSTR). Higher HMF yields can be obtained in a PFR at relative-
ly shorter residence times. Likewise, compared to a PFR,
a CSTR requires longer residence times to attain comparable
levulinic acid yields. A maximum calculated LA carbon yield of
57%, or 68% of the theoretical yield, can be obtained in a PFR
Analysis
at 1498C and a residence time of 500 min. The optimal operat-
ing conditions for HMF production are 2008C and a reaction
time of 10 s in a PFR-type reactor with a maximum attainable
carbon yield from glucose of 14%. Overall, from an economical
and operational point of view, there is a trade-off between the
reactor temperature and residence time. Shorter residence
times require higher temperatures, which can consequently
jeopardize the final yield of levulinic acid. Finally, we have
shown that a system of two consecutive PFRs has a higher per-
Samples were analyzed by means of HPLC with a Shimadzu LC-
2
0AT. Carbohydrates were detected with a refractive index detector
(
(
RID-10A), and products were detected with a UV-Vis detector
SPD-20AV) at wavelengths of 210 and 254 nm. A Biorad Aminex
HPX-87H sugar column was used. The mobile phase was 5mm
ꢀ1
H SO flowing at a rate of 0.6 mLmin . The column oven was set
2
4
to 308C. The TOC measurements were performed with a Shimadzu
TOC-VCPH Analyzer. Calibrations were performed with carbon
standards supplied by SpectroPure.
&10
&
www.chemsuschem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 12
ÝÝ
These are not the final page numbers!

Levulinic Acid Production
Modeling
[24] J. F. Saeman, Ind. Eng. Chem. 1945, 37, 43.
25] K. R. Heimlich, A. N. Martin, J. Am. Pharm. Assoc. Sci. Ed. 1960, 49, 592.
[
Experimental data was collected and used to compare with the
proposed kinetic model to estimate rate parameters of the reaction
pathways for levulinic acid production from glucose. The kinetic
model for the overall reaction paths was a set of coupled ordinary
differential equations (ODEs) and rate constants were correlated by
the Arrhenius equation to include temperature dependence. The
pre-exponential factors and activation energies were set as adjust-
able parameters to simultaneously correlate a complete set of con-
centration data for reactants and products at different tempera-
tures (120–1808C). The sum of absolute errors between estimated
and observed values was minimized to find the best fit with the
observed reaction rates of glucose dehydration and levulinic acid
formation. Matlab and Athena Visual Studio (v14.0) were used for
the numerical integration of ODEs and parameter estimations.
[26] S. W. McKibbins, J. F. Harris, J. F. Saeman, W. K. Neill, For. Prod. J. 1962,
12, 17.
[27] B. Singh, G. R. Dean, S. M. Cantor, J. Am. Chem. Soc. 1948, 70, 517.
[
[
[
28] C. Chang, X. Ma, P. Cen, Chin. J. Chem. Eng. 2006, 14, 708.
29] B. Girisuta, L. P. B. M. Janssen, H. J. Heeres, Green Chem. 2006, 8, 701.
30] B. Girisuta, L. P. B. M. Janssen, H. J. Heeres, Chem. Eng. Res. Des. 2006,
8
4, 339.
31] B. Girisuta, L. P. B. M. Janssen, H. J. Heeres, Ind. Eng. Chem. Res. 2007, 46,
696.
32] Q. Jing, X. Lꢁ, Chin. J. Chem. Eng. 2008, 16, 890.
[
1
[
[33] J. Shen, C. E. Wyman, AIChE J. 2012, 58, 236.
[34] J. Horvat, B. Klaic, B. Metelko, V. Sunjic, Tetrahedron Lett. 1985, 26, 2111.
[35] J. Horvat, B. Klaic, B. Metelko, V. Sunjic, Croat. Chem. Acta 1986, 59, 429.
[
[
36] K. D. Baugh, P. L. McCarty, Biotechnol. Bioeng. 1988, 31, 50.
37] P. R. Bienkowski, M. R. Ladisch, R. Narayan, G. T. Tsao, R. Eckert, Chem.
Eng. Commun. 1987, 51, 179.
[
38] K. Kato, T. Doihara, F. Sakai, N. Takahashi, Kenkyu Hokoku Nippon Senbai
Kosha Chuo Kenkyusho 1966, 108, 361.
Acknowledgements
[
[
39] A. Ohnishi, K. Kato, E. Takagi, Polym. J. 1975, 7, 431.
40] R. Krishna, M. R. Kallury, C. Ambidge, T. T. Tidwell, D. G. B. Boocock, F. A.
Agblevor, D. J. Stewart, Carbohydr. Res. 1986, 158, 253.
The authors gratefully acknowledge financial support from NSF-
CBET (Grant 0756663) and NSF MRI (Grant 0722802). We also sin-
cerely thank Alejandro Fernꢀndez for his contribution in the lab.
[41] G. C. A. Luijkx, F. van Rantwijk, H. van Bekkum, Carbohydr. Res. 1993,
42, 131.
2
[
[
42] P. C. Smith, H. E. Grethlein, A. O. Converse, Solar Energy 1982, 28, 41.
43] H. E. van Dam, A. P. G. Kieboom, H. van Bekkum, Starch/Staerke 1986,
3
8, 95.
Keywords: flow reactor · glucose · homogeneous catalysis ·
kinetics · reaction mechanism
[
[
[
44] Q. Xiang, Y. Lee, R. Torget, Appl. Biochem. Biotechnol. 2004, 115, 1127.
45] D. W. Harris, M. S. Feather, J. Am. Chem. Soc. 1975, 97, 178.
46] J. B. Binder, A. V. Cefali, J. J. Blank, R. T. Raines, Energy Environ. Sci. 2010,
[
[
1] R. Leonard, Ind. Eng. Chem. 1956, 48, 1330.
2] J. J. Bozell, L. Moens, D. C. Elliott, Y. Wang, G. G. Neuenscwander, S. W.
Fitzpatrick, R. J. Bilski, J. L. Jarnefeld, Resour. Conserv. Recycl. 2000, 28,
3
, 765.
47] M. Moliner, Y. Romꢂn-Leshkov, M. E. Davis, Proc. Natl. Acad. Sci. USA
010, 107, 6164.
[
[
[
2
227.
48] Y. Romꢂn-Leshkov, M. Moliner, J. A. Labinger, M. E. Davis, Angew. Chem.
Int. Ed. 2010, 49, 8954.
49] E. Nikolla, Y. Romꢂn-Leshkov, M. Moliner, M. E. Davis, ACS Catal. 2011, 1,
[
[
[
[
[
3] S. W. Fitzpatrick, ACS Symp. Ser. 2006, 921, 271.
4] D. C. Elliott, J. G. Frye, US 5883266, 1999.
5] S. Bayan, E. Beati, Chim. Ind. 1941, 23, 432.
6] G. W. Huber, S. Iborra, A. Corma, Chem. Rev. 2006, 106, 4044.
7] H. Heeres, R. Handana, D. Chunai, C. B. Rasrendra, B. Girisuta, H. J.
Heeres, Green Chem. 2009, 11, 1247.
4
08.
[
[
[
[
50] B. F. M. Kuster, H. M. G. Temmink, Carbohydr. Res. 1977, 54, 185.
51] C. Perez Locas, V. A. Yaylayan, J. Agric. Food Chem. 2008, 56, 6717.
52] E. F. L. J. Anet, Aust. J. Chem. 1961, 14, 295.
53] R. S. Assary, P. C. Redfern, J. R. Hammond, J. Greeley, L. A. Curtiss, J. Phys.
Chem. B 2010, 114, 9002.
54] M. Chidambaram, A. T. Bell, Green Chem. 2010, 12, 1253.
55] M. J. Antal, W. S. L. Mok, G. N. Richards, Carbohydr. Res. 1990, 199, 91.
56] R. S. Assary, P. C. Redfern, J. Greeley, L. A. Curtiss, J. Phys. Chem. B 2011,
[
[
8] L. Deng, Y. Zhao, J. Li, Y. Fu, B. Liao, Q.-X. Guo, ChemSusChem 2010, 3,
1
172.
9] I. T. Horvꢂth, H. Mehdi, V. Fabos, L. Boda, L. T. Mika, Green Chem. 2008,
0, 238.
[
[
[
1
[
[
10] J.-P. Lange, R. Price, P. M. Ayoub, J. Louis, L. Petrus, L. Clarke, H. Gosse-
link, Angew. Chem. 2010, 122, 4581; Angew. Chem. Int. Ed. 2010, 49,
115, 4341.
4
479.
11] J. Q. Bond, D. M. Alonso, D. Wang, R. M. West, J. A. Dumesic, Science
010, 327, 1110.
[
[
[
57] S. Caratzoulas, D. G. Vlachos, Carbohydr. Res. 2011, 346, 664.
58] K. Kato, H. Komorita, Agricultural and Biological Chemistry 1968, 32, 715.
59] Z. Srokol, A.-G. Bouche, A. van Estrik, R. C. J. Strik, T. Maschmeyer, J. A.
Peters, Carbohydr. Res. 2004, 339, 1717.
60] A. P. Dunlop, Ind. Eng. Chem. 1948, 40, 204.
61] D. L. Williams, A. P. Dunlop, Ind. Eng. Chem. 1948, 40, 239.
62] R. Weingarten, J. Cho, W. C. Conner, G. W. Huber, Green Chem. 2010, 12,
2
[
[
[
12] D. M. Alonso, J. Q. Bond, J. A. Dumesic, Green Chem. 2010, 12, 1493.
13] F. Leger, H. Hibbert, Can. J. Res., Sect. B 1938, 16, 68.
14] B. V. Timokhin, V. A. Baransky, G. D. Eliseeva, Russian Chemical Reviews
[
[
[
1999, 68, 73.
[
[
[
[
[
15] W. A. Farone, J. E. Cuzens, US 6054611, 2000.
16] J. Y. Cha, M. A. Hanna, Ind. Crops Prod. 2002, 16, 109.
17] Q. Fang, M. A. Hanna, Bioresour. Technol. 2002, 81, 187.
18] L. Yan, N. Yang, H. Pang, B. Liao, Clean Soil Air Water 2008, 36, 158.
19] L. Peng, L. Lin, J. Zhang, J. Zhuang, B. Zhang, Y. Gong, Molecules 2010,
1
423.
[
[
63] R. D. Fagan, H. E. Grethlein, A. O. Converse, A. Porteous, Environ. Sci.
Technol. 1971, 5, 545.
64] D. J. Hayes, S. Fitzpatrick, M. H. B. Hayes, J. R. H. Ross, The Biofine Pro-
cess—Production of Levulinic Acid, Furfural, and Formic Acid from Ligno-
cellulosic Feedstocks, Wiley-VCH, Weinheim, 2008.
15, 5258.
[
[
20] B. Girisuta, University of Groningen (Groningen), PhD thesis 2007.
21] D. W. Rackemann, W. O. S. Doherty, Biofuels Bioprod. Biorefin. 2011, 5,
198.
[
[
22] S. W. Fitzpatrick, Biofine Incorporated, US 4897497, 1990.
23] S. W. Fitzpatrick, US 5608105, 1997.
Received: November 9, 2011
Published online on && &&, 0000
ChemSusChem 0000, 00, 1 – 12
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
11
&
ÞÞ
These are not the final page numbers!

FULL PAPERS
R. Weingarten, J. Cho, R. Xing,
W. C. Conner, Jr, G. W. Huber*
&& – &&
Kinetics and Reaction Engineering of
Levulinic Acid Production from
Aqueous Glucose Solutions
Plug wins over stirring: A plug flow re-
actor is preferred over a continuously
stirred tank reactor for aqueous-phase
production of 5-hydroxymethylfurfural
and levulinic acid from glucose (estimat-
ed levulinic acid carbon yield of 57%—
68% of theoretical yield). There is no
distinct advantage to implementing
a system of two consecutive reactors
compared to a single-plug flow reactor.
&12
&
www.chemsuschem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 12
ÝÝ
These are not the final page numbers!