Hydrochloric Acid-Catalyzed Levulinic Acid

Formation from Cellulose: Data and Kinetic

Model to Maximize Yields
Jiacheng Shen
Northwest Irrigation and Soils Research Laboratory, United States Dept. of Agriculture, Kimberly, ID 83341
Charles E. Wyman
Center for Environmental Research and Technology, and Chemical and Environmental Engineering Dept.,
Bourns College of Engineering, University of California, Riverside, CA 92507

DOI 10.1002/aic.12556
Published online March 17, 2011 in Wiley Online Library (wileyonlinelibrary.com).

In this study, the kinetics of the acid catalyzed hydrolysis of microcrystalline cellulose
(Avicel PH101) to levulinic (LA) and formic (FA) acids was investigated in a batch reactor
over the following range of conditions: 160–200
C, hydrochloric acid concentrations of
0.309–0.927 M (11.3–33.8 g/l), cellulose concentrations of 49.8–149 mM (8.06–24.1 g/l),
and residence times of 0–50 min. The maximum LA yield of around 60% of theoretical
was achieved for an initial cellulose concentration of 99.6 mM, acid concentration 0.927
M, and 180–200
C. A mathematical model and its analytical solution were developed to
predict conversion of cellulose to LA and FA through glucose and hydroxymethyl-2-furfu-
ral based on an irreversible pseudo-first order reaction. Rate analysis of each reaction
indicated that the rate-controlling step shifted from LA formation initially to HMF forma-
tion later. V
C 2011 American Institute of Chemical Engineers AIChE J, 58: 236–246, 2012

Keywords: cellulose, levulinic acid, hydrochloric acid, kinetics, model, rate control

Introduction avenues to achieve the highest possible yields from cellu-

losic materials. Development of robust kinetic models can
Levulinic acid (H8C5O3, gamma-ketovaleric acid, LA) and
serve this goal well by providing new insight into interplays
formic acid (H2CO2, FA) are short-chain fatty acids that can
among the various reaction steps that can help identify ave-
provide potentially important platforms for production of liq-
nues that favor LA production.
uid fuels and chemicals.1 LA can be produced from glu-
Some mathematical models of LA formation have been
cose,2,3 cellulose,4,5 5-hydroxymethyl-2-furfural (HMF),6
normal corn starch,7 whole kernel grain sorghum,8 water developed, with the most common being based on that
hyacinth plant,9 bagasse and paddy straw,10 and wheat developed by Saeman in 1945.12 In this case, the author
straw.11 Although focus on cellulosic feedstocks is important assumed a two-step reaction from cellulose to decomposition
in the long term to achieve low costs and make a substantial products through glucose (Eq. 1) with the rate constant
impact on fuel consumption, more information is needed on including the effect of acid concentration along with a modi-
fied Arrhenius equation for its temperature dependence:
The views, opinions, and/or findings contained in this article/presentation are
those of the author/presenter and should not be interpreted as representing the offi-
cial views or policies, either expressed or implied, of the Defense Advanced
Cellulose ! Glu cos e ! Decomposition products (1)
Research Projects Agency or the Department of Defense.
Correspondence concerning this article should be addressed to J. Shen at jiacheg. A more detailed reaction mechanism of LA formation from
[email protected].
cellulose through glucose and HMF as well as some parallel
V
C 2011 American Institute of Chemical Engineers by-product reactions to humins was proposed by Girisuta

236 January 2012 Vol. 58, No. 1 AIChE Journal

et al.5 Based on this mechanism, they developed a model of and reaction time; (3) develop a kinetic model with an ana-
LA formation from cellulose that involves four differential lytical solution to describe LA production from cellulose; (4)
equations to track each of the major compounds. However, determine model parameters based on the experimental data;
because the cellulose content in solid residues after reaction and (5) apply the model to identify promising strategies to
was not measured in these experiments, cellulose concentra- enhance LA yields.
tion vs. time was not simulated by their model, and the model
is not completely verified by accounting for all of the reaction
participants. In addition, analytical solutions were not devel- Kinetic Model Development
oped for these equations, potentially limiting the utility of the The model assumptions were as follows:
model. Jing and Lu3 published a model with an analytical 1) the glucan in cellulose decomposes to LA and FA fol-
solution describing LA production based on a similar reaction lowing reaction to glucose and then 5-hydroxymethyl-2-
mechanism but only considered glucose as a reactant. Because furfural (HFM), with the final product of the series of reac-
cellulosic biomass could be a widely available feedstock for tions being decomposition products;
LA production, development of a model from cellulose to LA 2) two parallel reactions produce humins from glucose
with an analytical solution should be a valuable tool in and HMF; and
designing reaction systems that can realize the high yields vital 3) the reaction of cellulose to glucose is pseudo-homoge-
to success. neous irreversible first order, the other reactions are irrevers-
In this study, our objectives were to (1) develop data on ible first order, and the proton catalyst concentration is
production of LA and FA from microcrystalline cellulose assumed constant throughout the reaction so that it can be
(PH 101) in a batch reactor; (2) determine conditions to combined into the rate constants.
maximize LA yield based on the four operating parameters These assumptions result in the following reaction
of cellulose concentration, acid concentration, temperature, formula (2):

ð2Þ

According to the mass action law, the rate equations for describe cellulose, glucose, HMF, and LA concentrations
reactions (1–6) can be expressed, respectively, as over time:

dC
¼ k1 C (3) C ¼ C0 expðk1 tÞ (8)
dt
k1 C 0
dG G¼ fexpðk1 tÞ  exp½ðk2 þ k3 Þtg (9)
¼ k1 C  ðk2 þ k3 ÞG (4) k 2 þ k3  k 1
dt


dM k1 k2 C 0 expðk1 tÞ exp½ðk2 þ k3 Þt
¼ k2 G  ðk4 þ k5 ÞM (5) M¼ 
dt k2 þ k3  k1 k 4 þ k5  k 1 k4 þ k 5  k2  k3
k1 k2 C0 exp½ðk4 þ k5 Þt
dL
¼ k4 M  k 6 L þ (10)
dt
(6) ðk4 þ k5  k1 Þðk4 þ k5  k2  k3 Þ
8 9

: a >
ki ¼ ki0 H ¼ Ai H exp>
E
; i¼16 k1 k2 k4 C0 expðk1 tÞ  expðk6 tÞ
(7) L¼
RT k2 þ k3  k1 ðk4 þ k5  k1 Þðk6  k1 Þ

in which C, G, M, L, and H represent cellulose, glucose, exp½ðk2 þ k3 Þt  expðk6 tÞ

HMF, LA, and proton concentrations (M), respectively, k1, ðk4 þ k5  k2  k3 Þðk6  k2  k3 Þ
k2, k3, k4, k5, and k6 are rate constants that incorporate the k1 k2 k4 C0 fexp½ðk4 þ k5 Þt  expðk6 tÞg
proton concentration for reactions 1–6 (min1), k0i is a rate þ (11)
ðk4 þ k5  k1 Þðk4 þ k5  k2  k3 Þðk6  k4  k5 Þ
constant that do not include the proton concentration (l/
(mol.min)), Ai and Eai are pre-exponential factors and
The mol fraction of unreacted cellulose, yC, and the yields of
activation energies for reactions 1–6, respectively ((l/(mol
glucose, yG, HMF, yM, and LA, yL, can be expressed as:
min) and J/mol), R is the universal gas constant (J/(mol K),
and T is the reaction temperature (K). Integration of Eqs. 3–
6 with initial conditions C ¼ C0 and G ¼ M ¼ L ¼ 0 at t C or G or M or L
yC or G or M or L ¼ y ¼ ½0; 1 (12)
¼ 0 results in the following analytical expressions to C0

AIChE Journal January 2012 Vol. 58, No. 1 Published on behalf of the AIChE DOI 10.1002/aic 237
Figure 1. LA (solid) and FA (open) concentrations at Figure 2. Glucose (solid) and HMF (open) concentra-
three initial cellulose concentrations vs. reac- tions at three initial cellulose concentrations
tion time: 49.8 mM (&), 99.6 mM ^, and 149 vs. time: 49.8 mM (&), 99.6 mM ^, and 149
mM (D). mM (D).
Conditions: 200
C and acid 0.206 M HCl concentration. Conditions were the same as Figure 1.

105
C in an oven overnight. Then, the filter paper, including the

Materials and Methods
residues, was weighed, and the residue weight was calculated.
Materials The washed solids remaining after reaction were analyzed.
Avcel PH 101 microcrystalline cellulose (power with size
0.05 mm, Eluka BioChemika) and hydrochloric acid (Sigma– Analytical methods
Aldrich) were the raw material and catalyst for LA production,
Glucose, FA, LA, and hydroxymethyl furfural concentra-
respectively. LA (Sigma–Aldrich 98%), FA (Fisher Scientific,
tions were measured using a Waters model 2695 HPLC sys-
99%), D-(þ)-glucose (Sigma–Aldrich, 99.5%), and 5-hydroxy-
tem equipped with a 2414 refractive detector and a Waters
methyl-2-furaldehyde, (SAFC, 99%) were used as standards for
2695 autosampler using Millenium32 chromatography man-
HPLC analysis. Calcium carbonate (Fisher Scientific, 99.0%)
ager 3.2 software (Waters Co., Milford, MA). A Bio-Rad
was used to neutralize acid in samples before HPLC analysis.
Aminex HPX-87H column (Bio-Rad Laboratories, Hercules,
Experimental methods CA) was used for analyzing these compounds. The column
temperature was 65
C, the mobile phase was 0.005 M
Tubular batch reactors [Hastelloys C-276, O.D. 1=200 (12.7 H2SO4, the flow rate was 0.6 ml/min, and the HPLC was
mm) of 600 length (152.4 mm)] were used for LA and FA operated in the isocratic mode. The identities of the com-
production from Avicel PH 101. Appropriate amounts of pounds were authenticated by comparing their retention
Avicel PH 101, hydrochloric acid, and deionizied water were times with those of pure compounds (Sigma–Aldnch, St.
added to the tubes for each experiment to give a total volume, Louis, MO). Before analytical work, the acid in the filtrates
including the solid cellulose, of 10 ml, thereby allowing for was neutralized with calcium carbonate to a pH of 5 to 6,
expansion of the contents on heating. The tubes were capped and then the suspensions were centrifuged at 15,000 rpm for
tightly with Swedgelock fittings. The tube reactors were heated
using two 4 kW fluidized sand baths (Model SBL-2D, Techne
Co., Princeton, NJ), and their internal temperature was moni-
tored with a K type thermocouple probe (Omega CASS-18U-12,
Omega Engineering Co., Stamford, CT). The tubes were sub-
merged in the first sand bath that was set at 300
C to reduce the
heat up time and held there until their internal temperature was
within 5
C of the target value, with the time required being usu-
ally less than 1 min. At that point, the tubes were quickly moved
to a second sand bath that was controlled at the target tempera-
ture, with this time being defined as time zero. At various time
intervals, one of the tubes was removed from the sand bath and
dropped into cool water to stop the reaction. When the tempera-
tures had dropped to ambient levels, the tubes were opened, and
the tube contents were filtered through filter paper of known
weight. The filtrates were added to centrifugal tips for further
preparation of samples for HPLC analysis. The tubes were
washed three times with deionized water, and the wash water Figure 3. Maximum LA yields at three initial cellulose
was passed through filter paper to trap as much of the solids concentrations with other conditions the
from the reaction as possible. The filter paper was then dried at same as for Figure 1.

238 DOI 10.1002/aic Published on behalf of the AIChE January 2012 Vol. 58, No. 1 AIChE Journal
Figure 4. Fitted curves of concentration vs. time for three initial cellulose concentrations: top left, 49.8; top right,
99.6; and bottom left, 149 mM.
Experimental data (symbols) and kinetic model (lines) concentrations for cellulose decomposition to LA: cellulose (l,), glucose (þ, ),
HMF (&, .), and LA (^, ). Conditions: 49.8 mM cellulose concentration, 200
C, and 0.206 M HCl concentration.

5 min through a 0.2 lm memberane filter. The supernatants LA concentration ðg=lÞ  glu cos e MW
were used for HPLC analysis. LA yield ¼
Initial glu cos e concentration  LA MW
The overall reactions from glucose to LA and from glucan LA concentration ðg=lÞ
in cellulose to LA can be, respectively, expressed as: ¼ % (15)
Initial glu cos e concentration  0:664
C6 H12 O6 ! C5 H8 O3 þ HCOOH þ H2 O (13)

C6 H10 O5 ! C5 H8 O3 þ HCOOH (14) FA concentration ðg=lÞ  glu cos e MW

FA yield ¼
Initial glu cos e concentration  FA MW
The average molecular weight of glucan, a unit of cellulose, is
162, and because glucose has a molecular weight of 180 and FA concentration ðg=lÞ
¼ (16)
LA yields reported in the literature5 are based on glucose, LA Initial glu cos e concentration  0:236
and FA concentrations (g/l) measured from HPLC were
converted to LA and FA yields as a percent of the theoretical Cellulose contents in the residues were analyzed by converting
maximum as follows: the cellulose into glucose using the NREL-recommended

Table 1. Kinetic Model First-Order Rate Constants at Three Initial Cellulose Concentrations
Cellulose
Concen. (mM) k10 (l/(mol min) k20 (l/mol min) k30 (l/mol min) k40 (l/mol min) k50 (l/mol min) k60 (l/mol min) Cwp
49.8 0.834  0.010 2.23  0.15 0 7.91  0.75 6.94  1.47 0.0200  0.0033 2.04E-6
99.6 0.602  0.033 3.17  0.40 0 6.16  0.43 5.45  0.83 0.0218  0.0068 1.47E-6
149.1 0.823  0.059 3.25  0.11 0 16.6  0.92 13.9  1.81 0.0290  0.0061 2.01E-6
Average 0.753  0.034 2.89  0.22 0 10.2  0.7 8.76  1.37 0.0236  0.0054 1.84E-6

AIChE Journal January 2012 Vol. 58, No. 1 Published on behalf of the AIChE DOI 10.1002/aic 239
Figure 5. Smoothed curves of yields vs. time for three initial cellulose concentrations using the average rate con-
stants in Table 1: top left, 49.8; top right, 99.6; and bottom left, 149 mM.
Experimental data (symbols) and kinetic model yields (lines) during cellulose decomposition: cellulose (l,), glucose (þ, ), HMF
(&, .), and LA (^, ). Conditions are the same as for Figure 4.

method.13 Briefly, about 0.3 g of residues was added to 3 ml of for 1 h. After cooling, the slurry was filtered through a
72 wt % sulfuric acid, and the suspension was hydrolyzed at crucible. About 2 ml of filtrate were neutralized by calcium
50
C for 1 h. Then, the slurry was diluted to 4 wt % sulfuric carbonate to a pH 5–6, and then the suspensions were
acid by adding 84 ml deionized water and autoclaved at 121
C centrifuged at 15,000 rpm for 5 min through the 0.2 lm
membrane filter. The supernatants were used for HPLC
analysis.

Figure 6. Experimental LA concentrations vs. time for

three temperatures: 160
C (&), 180
C ^, and
200
C (D).
Conditions: 99.6 mM initial cellulose concentration, and Figure 7. Maximum LA yields at three temperatures with
0.927 M HCl concentration. other conditions the same as for Figure 6.

240 DOI 10.1002/aic Published on behalf of the AIChE January 2012 Vol. 58, No. 1 AIChE Journal
Figure 8. Fitted curves of concentrations vs. time for three temperatures: top left 160
C, top right 180
C, and bot-
tom left 200
C.
Experimental data (symbols) and kinetic model concentrations (lines) results for cellulose decomposition to LA: cellulose (l,), glucose
(þ, ), HMF (&, .), and LA (^, ). Conditions: 99.6 mM cellulose concentration and 0.927 M HCl concentration.

Results and Discussion cules. The two concentration peaks observed for glucose and
Effect of initial cellulose concentration HMF resulted from the fact that both are intermediates in the
on LA production series of reactions from cellulose to LA and FA (Figure 2).
Figure 3 reveals that increasing the initial cellulose concentra-
Kinetic experiments were conducted at the following con-
tion decreased the maximum LA yield slightly from 51.5% to
ditions: 200
C, acid concentration of 0.206 M, and cellulose
48.4% (a 5.8% decrease). Although large decreases in LA
concentrations of 49.8 mM, 99.6 mM, and 149 mM. Plots of
yield have been reported as the initial cellulose concentration
concentrations and yields for LA, FA, glucose, and HMF vs.
increased, our smaller drop in yields may be due to our lower
time are shown in Figures 1–3, respectively. Thus, we see
that increasing the initial cellulose concentration from 49.8 cellulose concentrations (49.8 to 149 mM) than those (105 to
mM to 149 mM increased the LA concentration from 24.6 864 mM) of Girisuta et al.’s experiments.5
mM to 69.3 mM (Figure 1). Furthermore, the LA concentra- A MATLAB program was used to fit the parameters in Eqs.
tion increased by almost the same multiple (2.82 times) as the 8 to 11 to best match the experimental concentrations of cellu-
cellulose concentration (2.99 times). With increasing reaction lose, glucose, HMF, and LA, with the resulting curves pre-
time, LA and FA concentrations increased rapidly initially, sented in Figure 4 and the corresponding rate constants k0i
reached the maximum values, and dropped somewhat at the reported in Table 1. These fitted curves facilitate following
end reaction period. Figure 1 also shows that the concentra- trends in reactant, intermediate, and product concentrations
tions of LA and FA were almost the same because glucose over the reaction time. The model nearly perfectly followed
decomposition produced equal numbers of LA and FA mole- the LA concentration data over time, but deviations were found

Table 2. Kinetic Model First-Order Rate Constants at Three Temperatures

Temperature (
C) k10 (l/mol min) k20 (l/mol min) k30 (l/mol min) k40 (l/mol min) k50 (l/mol min) k60 (l/mol min)
160 0.0665  0.0033 0.0845  0.020 0 0.271  0.057 0.187  0.041 0.00296  0.00077
180 0.251  0.007 0.464  0.057 0 1.63  0.16 0.856  0.206 0.00648  0.00160
200 0.615  0.005 2.10  0.09 0 7.87  0.13 5.92  0.26 0.0118  0.0007
Ratio of ki,200/ki,160 9.25 24.9 N/A 29.0 31.7 3.99

AIChE Journal January 2012 Vol. 58, No. 1 Published on behalf of the AIChE DOI 10.1002/aic 241
Figure 9. Experimental points (symbols) and best fit Figure 10. Experimental LA concentrations vs. time for
lines for plot of ln of rate constants vs. three acid concentrations: 0.309 M (&),
inverse absolute temperature: k1 (&, 2), k2 0.618 M ^, and 0.927 M (D).
(^, 2), k4 (D, 2), k5 (x, 2), and k6 (1, 2). Conditions: 99.6 mM initial cellulose concentration and
200
C.
Conditions: 99.6 mM initial cellulose concentration, 0.927
M HCl concentration, and 473 K.
about 20 min at 180
C but only 3 min at 200
C. After 3 min
between the model predictions and experimental data for at 200
C, the LA concentration dropped, implying that the
glucose concentrations at initial cellulose concentrations of maximum LA concentration depended strongly on both temper-
49.8 mM (Figure 4 top left) and 149 mM (Figure 4 bottom ature and time. However, of greatest importance, the LA yield
shown in Figure 7 did not change significantly over the temper-
left). Because the rate constants k0i in Table 1 did not change
ature range of 180 to 200
C but held essentially constant at
monotonically with increasing initial cellulose concentra-
about 60%, with a possibly slightly higher value of 61% at
tion, the rate constants appeared to be independent of the 180
C for an acid concentration 0.927 M, an initial cellulose
initial cellulose concentration. On this basis, average values concentration 99.6 mM, and a reaction time 20 min.
(Table 1) of the rate constants k0i calculated for the three ini- By contrast, Girisuta et al.5 reported that the LA yield
tial cellulose concentrations were used in Eq. 12 to predict decreased from 61% at 150
C to 42% at 200
C. This differ-
yield changes with time, as shown for cellulose, glucose, ence might be explained by two factors: (1) our lower cellu-
HMF, and LA in Figure 5. In this case, the LA yield was lose concentrations reduced the effect of temperature on LA
predicted to be around 42% over the range of initial cellu- yield as intrinsic kinetics were approached due to less mass
lose concentrations for this model that did not include inter- transfer resistance and (2) hydrochloric acid is a stronger
nal and external mass transfer effects. catalyst than the sulfuric acid used by Girisuta et al.
Cellulose, glucose, HMF, and LA concentrations based on
the experimental data and kinetic model are shown in Figure 8,
Effect of temperature on LA concentration with the rate constants ki0 for the model listed in Table 2.
Figure 6 shows the concentrations of LA for a cellulose Although the LA data display linear behavior in Figure 8
concentration of 99.6 mM and an acid concentration of (top left) and nonlinear curves in Figure 8 (top right and bot-
0.926 M for temperatures of 160, 180, and 200
C. As tem- tom left), the model from Eqs. 8 through 11 follows both
perature was increased from 160 to 200
C, the maximum trends well. Unlike the nonmonotonic change in k0i at various
LA concentration increased from 36.9 to 67.1 mM. Although initial cellulose concentrations described in the last Section,
the LA concentration at 160
C was the lowest among the three all of the rate constants k0i increased with increasing tempera-
temperatures, the reaction was incomplete as shown in Figure 6 ture in the following order: k04 [k05[ k02 [ k01 [ k06. Thus,
with about 22 mM of glucose that could be further reacted to reaction 5 describing humins formation from HMF had the
higher LA concentrations left at the end of this experiment. highest rate constant except for reaction 4, implying that a
Because the rate of reaction increased with temperature, the very large portion of the solid residues/humins resulted from
time to achieve the maximum LA concentration dropped con- that reaction. Furthermore, the ratio k05200/k05160 for reaction 5
siderably with increasing temperature, as expected. For exam- 0 0
in Table 2 is largest (31.7) among all ratios of ki,200 /ki,160,
ple, the time to reach the maximum LA concentration was suggesting that humins formation from HMF was most

Table 3. Kinetic Model Pre-exponent Factors and Activation Energies

k10 (l/mol min) k20 (l/mol min) k30 (l/mol min) k40 (l/mol min) k50 (l/mol min) k60 (l/mol min)
Ai (l/mol.min) 2.36E10 2.80E15 N/A 5.62E16 8.57E16 3.95E4
Ai (l/mol.s) 3.93E8 4.67E13 N/A 9.37E14 14.3E14 658
Eai (J/mol) 9.56E4 13.7E4 N/A 14.4E4 14.7E4 5.90E4
R2i 0.992 1.000 N/A 1.000 0.991 0.996

242 DOI 10.1002/aic Published on behalf of the AIChE January 2012 Vol. 58, No. 1 AIChE Journal
 From this plot, the activation energies and pre-exponential
factors given in Table 3 were determined for each reaction
from the slopes and intercepts of the straight lines for the three
temperatures covered. Thus, we see that the activation
energies Ea1 and Ea6 for reactions 1 and 6 were much smaller
than those for reactions 2, 4, and 5, and Ea4 and Ea5 had almost
same values. As a result, increasing temperature would cause
approximately the same increase in reaction rates for reactions
4 and 5, resulting in little change in LA yields with
temperature. In parallel with this observation, the pre-
exponential factors A1 and A6 were much smaller than A4
and A5, and analyzing them in terms of reaction rate collision
theory gives14:

A ¼ pz (18)
Figure 11. Maximum LA yields at acid concentrations
with other conditions the same as for Figure 10. in which A is the pre-exponential factor in the Arrhenius
equation, z is the frequency factor, and p is the steric factor. Steric
factors for reactions between polyatomic molecules are less than
sensitive to temperature and has the largest activation energy those for reactions involving a single-atomic molecule.14 Against
among these reactions. If the rate constants follow the Arrhe- this, a polymer, cellulose, is the reactant in reaction 1, two
nius expression of Eq. 7, lnk0i should be a linear function of polyatomic molecules take part in reaction 6, and only one
1/T as shown in Eq. 17 and confirmed in Figure 9 for each polyatomic molecule is reacted in each of the other reactions.
rate constant: Therefore, the steric factors and pre-exponential factors of
reactions 1 and 6 would be expected to be smaller than for the
Ea
ln ki0 ¼ ln Ai  i¼16 (17) other reactions, consistent with the model predictions.
RT

Figure 12. Fitted curves of concentrations vs. time for reaction with three concentrations of hydrochloric acid: top
left 0.309, top right 0.618, and bottom left 0.927 M.
Experimental data (symbols) and kinetic model concentration (lines) for cellulose decomposition to LA: cellulose (l,), glucose (þ,
), HMF (&, .), and LA (^, ). Conditions: 99.6 mM cellulose concentration and 200
C.

AIChE Journal January 2012 Vol. 58, No. 1 Published on behalf of the AIChE DOI 10.1002/aic 243
 Table 4. Kinetic Model First-Order Rate Constants at Three Acid Concentrations
Acid Concen. (M) k10 (l/mol min) k20 ((l/mol min) k30 (l/mol min) k40 (l/mol min) k50 (l/mol min) k60 (l/mol min)
0.309 1.35  0.01 1.80  0.23 0 3.17  1.09 3.08  0.98 0  0.0016
0.618 0.984  0.034 3.81  0.58 0 10.0  0.4 8.76  0.83 0.00526  0.00030
0.927 0.615  0.005 2.10  0.09 0 7.87  0.13 5.92  0.25 0.0118  0.0007

Effect of hydrochloric acid concentration transfer and surface reaction controlled.15 The fact that the
on levunilic acid changes in rate constant with temperature followed the
The effect of HCl acid concentration on LA concentra- Arrhenius equation in Figure 9 demonstrated that acid-cata-
tions and yields are summarized in Figures 10 and 11 at the lyzed reaction of cellulose to glucose is not controlled by
following conditions: cellulose concentration of 99.6 mM, external mass transfer. The importance of internal mass
200
C, and acid concentrations of 0.309, 0.618, and 0.927 transfer limitations in heterogamous chemical kinetics can be
M. Thus, we see that increasing acid concentration caused a judged by the Weis-Prater (Eq. 19) criterion15:
slight increase in the maximum LA concentration and yield
from 52.3 mM and 54.8% at acid concentration 0.309 M to Cwp ¼ g/2 (19)
57.1 mM and 59.7% at acid concentration 0.927 M, respec- where Cwp is the Weise-Prater parameter (dimensionless) and
tively. This outcome is consistent with the role of hydrochlo- g is the effectiveness factor of internal mass transfer
ric acid as a catalyst not changing chemical equilibrium (dimensionless). For an irreversible first-order reaction g is
under ideal conditions even though the chemical reaction
rate would increase with increasing acid concentration. As a  
3 1 1
result, the maximum LA concentration and yield changed lit- g¼  (20)
/ tan h/ /
tle for the three acid concentrations applied. However,
increasing acid concentration reduced the time needed to where / is the Thiele modulus. For a first-order reaction in
attain the maximum LA concentration and yield: for the con- spherical particles, / is equal to
ditions tested, the time to the highest LA concentrations was
rffiffiffiffiffiffi
about 20 min with an acid concentration 0.309 M but only 6 R k1
min at 0.926 M. This result confirmed that acid accelerated /¼ (21)
3 De
the reactions. Moreover, LA concentrations and yields
dropped if the solution was held beyond the optimum time where k1 is the reaction rate constant (s1), De is the effective
due to by-product formation. diffusion coefficient (m/s), and R is the particle radius (0.025
Figure 12 presents cellulose, glucose, HMF, and LA con-  103 m for Avicel particles) (m). The diffusion coefficient
centrations vs. time based on the kinetic model, with the rate of protons in water at 298 K is 9.311  109 m2.16 The value
constants determined by the MATLAB program for each at 473 K was corrected with Eq. 2217:
reaction shown in Table 4. Thus, we can see that the rate
constants k01 and k06 in Table 4, respectively, decreased and De 298 l298 T473
De 473 ¼ (22)
increased monotonically with increasing acid concentration, T298 l473
but the rate constants k02, k04, and k05 did not.
Application of the Nelder-Mead approach in MATLAB to
the proposed model with the fitted constants in Table 2 pre-
dicted that a maximum LA yield of 59.2% would occur at a
time of 20.4 min and acid concentration of 0.899 M at
180
C (Figure 13). This predicted LA yield is somewhat less
than that determined experimentally (59.7% in Figure 11)
but is in the range of experimental error. Figure 13 also
shows that the LA yield is insensitive to both reaction time
and acid concentration: the LA yield changed by only about
0.2% when the reaction time and acid concentration were
increased from 20 min and 0.85 M to 21 min and 0.95 M,
respectively.

Identification of probable rate-limiting factors

The hydrolysis of acid-catalyzed cellulose to glucose is a
heterogamous reaction that may be controlled by external Figure 13. Simulated LA yield surface with respect to
and internal mass transfer and surface reaction. The relation- reaction time and acid concentration apply-
ship between the observed process rate and temperature can ing the model with the fitted constants in
be used to identify the rate-limiting step. The rate would Table 2.
change linearly with temperature if external mass transfer Conditions: cellulose concentration 99.6 mM and tempera-
were limiting and change exponentially if internal mass ture 180
C.

244 DOI 10.1002/aic Published on behalf of the AIChE January 2012 Vol. 58, No. 1 AIChE Journal
Figure 14. Reaction rates vs. time from the kinetic models for cellulose decomposition to LA at three tempera-
tures: top left 160
C, top right 180
C, and bottom left 200
C.
Cellulose (), glucose (), HMF (.), and LA (). Conditions: 99.6 mM cellulose concentration and 0.927 M HCl concentration.

where l is the water viscosity at T (Pa s) (9  104 for 298 K but remained positive. Initially, glucose formation controlled
and 1.36  104 Pa s for 473 K). Applying these values to the rate of the entire series of reactions because it had the
Eqs. 19–22, the Weise-Prater parameters are much less than 1 smallest reaction rate among all the absolute reaction rates.
for reaction 1 at the three initial cellulose concentrations in However, as shown in the figures, HMF formation soon had
Table 1. Therefore, the internal mass transfer is predicted to be the smallest absolute reaction rate, with the result that the reac-
negligible. tion from glucose to HMF became rate controlling in the series
of reactions. This change in the rate-controlling step was de-
pendent on reaction temperature, with the transition occurring
Rate-controlling step in the reaction of cellulose to LA
earlier at higher temperatures. For example, this shift occurred
Equations 3–6 are rate expressions for consumption of cel- after about 5 min at 160
C but less than 1 min at 200
C.
lulose and formation of glucose, HMF, and LA. After the
rate constants in these equations were determined by the
MATLAB program, the reaction rates were calculated to Conclusions
determine the rate-controlling step. The rates of cellulose Data were developed on the effects of initial cellulose
decomposition at 160, 180, and 200
C shown in Figure 14 concentration, hydrochloric acid concentration, and tempera-
are always negative because it is consumed. On the other ture on LA concentrations and yields from cellulose in a
hand, the net reaction rate to LA changed from positive to batch system. The highest LA yield of about 60% was
negative as LA was initially produced and then consumed, observed at an initial cellulose concentration of 99.6 mM, an
whereas the reaction rates of glucose and HMF quickly acid concentration of 0.927 M, and 180–200
C in our
increased before decreasing (except for Figure 14 top left) experiments. Higher acid concentrations and temperatures

AIChE Journal January 2012 Vol. 58, No. 1 Published on behalf of the AIChE DOI 10.1002/aic 245
accelerated the reaction rate and allowed the maximum LA 2. Girisuta B, Janssen LPBM, Heeres HJ. A kinetic study on the con-
yield to be reached sooner, but the maximum LA yield var- version of glucose to levulinic acid. Chem Eng Res Des. 2006;84:
339–349.
ied little with changing acid concentration and temperature. 3. Jing Q, Lu XY. Kinetics of non-catalyzed decomposition of glucose
This outcome is consistent with acid acting as a catalyst to in high-temperature liquid water. Chin J Chem Eng. 2008;16:890–
accelerate the reaction rate but not change chemical equilib- 894.
rium. The maximum LA concentration was approximately 4. Seri K, Sakaki T, Shibata M, Inoue Y, Ishida H. Lanthanum(III)-cat-
alyzed degradation of cellulose at 250 degrees C. Bioresour Technol.
proportional to the initial cellulose concentration, showing
2002;81:257–260.
that cellulose concentration had little impact on yields. 5. Girisuta B, Janssen LPBM, Heeres HJ. Kinetic study on the acid-
A kinetic model that included a series of pseudo-first catalyzed hydrolysis of cellulose to levulinic acid. Indus Eng Chem
order reactions producing LA and FA from cellulose through Res. 2007;46:1696–1708.
glucose and HMF as well as parallel reactions for humins 6. Girisuta B, Janssen LPBM, Heeres HJ. A kinetic study on the
decomposition of 5-hydroxymethylfurfural into levulinic acid. Green
formation successfully replicated the time history of concen- Chem. 2006;8:701–709.
trations and yields for cellulose, glucose, HMF, and LA. All 7. Cha JY, Hanna MA. Levulinic acid production based on extrusion
of the rate constants increased with increasing temperature, and pressurized batch reaction. Ind Crops Prod. 2002;16:109–118.
in a manner consistent with Arrhenius dependence. The fact 8. Fang Q, Hanna MA. Experimental studies for levulinic acid produc-
that LA formation (reaction 4) from HMF had the same acti- tion from whole kernel grain sorghum. Bioresour Technol. 2002;81:
187–192.
vation energy as humins formation (reaction 5) from HMF 9. Girisuta B, Danon B, Manurung R, Janssen LPBM, Heeres HJ.
suggests that LA yield is less affected by temperature, con- Experimental and kinetic modelling studies on the acid-catalysed hy-
sistent with our experimental data. Over most of the reaction drolysis of the water hyacinth plant to levulinic acid. Bioresour
time, HMF formation was the slowest step in the series of Technol. 2008;99:8367–8375.
10. Yan LF, Yang NK, Pang H, Liao B. Production of levulinic acid
reactions and controlled overall reaction progress. from bagasse and paddy straw by liquefaction in the presence of
hydrochloride acid. Clean-Soil Air Water 2008;36:158–163.
11. Chang C, Cen PL, Ma XJ. Levulinic acid production from wheat
Acknowledgments straw. Bioresour Technol. 2007;98:1448–1453.
12. Saeman JF. Kinetics of wood saccharification hydrolysis of cellulose
The authors thank the Defense Advanced Research Projects Agency and decomposition of sugars in dilute acid at high temperature. Ind
(DARPA) and University of Massachusetts for providing financial sup- Eng Chem. 1945;37:43.
port (contract No. 09-005334-000) for this project. They also appreciate 13. Sluiter A, Hames B, Ruiz R, Scarlata C, Sluiter J, Templeton D,
the Center for Environmental Research and Technology of the Bourns Crocker D. Determination of structural carbohydrates and lignin in
College of Engineering (CE-CERT) at the University of California, Riv- biomass. Lab Anal Proced. (LAP) 2008; 4/25.
erside (UCR) for providing key equipment and facilities. The second 14. Upadhyay SK. Chemical Kinetics and Reaction Dynamics. New
author is grateful to the Ford Motor Company for funding the Chair in York: Springer, 2006:93–99.
Environmental Engineering at the Center for Environmental Research 15. Fogler SH. Elements of Chemical Reaction Engineering, 4th ed.
and Technology of the Bourns College of Engineering at UCR that aug- Upper Saddle River, NJ: Prentice Hall, 2005:838–849.
ments support for many projects such as this.
16. CRC Handbook of Chemistry and Physics, 74th ed. In: Section 5
Thermochemistry, Electrochemistry, and Kinetics-90. Lide DR,
editor. Boca Raton, FL: CRC press, Inc. 2000:5–90.
Literature Cited 17. Poling BE, Prausnitz JM, O’Connell JP. In: Lide DR, editor.
1. Bozell JJ, Moens L, Elliott DC, Wang Y, Neuenscwander GG, Fitz- Properties of Gases and Liquids, 5th ed. New York: McGraw-Hill,
patrick SW, Bilski RJ, Jarnefeld JL. Production of levulinic acid and 2001:11.39.
use as a platform chemical for derived products. Resour Conserv
Recyl. 2000;28:227–239. Manuscript received July 11, 2010, and revision received Dec. 20, 2010.

246 DOI 10.1002/aic Published on behalf of the AIChE January 2012 Vol. 58, No. 1 AIChE Journal