Académique Documents
Professionnel Documents
Culture Documents
pubs.acs.org/IECR
ABSTRACT: The heterogeneously catalyzed transesterification reaction for the production of biodiesel from triglycerides was
investigated for reaction mechanism and kinetic constants. Three elementary reaction mechanisms Eley-Rideal (ER), Langmuir-
Hinshelwood-Hougen-Watson (LHHW), and Hattori with assumptions, such as quasi-steady-state conditions for the surface species
and methanol adsorption, and surface reactions as the rate-determining steps were applied to predict the catalyst surface coverage and the
bulk concentration using a multiscale simulation framework. The rate expression based on methanol adsorption as the rate limiting in
LHHW elementary mechanism has been found to be statistically the most reliable representation of the experimental data using
hydrotalcite catalyst with different formulations.
r 2011 American Chemical Society 4818 dx.doi.org/10.1021/ie101403f | Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE
catalytic oxidation of CO on Pt. By adapting these modeling tools and appropriate mechanism that may be valid for the whole
insights, this work aims to integrate catalytic surface kinetic Monte range of formulations or suggest the best mechanism for
Carlo (KMC)23 and bulk scale mean field (MF) simulations24 to individual formulations.
validate the kinetic parameters estimated using genetic algorithm Determination of Kinetic Rate Constants Using GA. The
(GA) based optimization approach.25,26 The kinetic parameters steps for the estimation of the rate constants by applying GA-
obtained correspond to the appropriate match between simulation based optimization are as follows.
and experimental results of evolution in bulk concentrations. 1. The initial concentrations in the batch reactor (Ci(0)), size
The overall modeling strategy is discussed in the next section. of the reactor, and batch time are specified (Appendix A).
Estimation of kinetic rate constants using GA based optimization The bounds for the rate constants kj are also provided as the
methodology, followed by their validation using multiscale KMC/ inputs to the GA (Appendix B).
MF simulation framework is outlined. Thereafter, ER, LHHW, 2. The rate constants are decision variables and their initial
and Hattori mechanisms alongside the derivations of the kinetic values are guessed by GA using random number generator
rate expressions are illustrated. The results of various mechanisms in between their respective bounds.
are quantitatively analyzed and compared for the selection of the 3. The bulk concentration (Ci) of specie i is a function of
most appropriate mechanism that may be valid for the whole range kinetic rates (Ri), eq 4.
of formulations. Simultaneously, the analysis may also suggest the
dCi
best mechanism for individual formulations. ¼ Ri "i∈T, D, M, G, MoOl, CH3 OH ð4Þ
dt
’ METHODOLOGY 4. To account for the nonideality of a mixture as in here, the
The overall strategy for an estimation of kinetic rate constants correlation between the activity of species ([i]) and its
based on a reaction mechanism is illustrated as follows (Figure 1). concentration Ci in eq 5 was applied. The UNIFAC
1. The rate parameters involved in an assumed mechanism contribution method27 outlined in Appendix C was used
were estimated using GA based optimization. to calculate the activity coefficients, γi. Table 1 exemplifies
2. Using the rate constants obtained from step 1, the distribution typical values of activity coefficients of species.
of species on the surface of a catalyst formulation as well as the
changes in bulk specie concentrations were predicted by a ½i ¼ Ci γi "i∈T, D, M, G, MeOl, CH3 OH ð5Þ
multiscale KMC/MF simulation framework implemented.
This framework was used to simultaneously capture the effect 5. The set of ordinary differential equations (eq 4) was solved
of surface adsorption-reaction-desorption on the bulk specie by ode45 solver in MATLAB with a time gap of 1 s. The
concentrations. An iteration of rate constants between the GA- bulk concentrations profiles of species in the reactor is
based optimization step, and the multiscale KMC/MF simula- obtained as a function of time.
tion step may be involved, until the best fit of concentration 6. The GA optimization, based on the works of Bhat et al.25
profiles against experimental results is obtained. and Xu et al.,26,28 implemented in MATLAB (Appendix B)
3. The mechanism is applied to various catalyst formulations. was then applied to minimize the residual sum of square
4. New mechanisms and assumptions were then considered for (RSSQ) of errors between the experimentally observed
the parametric prediction using the above two frameworks, GA- and the model predicted concentrations of species (i = 1 to
based optimization and multiscale KMC/MF simulation, until nspc), at subsequent time points (j = 1 to ntime), in eq 6,
all three mechanisms with given assumptions are investigated. by adjusting the rate constants within their specified ranges.
5. Statistical reliance and comparison between mechanisms It is assumed that the rate of reaction for adsorp-
were performed. This analysis may propose the most tion of methanol is a low value in the range of 10-1, while
4819 dx.doi.org/10.1021/ie101403f |Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE
Hattori mechanism also considers the adsorption of triglyceride Table 4. Elementary Reactions in LHHW Mechanism
on the surface of the catalyst as a rate determining step. Hattori
elementary reactions equilibrium reaction constants
mechanism differs from LHHW mechanism, where in the formation
of intermediate species from the reactions between adsorbed þ CH3 OH S CH3 OH ½CH3 OH
methanol and adsorbed triglyceride, diglyceride and monoglyceride K1 ¼
½ ½CH3 OH
is considered. The adsorbed methanol and triglyceride react to form
adsorbed intermediate (TsCH3OH*) and an empty site, respec- T þ S T ½T
tively. Subsequently, the adsorbed intermediate (TsCH3OH*) K2 ¼
½ ½T
decomposes into the production of adsorbed diglyceride and bulk
methyl oleate, respectively. Adsorbed diglyceride, monoglyceride CH3 OH þ T S D þ MeOl ½D½MeOl
and glycerol thereafter desorb from the catalyst surface into the bulk K3 ¼
½CH3 OH½T
phase. TsCH3OH*, DsCH3OH*, and MsCH3OH* represent the
intermediates from the reactions between adsorbed methanol and CH3 OH þ D S M þ MeOl ½M½MeOl
adsorbed T, D, and M, respectively. The elementary reaction K4 ¼
½CH3 OH½D
expressions in Table 6 result in the kinetic rate expressions in Table 7.
kj indicates the forward reaction kinetic rate constants for 11 CH3 OH þ M S G þ MeOl ½G½MeOl
rate determining steps, and Kj represents the equilibrium con- K5 ¼
½CH3 OH½M
stant of reaction j, respectively, in Table 6.
• ER quasi-steady-state
• ER methanol adsorption as rate determining
• LHHW surface reaction as rate limiting The statistical significance of a mechanism is achieved by Chi
• LHHW methanol adsorption as rate limiting square, χ2, test. The results of kinetic parameters and Chi square test
• Hattori methanol adsorption as rate determining for statistical significance are presented in Table 8. It is calculated as
4821 dx.doi.org/10.1021/ie101403f |Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE
the ratio of the residual sum of the square of the errors (RSSQ) sorbed diglyceride (þ), 97.1% in case b Figure 5. A fraction of the
between the predicted and the experimental values for the concen- adsorbed diglyceride is subsequently converted into adsorbed
trations of the species (i = 1 to nspc), at time points (j = 1 to ntime), monoglyceride (ω) and finally into glycerol (O), 25.3% of each
and the experimental values at given data points, in eq 7.28 in case c Figure 5 over the duration of the reaction (10 800 s).
The kinetic parameters (Table 8) thus obtained by following
time X
nX nspc experimental 2 the strategy in Figure 1 for ER quasi-steady-state reaction mech-
ðCi ðjÞ - Ci ðjÞÞ anism represents the experimental observation for M2.93Al hydro-
χ2 ¼ experimental
ð7Þ
j¼1 i¼1 Ci ðjÞ talcite catalyst adequately. These frameworks were further applied to
the other hydrotalcite catalysts with different concentrations of Mg
and Al (Appendix A).
The simulation results of ER quasi-steady-state mechanism Results of Application of ER Quasi-Steady-State Mech-
(Table 3) on Mg2.93Al hydrotalcite catalyst are illustrated in anism for All Hydrotalcite Catalysts. The RSSQ and χ2 obtained
detail, while other system results are only summarized. for Mg1.82Al hydrotalcite catalyst are 0.006 and 0.0218, respec-
Results of Application of ER Quasi-Steady-State Mechan- tively, indicating statistically good representation of the experi-
ism to Hydrotalcite Catalyst Mg2.93Al. The results of kinetic mental data for Mg1.82Al hydrotalcite catalyst. Table 8 demon-
parameters (k1-7) in ER quasi steady state mechanism (Table 3) on strates statistically acceptable χ2 values of 0.0174, 0.0218, and
Mg2.93Al hydrotalcite catalyst using GA and multiscale KMC/MF 0.0044 for the hydrotalcite catalysts, Mg0.82Al, Mg1.82Al, and
simulation framework are provided in Figures 3-5, respectively. Mg2.93Al, respectively. However, the results for Mg1.38Al indicate
The RSSQ resulted (0.008) is higher than the specified for GA a higher value of χ2 (0.2611), revealing inconsistency in predict-
optimization. However, the χ2 value of 0.0044 in Table 8 is smaller ing the ER quasi-steady-state model for all four catalysts.
than 0.01 required for 99.5% level of significance (confidence) with ER Elementary Reaction Mechanism (Methanol Adsorp-
7 degrees of freedom, in chi square test for statistical significance,29 tion as Rate Limiting). From the illustration earlier, it was rational
implying higher level of significance/applicability of the mechanism. to assume methanol adsorption as a rate determining step in the ER
The multiscale simulation framework results into the time mechanism. The resulting mechanism was further simplified by the
evolution of surface species, based on which the quasi-steady hypo- assumption of equal rate constants of all the adsorption equilibrium
thesis assumed at the first place can be validated. The results of the steps, based on the work of Dossin et al.2 The equilibrium constant for
KMC/MF simulation presented in Figure 4 indicate a rapid initial the adsorption of methanol (Keq) was assumed to be constant and
change in the concentration of the species on the surface. However, equal to unity. The kinetic rate expression was reduced to two
subsequently steady-state surface coverage fraction of 0.05 and parameters kf (rate of forward reaction for adsorption of methanol),
0.0003 were attained by the adsorbed species M* and CH3OH*, and KA (adsorption equilibrium constant for diglyceride, monogly-
respectively, after 1000 s, while D* and G* eventually reached to ceride, glycerol, and methyl oleate) (eqs 8-10). Table 8 illustrates the
their steady states at 0.09 and 0.84, respectively, after 10 000s χ2 of 0.028, 0.03, 0.1722, and 0.013 for Mg0.82Al, Mg1.38Al, Mg1.82Al,
(Figure 4). Hence, a steady state equilibrium attained by surface and Mg2.93Al, respectively, which is much less than 1.72 required for
species reinforces the assumption on their quasi-steady-state. individual catalysts for the two parameter kinetic model with 9 degrees
Figure 5 shows the evolution of catalyst surface resulted from of freedom at 99.5% level of fit.29 The degree of freedom is the
KMC simulations. Initially the catalyst surface lattice is made up number of experimental points - the number of kinetic parameters - 1.
of empty sites ( 3 ) (Figure 5 case a). These lattice sites are con- Hence the kinetic model based on ER mechanism with methanol
verted into adsorbed methanol (∼). However, the reaction of adsorption as rate limiting, can be applied to all four hydrotalcite
methanol with high concentration of triglyceride in bulk phase is catalysts. The χ2 value certainly improves for Mg1.38Al case from the
fast. Hence most of the adsorbed methanol is converted into ad- quasi steady state assumption.
½D½MeOl
kf ½CH3 OH - 1=Keq
½T
rT ¼ ! ð8Þ
KA ½D½MeOl
1þ þ KA ½MeOl þ KA ½T þ KA ½D þ KA ½M þ KA ½G
Keq ½T
½M½MeOl
kf ½CH3 OH - 1=Keq
½D
rD ¼ ! ð9Þ
KA ½M½MeOl
1þ þ KA ½MeOl þ KA ½T þ KA ½D þ KA ½M þ KA ½G
Keq ½D
½G½MeOl
kf ½CH3 OH - 1=Keq
½M
rM ¼ ! ð10Þ
KA ½G½MeOl
1þ þ KA ½MeOl þ KA ½T þ KA ½D þ KA ½M þ KA ½G
Keq ½M
Table 5. Elementary Reaction Rate Expressions for LHHW Mechanism with Surface Reaction As Rate Limiting
rate determining reaction overall reaction rate
1 K6 K7 ½D½MeOl
k1 ½CH3 OH -
þ CH3 OH S CH3 OH K1 K3 K2 ½T
r1 ¼
K6 K7 ½D½MeOl
1þ þ K6 ½MeOl þ K2 ½T þ K7 ½D þ K8 ½M þ K9 ½G
K3 K2 ½T
1 K6 K7 ½D½MeOl
k2 ½T -
T þ S T K2 K3 K1 ½CH3 OH
r2 ¼
K6 K7 ½D½MeOl
1 þ K1 ½CH3 OH þ K6 ½MeOl þ þ K7 ½D þ K8 ½M þ K9 ½G
K3 K1 ½CH3 OH
1
k3 K1 K2 ½CH3 OH½T - K7 K6 ½D½MeOl
CH3 OH þ T S D þ MeOl K3
r3 ¼
ð1 þ K1 ½CH3 OH þ K6 ½MeOl þ K2 ½T þ K7 ½D þ K8 ½M þ K9 ½GÞ2
1
k4 K1 K7 ½CH3 OH½D - K8 K6 ½M½MeOl
CH3 OH þ D S M þ MeOl K4
r4 ¼
ð1 þ K1 ½CH3 OH þ K6 ½MeOl þ K2 ½T þ K7 ½D þ K8 ½M þ K9 ½GÞ2
1
k5 K1 K8 ½CH3 OH½M - K9 K6 ½G½MeOl
CH3 OH þ M S G þ MeOl K5
r5 ¼
ð1 þ K1 ½CH3 OH þ K6 ½MeOl þ K2 ½T þ K7 ½D þ K8 ½M þ K9 ½GÞ2
K3 K2 K1 ½T½CH3 OH
k6 - K6 ½MeOl
K7 ½D
MeOl þ S MeOl r6 ¼
K3 K2 K1 ½T½CH3 OH
1 þ K1 ½CH3 OH þ þ K2 ½T þ K7 ½D þ K8 ½M þ K9 ½G
K7 ½D
K3 K2 K1 ½T½CH3 OH
k7 - K7 ½D
K6 ½MeOl
D þ S D r7 ¼
K3 K2 K1 ½T½CH3 OH
1 þ K1 ½CH3 OH þ K6 ½MeOl þ K2 ½T þ þ K8 ½M þ K9 ½G
K6 ½MeOl
K4 K1 K7 ½D½CH3 OH
k8 - K8 ½M
K6 ½MeOl
M þ S M r8 ¼
K4 K1 K7 ½D½CH3 OH
1 þ K1 ½CH3 OH þ K6 ½MeOl þ K2 ½T þ K7 ½D þ þ K9 ½G
K6 ½MeOl
K5 K8 K1 ½M½CH3 OH
k9 - K9 ½G
G þ S G K6 ½MeOl
r9 ¼
K5 K8 K1 ½M½CH3 OH
1 þ K1 ½CH3 OH þ K6 ½MeOl þ K2 ½T þ K7 ½D þ K8 ½M þ
K6 ½MeOl
The lowest rate constant corresponds to methanol adsorption LHHW Elementary Reaction Mechanism (Surface Reaction
(kf) with values of 1.6 10-6, 1.71 10-6, 7.33 10-6, and as Rate Limiting). The kinetic rate constants for LHHW
7.12 10-6 for the four hydrotalcite catalysts, with increasing Mg mechanism comprising of elementary steps in Table 4 are shown
concentration, respectively, revalidating the assumption (Table 8). in Table 8. As can be seen from the results, the equilibrium reac-
However, ER mechanism assumes no adsorption for triglyceride, tion rate constant between adsorbed methanol and adsorbed
diglyceride, and monoglyceride in the bulk to react with adsorbed diglyceride (K4), is the fastest with the values of 0.37, 0.48, and
methanol. Triglyceride, diglyceride, and monoglyceride are large 0.44 for Mg0.81Al, Mg1.38Al, and Mg2.93Al respectively (in the
molecules, and therefore, their adsorption onto the catalyst case of Mg1.82Al reaction between adsorbed methanol and ad-
surface would be difficult and hence can be regarded as rate sorbed triglyceride (K3) is found to be the fastest with a value of
determining steps, such as in LHHW and Hattori mechanisms 0.15). It is consistently identified that K1, the equilibrium con-
discussed as follows. stant of adsorption of methanol is the slowest with 5.97 10-5,
4823 dx.doi.org/10.1021/ie101403f |Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE
Table 6. Elementary Reactions in Hattori Mechanism tively, is less than 1.34 required for 3-parameter models with 8
degrees of freedom for 99.5% level of significance.29 Thus,
elementary reactions equilibrium reaction constants
LHHW mechanism with methanol adsorption as the rate deter-
þ CH3 OH S CH3 OH ½CH3 OH mining step represents the given experimental observations in
K1 ¼
½ ½CH3 OH Appendix A, adequately.
Hattori Elementary Reaction Mechanism (Methanol Ad-
T þ S T ½T
K2 ¼ sorption as Rate Limiting). As discussed in the previous sec-
½ ½T
tion the formation of intermediates is an important element in
CH3 OH þ T S ½TsCH3 OH þ ½½TsCH3 OH the Hattori elementary reaction model (Table 6) that differenti-
K3 ¼
½CH3 OH½T ates it from the LHHW and ER mechanisms. The kinetic rate
constants in the Hattori mechanism with methanol adsorption as
½TsCH3 OH S D þ MeOl ½D½MeOl the rate determining step are shown in the Table 8. The slowest
K4 ¼
½TsCH3 OH reaction rate is the adsorption of methanol (K1) as expected, with
the values of 4.45 10-3, 5.09 10-3, 8.11 10-3, and 0.014
CH3 OH þ D S ½DsCH3 OH þ ½½DsCH3 OH
K5 ¼ for Mg0.81Al, Mg1.38Al, Mg1.82Al, and Mg2.93Al, respectively. The
½CH3 OH½D
fastest reaction rates are K3, corresponding to the reaction be-
½M½MeOl tween adsorbed methanol and adsorbed triglyceride, 0.16 and
½DsCH3 OH S M þ MeOl K6 ¼
½DsCH3 OH 0.58 for Mg0.81Al and Mg2.93Al and K4, the decomposition of
intermediate species [TsCH3OH*] (Table 3), 0.29 and 0.32 for
CH3 OH þ M S ½MsCH3 OH þ ½½MsCH3 OH Mg1.38Al and Mg1.82Al, respectively. Table 8 indicates χ2 of 0.05,
K7 ¼
½CH3 OH½M 0.14, 0.30, and 0.009 for the four hydrotalcite catalysts with in-
creasing Mg concentration, respectively. However, since the
½MsCH3 OH S G þ MeOl ½G½MeOl number of experimental points for this mechanism is the same
K8 ¼
½MsCH3 OH as the number of kinetic parameters, statistical significance cri-
teria cannot be applied to this model.
D þ S D ½D Comparison between Kinetic Models. In the previous
K9 ¼
½ ½D section, the kinetic rate constants and the model fit χ2 with the
½M
assumptions of quasi-steady-state, methanol adsorption, and sur-
M þ S M K10 ¼ face reaction as rate limiting steps in ER, LHHW, and Hattori
½ ½M
elementary reaction mechanisms, are determined in Table 8. The
G þ S G ½G increase in the number of parameters increases the complexity of
K11 ¼ the model and reduces the limit on χ2 for the statistical signi-
½ ½G
ficance of the fit.29 Hence, different models based on the statis-
tical criterion PAB as defined below29 are further compared.
0.0001, 3.3 10-5, and 0.0001 for all four hydrotalcite catalysts,
respectively (Table 8). χ2 of 0.0061, 0.0203, 0.0092, and 0.0129 χA 2 =ðN - nA Þ
PAB ¼ ð11Þ
are predicted for Mg0.81Al, Mg1.38Al, Mg1.82Al, and Mg2.93Al, χB 2 =ðN - nB Þ
respectively, which are larger than 0.00 required for ten para-
meter kinetic model with 1 degree of freedom at 99.5% level of Here, χA2 and χB2 are the Chi-square, while nA and nB are the
fit.29 Hence, the kinetic model based on LHHW surface reac- number of parameters, for model A and B, respectively, and N is
tion rate limiting cannot be applied to any of the four hydrotalcite the number of experimental points. If PAB is lesser than one,
catalysts at 99.5% level of significance. model A fits the data better than model B and vice versa. The
LHHW Elementary Reaction Mechanism (Methanol Ad- number of parameters for a model fit should be less than the
sorption as Rate Limiting). The increase in the number of number of experimental points by at least one. Overall reaction
kinetic rate constants although increases the degrees of freedom rate determined for Hattori elementary reaction with methanol
but reduces statistical reliability because of lighter fitting with adsorption as the rate determining step has 12 parameters, which
experimental data. Dossin et al.2 developed a rate mechanism are the same as the number of experimental points. Hence, this
based on ER mechanism and adsorption of methanol as the rate model cannot be used for statistical analysis of this set of experi-
determining step. The LHHW elementary mechanism with me- mental data.
thanol adsorption as a rate limiting step involving three param- A comparison of performance using the statistical criterion
eters, the rate coefficient for methanol adsorption kMeOH, the PAB between the kinetic models on the four hydrotalcite catalysts
adsorption equilibrium coefficient of the overall transesterifica- is shown in Table 9. The least number of parameters (in this case 3)
tion reaction Keq, and the adsorption equilibrium constant of the are involved in LHHW and ER mechanisms with methanol adsorp-
alcohols KA, is thus statistically more reliable than the mechanism tion as the rate determining steps. Hence, these two mechanisms
with the assumption of surface reaction as the rate limiting step. would be most recommended ones from statistical reliability point
The kinetic rate constants obtained are shown in Table 8. of view. At the same time, simpler, but adequately detailed reaction
Similar to all other mechanisms, adsorption of methanol is kinetics model can be integrated to multiscale reactor simulation
identified as the slowest and the rate determining step with the frameworks.24 By comparing the PAB values in Table 9, the follow-
rate constants of 9.8 10-3, 1.53 10-2, 1.37 10-2, and ing sequences from the best performing to the worst performing
0.06, for Mg0.81Al, Mg1.38Al, Mg1.82Al, and Mg2.93Al, respec- mechanisms for individual catalysts are obtained. Mg0.81Al: ER
tively. χ2 of 0.052, 0.004, 0.033, and 0.011 achieved for (methanol adsorption) > LHHW (surface reaction) > ER (quasi-
Mg0.81Al, Mg1.38Al, Mg1.82Al, and Mg2.93Al (Table 8), respec- steady-state) > LHHW (methanol adsorption). Mg1.38Al: LHHW
4824 dx.doi.org/10.1021/ie101403f |Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE
Table 7. Elementary Reaction Rate Expressions for Hattori Mechanism with Methanol Adsorption As Rate Limiting
rate determining reaction overall reaction rate
K9 ½D½MeOl
k1 ½CH3 OH - 1=K1
þ CH3 OH S CH3 OH K4 K3 K2 ½T
r1 ¼
K9 ½D½MeOl K9 ½D½MeOl K10 ½M½MeOl K11 ½G½MeOl
1þ þ þ þ þ K2 ½T þ K9 ½D þ K10 ½M þ K11 ½G
K4 K3 K2 ½T K4 K6 K8
K9 ½D½MeOl
k2 ½T - 1=K2
T þ S T K4 K3 K1 ½CH3 OH
r2 ¼
K9 ½D½MeOl K10 ½M½MeOl K11 ½G½MeOl K9 ½D½MeOl
1 þ K1 ½CH3 OH þ þ þ þ þ K9 ½D þ K10 ½M þ K11 ½G
K4 K6 K8 K4 K3 K1 ½CH3 OH
K9
k3 K1 K2 ½CH3 OH½T - 1=K3 ½D½MeOl
CH3 OH þ T S ½TsCH3 OH þ r3 ¼ K4
2
K9 ½D½MeOl K10 ½M½MeOl K11 ½G½MeOl
1 þ K1 ½CH3 OH þ þ þ þ K2 ½T þ K9 ½D þ K10 ½M þ K11 ½G
K4 K6 K8
(methanol adsorption) > ER (methanol adsorption) > LHHW (methanol adsorption) > LHHW (surface reaction). Hence, the
(surface reaction) > ER (quasi-steady-state). Mg1.82Al: LHHW LHHW (methanol adsorption) provides consistently better repre-
(methanol adsorption) > ER (quasi-steady-state) > LHHW sentation of the experimental data compared to ER (methanol
(surface reaction) > ER (methanol adsorption). Mg2.93Al: ER adsorption), except Mg0.81Al, to which ER (methanol adsorption)
(quasi-steady-state) > LHHW (methanol adsorption) > ER applies better than any other mechanism, illustrated in Table 9.
4825 dx.doi.org/10.1021/ie101403f |Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE
Hattori methanol adsorption as rate determining Figure 3. Comparison between (a) conversion of triglyceride (T), (b)
K Mg0.81Al Mg1.38Al Mg1.82Al Mg2.93Al
moles of diglyceride (D), and (c) moles of monoglyceride (M), for
Mg2.93Al hydrotalcites obtained from ER quasi-steady-state model.
1 4.45 10-3 5.09 10-3 8.11 10-3 0.0140
2 0.0945 0.1376 0.0216 1.4572
3 0.1642 0.1607 0.1662 0.5854 Also, considering lower conversion resulting with lower molar
4 0.1218 0.2941 0.3244 0.4282 fraction of Mg in hydrotalcite catalyst (Appendix A), biodiesel
5 0.0119 0.0498 0.0146 0.0708
reactors are expected to incorporate hydrotalcite catalysts with
6 0.1344 0.1392 0.1893 0.0289
higher molar fraction of Mg, in which cases LHHW with methanol
7 0.0186 0.0073 0.0684 0.0639
adsorption as the rate-determining step adequately and reliably
represents the kinetic data. It also satisfies χ2 limit of 1.3429 for all
8 0.0750 0.0318 0.0194 0.0648
four catalysts.
9 0.0001 0.0018 8.34 10-5 0.0151
10 0.0502 0.0024 0.0887 0.0782
11 0.0644 0.0637 0.0130 0.0115 ’ CONCLUSIONS
k1 0.0002 0.0006 8.05 10-5 0.0080 Three kinetic mechanisms, ER, LHHW, and Hattori, based on
χ2 0.0515 0.1454 0.2988 0.0092 assumptions of quasi-steady state for the surface species and
4826 dx.doi.org/10.1021/ie101403f |Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE
’ APPENDIX A
Experimental Data. Wilson and coworkers14 developed a Figure 5. Evolution of catalyst surface from KMC simulations:
series of hydrotalcite catalyst with the general formulae of * ( 3 ), CH3OH* (∼), D* (þ), M* (ω), G* (O). Case a: * (100%),
[Mg(1-x)Alx(OH)2]xþ(CO3)x/n2- with x in a range of 0.25- CH3OH* (0), D* (0), M* (0), G* (0) at t = 0 s. Case b: * (2.9%),
0.55. Table 10 shows the nominal Mg/Al ratio along with the CH3OH* (0.07%), D* (97.1%), M* (0), G* (0) at t = 2000 s. Case c:
distribution of Mg and Al in the bulk and surface of these * (5.4%), CH3OH* (0.1%), D* (43.9%), M* (25.3%), G* (25.3%) at
catalysts. The effect of increasing Mg fraction on the activity t = 10 800 s.
and surface area of these hydrotalcite catalysts is shown in
Table 10. With the increase in Mg content the surface area of
tributyrate and 0.3036 mol (12.5 cm3) methanol. The batch
the catalyst initially decreases and then increases. The increase
reactor was run for 3 h.
in activity with the increase in the Mg molar fraction led to
higher conversion of triglyceride and more production of methyl
oleate. ’ APPENDIX B
The transesterification reactions were performed in stirred Genetic Algorithm (GA). The GA algorithm used in this
batch reactor at 333 K using 0.01 mol (3 cm3) of glyceryl simulation is based on the studies by Bhat et al.,25 Xu et al.,26
4827 dx.doi.org/10.1021/ie101403f |Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE
PAB
and Haupt and Haupt.28 The algorithm is implemented in Table 10. Properties of Hyrotalcite Catalysts16
MATLAB.
• Specify the initial inputs to the genetic algorithm. Popula- nominal surface activity/mmol glyceryl
tion size is 1000, while maximum number of iterations catalyst Mg/Al area min-1 tributyrate
is 100 000. The bounds of control variables (rate con- composition ratio (m2/g) g(cat)-1 conversion %
stants) are specified between 10-1 and 2 orders of magni-
Mg0.81Al 1:1 166.4 ( 8.3 0.004 42.4
tude.
Mg1.38Al 2:1 121.9 ( 6.1 0.01 49.2
• Initial chromosomes are generated, with each chromosome
representing a feasible solution in terms of decision vari- Mg1.82Al 3:1 92.5 ( 4.6 0.024 55.3
ables. Mg2.93Al 4:1 104.1 ( 5.2 0.025 74.8
• The objective function is generated for these sets of rate
constants. s i and q i are the measure of molecular van der Waals volume
• The chromosomes are ranked based on the objective and molecular surface area, respectively.
function.
• The best x percent (50%) chromosomes are kept for the li ¼ 5ðsi - qi Þ - ðsi - 1Þ
next iteration. P ðiÞ P ðiÞ ð16Þ
si ¼ vp S p q i ¼ vp Qp
• Generate a new set of chromosomes by mutation, crossover p p
in remaining chromosomes. The cost of the new chromo-
somes is evaluated on the objective function. Sp and Qp are group volume and area parameters, ν(i) p is the number
• Simulations are repeated until objective function is a very of p groups present in molecule i. The group activity coefficient Γp
low value within the expected tolerance limits or the is a function of area parameter Qp, area fraction θi, and the group
maximum number of iterations (100 000) is exceeded. interaction parameters ψip and aii0 . T is the temperature.
8 9
>
> >
>
’ APPENDIX C < X
nspc: Xnspc
θi Ψpi =
ln Γp ¼ Qp 1 - lnð θi Ψip Þ -
Prediction of Activity Coefficients of Species. In the
> P
nspc
UNIFAC method the activity coefficient (γi ) is made up of >
: i i θi Ψi0 i >
>
;
two parts: the first part provides the contribution of molec- i0
Φi θi Φi X
nspc ’ AUTHOR INFORMATION
ln γCi ¼ ln þ 5qi ln þ li - xi0 li0 ð13Þ
xi Φi xi i0 Corresponding Author
*E-mail: jhuma.sadhukhan@manchester.ac.uk. Phone: þ44-161-
Here, x i is the mole fraction, Φ i and θ i are the segment and 3064396. Fax: þ44 -161-236 7439.
area fractions for species i, respectively.
X
allgroups ’ ACKNOWLEDGMENT
ln γRi ¼ vðiÞ ðiÞ
p ðln Γp - ln Γp Þ ð14Þ Financial support from EPSRC (EP/D04829X/1) of the U.K.
p
for undertaking this research is gratefully acknowledged.
In these equations, Γ p and Γ (i)
p are the group residual activity
coefficient and the residual coefficient of group p in reference ’ NOMENCLATURE
solution containing only molecules of type i. [i] = activity of species i, mol m-3
Ci = concentration of species i, mol m-3
xi qi xi s i Ci(j) = concentration of species i at time point j, mol m-3
θi ¼ P Φi ¼ P ð15Þ Cexperimental (j) = concentration of species i obtained from
xi0 qi0 xi0 si0 i
i0 i0 experimental studies at time point j, mol m-3
4828 dx.doi.org/10.1021/ie101403f |Ind. Eng. Chem. Res. 2011, 50, 4818–4830
Industrial & Engineering Chemistry Research ARTICLE