Jesús Estebana,
Elena Fuentea,
María González-Miquelb,
Ángeles Blancoa,
Miguel Ladero*a and
Félix García-Ochoaa
aDepartment of Chemical Engineering, Complutense University of Madrid, Avda. Complutense s/n. 28040, Madrid, Spain. E-mail: mladero@quim.ucm.es; Fax: +34-913944179; Tel: +34-913944164
bSchool of Chemical Engineering and Analytical Science, The University of Manchester, The Mill Sackville Street, M139PL, Manchester, UK
First published on 8th October 2014
This study focuses on the thermal reaction between glycerol and ethylene carbonate to obtain glycerol carbonate and ethylene glycol under solventless homogeneous operation, the process being a transcarbonation of glycerol or a glycerolysis of ethylene carbonate. As the two reagents constitute an immiscible system at 40 °C evolving into a single phase at 80 °C, the evolution of phases with temperature was studied by focused beam reflectance measurement. As the biphasic system was inert, runs were completed under a monophasic regime from 100 to 140 °C with molar ratios of ethylene carbonate to glycerol of 2 and 3, achieving quantitative conversion of glycerol, as corroborated by a thermodynamic study. Second order potential kinetic models were proposed and fitted to the data. Finally, a comparison with analogous catalytic approaches was made, showing that this process performs better material-wise.
GlyCarb has received increasing interest as a potential bio-based product.3 It has shown outstanding properties in many applications as surfactant and solvent.4 Its use as a green-based solvent has been credited in Li-ion batteries,5 analytical applications,6 cosolvent with ionic liquids7 or as a solvent in immobilized liquid membranes for selective carbon dioxide separation from CO2/N2 mixtures.8 Moreover, its inclusion in building materials has proven effective for rapid hardening, reducing shrinkage of the material and improving compressive strength.9
Moreover, GlyCarb can also be regarded as a building block. Atom transfer radical polymerization initiators can be synthesized to yield polymers with end-functional five-membered cyclic carbonate groups for application as coatings, macromolecular surfactants and adhesives.10 GlyCarb has substituted the less environmentally friendly glycidol in the synthesis of hyperbranched polyethers.11 By acylation of GlyCarb, several esters have been obtained with surfactant features as well as thermal and oxidation stability.12 Secondary amines also react with it to produce alkyl glycerol carbamates used as thickeners in surface-active preparations.13
Traditional production of GlyCarb used to be accomplished with phosgene. Nevertheless, this hazardous method has been substituted by alternative procedures. Reaction of glycerol with urea at 140 to 150 °C under vacuum conditions and catalysts like rare earth metal oxides, La2O3, Zn and Mn sulphates or calcined Zn hydrotalcites yields 86% GlyCarb in the best case scenario.14–19 Direct addition of CO2 was tested with tin-based catalysts under solventless conditions at 180 °C and 5 MPa;20 in the presence of methanol as solvent, conditions were lowered to 80 °C and 3.5 MPa improving the yield of the process.21 Even supercritical conditions (40 °C, 10 MPa) with basic ion exchange resins and zeolites22 were tried. In none of these cases the yields achieved were higher than 35%.21 Similarly, carbonylation via the addition of CO and O2 mixtures with a palladium-based catalyst has been undertaken with a yield as high as 92% and almost total selectivity.23
However, the most followed trend is the use of organic carbonates to perform the transesterification of glycerol, known for over fifty years,24 due to the high yield obtained at low temperature. Particularly, transesterification to GlyCarb with dialkyl carbonates has been much more widely covered11,25–31 than that with ethylene carbonate (EtCarb), with fewer references being found until the present date.19,22,32,33 As presented in Scheme 1, an additional advantage of the latter reaction is that ethylene glycol (MEG) is obtained as well, being this a product extensively used as antifreeze and other applications. This substance is obtained by hydrolysis of ethylene oxide (EO), giving MEG and oligoglycols. The best procedure seems to be Shell's OMEGA process, in which EO is carbonated and subsequently hydrated to yield 99.5% MEG, a further effort to reach the most valuable product and the maximum achievable exploitation of feedstock.34–37
At temperatures above the melting point of EtCarb (36 °C), the two reactants of the proposed reaction constitute a liquid–liquid dispersion up to a certain temperature. In order to study heterogeneous systems, focused beam reflectance measurement (FBRM) has been successfully applied. This technique has been used to measure particle and droplet sizes in suspensions of solids in liquids,38 flocculation processes39 and liquid–liquid dispersions.40,41 An in situ monitoring of the evolution of the dispersion with temperature is herein proposed to determine the phase changing behaviour of the system.
According to literature, thus far, GlyCarb synthesis has only been pursued through catalytic procedures, with the concomitant need for removing the catalyst from the final product and/or regenerating it after a certain operation time. Hence, the aim of this work is to develop a novel and more sustainable solventless thermal process at atmospheric pressure and low to moderate temperatures. A thermodynamic study of the reaction at the conditions tested is presented together with a kinetic model for this thermal reaction.
![]() | ||
Fig. 1 Evolution of the mean chord length and droplet count per second with temperature. Conditions: stirring speed (SS) of 750 rpm and Gly at a EtCarb to Gly molar ratio (M) of 3. |
Values of the mean chord length at 40 °C are in the millimetre range, while at temperatures equal or higher than 75 °C they are in the micrometer range. At 80 °C, the mean chord length was negligible, indicating that the size of the droplets was smaller than the detection limit of the FBRM (1 µm), so it can be inferred that EtCarb and Gly gradually dissolved in each other so that the dispersion system evolved into a single phase above 80 °C. This is confirmed by the exponential reduction of the number of counts or events per second detected by the FBRM.
To observe the influence of the number of phases in the reacting system and the temperature, some preliminary runs were conducted at different temperatures allowing batches of the reagents (molar ratio of EtCarb to Gly equal to 2) to interact under agitation for 24 hours (data not shown). No products were detected while operating at temperatures within which the system showed liquid–liquid biphasic behaviour, at 75 °C and lower temperatures. Only the presence of products was observable when operating at 80 °C, though the conversion X only amounted to 9.3% after the mentioned period of time. Thus, it can be said that the system was almost inert while in a liquid–liquid biphasic state and the thermal reaction only took place at appreciable rate under homogeneous conditions and temperature equal to or higher than 100 °C. At atmospheric pressure and 140 °C, some evaporation phenomena were observed, thus being this the highest temperature selected for further studies.
Application of the well-known Kirchhoff laws for reaction enthalpy and entropy as functions of temperature leads to values of these thermodynamic functions at 80 to 140 °C. Gibb's free energy values have been estimated from the mentioned functions and the equilibrium constants from Gibbs free energy at the temperature values were also computed (Appendix 1). For the mentioned calculations to be performed, several literature references were consulted.43–49 The main results are summarized in Table 1. It can be seen from the value of the equilibrium constant at temperatures equal to or higher than 100 °C that the global reaction is shifted towards the products, while reaction enthalpies imply endothermicity. The influence of the entropy in the equilibrium is decisive, being the most influential term in Gibb's free energy, with a higher impact as temperature rises. For this system, the following equations relate the thermodynamic functions and the equilibrium constant with temperature:
ΔH0r = 166.21 − 0.1556T[K] | (1) |
ΔS0r = −9193 + 30.49T[K] | (2) |
ΔG0r = 4862 − 14.93T[K] | (3) |
K = 10−5![]() | (4) |
Temperature (°C) | ΔH0r (kJ mol−1) | ΔS0r (J mol−1 K−1) | ΔG0r (kJ mol−1) | K |
---|---|---|---|---|
100 | 108.26 | 2186.66 | −707.37 | 14.42 |
110 | 106.53 | 2487.25 | −846.08 | 20.78 |
120 | 104.91 | 2790.85 | −991.89 | 30.04 |
130 | 103.41 | 3097.22 | −1144.77 | 43.55 |
140 | 102.04 | 3406.11 | −1304.69 | 63.29 |
Runs for 48 hours were performed at temperatures in the 100 to 140 °C interval and at EtCarb to Gly molar ratios of 2 and 3, in order to reach equilibrium conditions in the relevant operational range. In parallel, calculation of values of conversion at equilibrium was performed from the equilibrium constant at each temperature using eqn (5).
![]() | (5) |
Experimental conversion values for Gly, with their absolute errors, are shown in Fig. 2, together with the computed conversions for the ideal and real liquid approaches. It can be inferred that the system approaches total conversion as the temperature rises and the EtCarb to Gly molar ratio increases, in agreement with the equilibrium constant values. At the same time, although activity coefficients are far from the unit, especially in the case of glycerol and glycerol carbonate, the effect of the high concentration of reagents and products in solution is almost negligible. Moreover, computed values for conversion at equilibrium are in agreement with experimental values, given the absolute error intervals for the latter.
![]() | ||
Fig. 2 Experimental and calculated conversion of Gly at initial molar ratios of EtCarb to Gly of 2 (a) and 3 (b). |
For kinetic model fitting purposes, the software Aspen Custom Modeler was employed. In this program, an algorithm for non-linear regression based on the Levenberg–Marquardt method was applied simultaneously with the numerical integration of the proposed kinetic equation corresponding to each model through a fourth-order Runge–Kutta method.
Initially, correlation of each model was realized at individual temperatures. After obtaining the value of the kinetic constant (or constants) at each temperature, estimates of Ea/R parameters were retrieved, from which simultaneous correlation or each kinetic model to all data at all temperatures was performed to obtain the multivariable fitting parameters.
Table 2 compiles the diverse kinetic models utilized to fit to the experimental data gathered that were proposed in this work. Model 1 was defined as a potential model of first order with respect to EtCarb, keeping the concentration of Gly constant, and a part of the apparent kinetic constant thereof obtained. Model 2 considered an analogous situation, in which only the concentration of Gly was regarded as influential to the kinetic model, becoming the concentration of EtCarb a part of the apparent constant. These models imply that one of the reagents is the main component of the phase were the reaction takes place; the other phase is mainly composed by the other reagent. The dispersed phase droplets would be forming a nanoemulsion, so an FBRM analysis would be not able to detect it (The FBRM herein employed had a lower limit of 1 µm for the diameter of the detected particle).
Model number | Rate equations |
---|---|
1 | r = k1CEtCarb = k1CEtCarb0(M − X) |
2 | r = k2CGly = k2CGly0(1 − X) |
3 | r = k3CGlyCEtCarb = k3CGly02(1 − X)(M − X) |
4 | r = k4CGlyCEtCarb − k5CMEGCGlyCarb = k4CGly02(1 − X)(M − X) − k5(CGly0X)2 |
Model 3 describes an overall second order potential kinetic model, with partial first orders with respect to the concentrations of the reactants. Successful fitting of this type of model for the transesterification of dimethyl carbonate and ethanol has been reported.50 Likewise, second order potential kinetic models have been applied to esterification reactions.51,52 Finally, Model 4 still considers a reversible second order potential model. Said situation would be described with a reverse reaction from the products to the reactants. The latter model was tested after some results from the equilibrium runs suggested a conversion slightly lower than one for EtCarb to Gly molar ratio of 2 and temperature varying from 100 to 120 °C.
To select one of the proposed models, statistical criteria defined in the experimental section were used, as well as physicochemical criteria as the value of activation energies.
Table 3 compiles the statistical and fitting parameters calculated after multivariable correlation of all data. Regarding the parameters of the models, a definition of the dependence of the kinetic constants kj with temperature was made following a modified Arrhenius equation suitable for computational purposes:
![]() | (6) |
Model | Parameter | Value | ±Error | F95 | AIC | RMSE | VE (%) |
---|---|---|---|---|---|---|---|
1 | ln![]() |
2.64 | 0.98 | 718 | −3.90 | 0.14 | 77.61 |
Ea1/R | 3656 | 388 | |||||
2 | ln![]() |
13.04 | 0.67 | 4407 | −5.62 | 0.06 | 96.01 |
Ea1/R | 7154 | 263 | |||||
3 | ln![]() |
11.72 | 0.25 | 34426 | −7.69 | 0.02 | 99.50 |
Ea1/R | 7436 | 100 | |||||
4 | ln![]() |
11.22 | 0.29 | 25335 | −8.08 | 0.02 | 99.67 |
Ea1/R | 7217 | 116 | |||||
ln![]() |
21.64 | 3.73 | |||||
Ea2/R | 5143 | 1472 |
The activation energies range from 30 kJ mol−1 to around 60 kJ mol−1. The activation energies of processes controlled by the chemical reaction step usually acquire values between 40 and 200 kJ mol−1. These figures can be expected for a homogeneous reacting system.
Due to the value of Ea1/R being below this interval and the poor degree of fitting shown, especially concerning the variation explained (VE), Model 1 was dismissed. Model 2 showed better agreement between experimental and predicted values, increasing significantly the adjusted Fischer parameter (F) and VE. Correlation of Models 3 and 4 lead to a further marked enhancement of all statistical criteria. Nevertheless, as inferred from the results in Table 3, there is no clear evidence that Model 3 is better than Model 4 or vice versa: while F is higher for Model 3 and RMSE has a similar value, VE and AIC show both slightly worse values in terms of goodness of agreement than those obtained for Model 4. In any case, as stated, second order models had been proposed in literature to describe comparable chemical reactions,50–52 and these results further probes the observations in the FBRM studies: the system is homogeneous.
Nonetheless, taking into account results from the thermodynamic studies, it can be said that only the direct reaction takes place when the molar ratio of EtCarb to Gly is equal or higher than 3, while at lower values, the influence of the reverse reaction is considerable, mainly at temperatures of 120 °C or lower. This leads to select Model 3 as the most adequate to represent the transesterification of glycerol and ethylene carbonate in the more common situation of M ≥ 3, and to select Model 4 as the most precise for lower values of the reagent's molar ratio. In Fig. 3, the reasonable fitting of Model 3 to experimental results can be observed.
Finally, Fig. 4 shows the evolution of the relative error of the prediction for Model 3, with a positive value implying an underestimation of such model with respect to the observed data. While the relative error is higher at short times than afterwards, partly due to the absolute values of the variable X being much smaller, no clear trend in regards of an under or overestimation can be observed for the two sets of experiments; thus, this is further proof for the validation of the model, acceptable even for M = 2 if experimental error bars showed in Fig. 3 are taken into account.
![]() | ||
Fig. 4 Evolution of the error of prediction of Model 3 with respect to the experimental measures operating with (a) M = 2 and (b) M = 3 Conditions: fixed agitation speed of 750 rpm. |
First, it can be said that the atom economy and the carbon efficiency, as defined in the experimental section of this work (eqn (14) through (16)) of the transesterification of Gly and EtCarb to give GlyCarb and MEG are virtually 100%, and the E-factor value is equal to zero in all cases if only the synthetic process herein studied is considered. None of the references cited the use of any solvents and the catalysts employed, where applicable, could be subject to reutilization. The process studied is a one-step solventless reacting system in all cases.
Regarding the comparison between the thermal process herein reported and the catalytic ones found in literature, mass productivity (MP) and reaction mass efficiency (RME) were computed on the basis of a reference experiment (at 100 °C and a molar ratio of EtCarb to Gly of 2). The same parameters were calculated for the experiments that achieved the best yields to products in the other references.
Table 4 compiles the mentioned calculations along with the operating conditions and catalysts described in the other references. The values of MP and RME show that the process herein proposed performs better than the rest. This can be ascribable to the fact that no catalyst was used and virtually total conversion was achieved. When catalysts were used, yields to the products were lower given the activity limitations to the completion of the reaction; whereas in the thermal process, the final yield to product (equal, being this an elemental reaction, to the conversion of glycerol in percentage) is only restricted by thermodynamic considerations, not by mass transfer or deactivation of the catalyst. Also, the values of MP and RME herein obtained apply to batch processes, where no recycling of the molar excess of EtCarb used in is contemplated. Should this excess be recycled, the values of these sustainability parameters could be further improved in all cases.
Reference | YGC (%) | Catalyst | Reaction conditions | MP (%) | RME (%) |
---|---|---|---|---|---|
a At atmospheric pressure except otherwise specified.b Supercritical conditions. | |||||
19 | 91.0 | Al–Mg mixed oxide derived from hydrotalcite with Al/Mg molar ratio of 0.25 | T = 50 °C; M = 2; 7 wt% of catalyst with respect to the total weight of reactantsa | 45.92 | 61.80 |
22 | 32.2 | Purosiv zeolite | P = 13 MPa; T = 74 °C; M = 0.63; 125% wt% of catalyst with respect to total weight of reactantsb | 11.70 | 24.68 |
33 | 83.8 | Tri-n-butylamine supported on MCM-41 molecular sieve | T = 80 °C; M = 2; 3.1% wt% of catalyst with respect to the total weight of reactantsa | 44.92 | 56.29 |
This work | 96.9 | None: thermal reaction | T = 140 °C; M = 2a | 50.26 | 66.37 |
Fischer's F is based on a null hypothesis which accounts for the adequacy of the model to the observed values of the variable compared to given values of F at 95% confidence (or other). It is defined according to the following equation
![]() | (7) |
Furthermore, the Akaike's information criterion (AIC) has been regarded given that it has previously been applied as a standard of judgment for kinetic model discrimination.53,54 This parameter relates the amount of experimental data available to the number of parameters of the model proposed and is defined following eqn (8), being the model better when the AIC value, always negative, is lower:55
![]() | (8) |
In addition to the F and AIC, the residual mean squared error (RMSE) has been regarded as measure of the difference of the values of the variable being evaluated predicted by the model to those obtained experimentally considering, once again, the number of data available together with the parameters.54 As this parameter is related to the sum of variances, the better the model fits to data, the lower the value of RMSE is:
![]() | (9) |
Finally, if the variation between adjacent data is considered, the percentage of variation explained (VE) also gives information of the quality of fit for each measured variable, being best when all experimental trends are well explained by the tested model (a value near or equal to 100%). It is quantified using eqn (10):
![]() | (10) |
![]() | (11) |
![]() | (12) |
![]() | (13) |
In eqn (11) through (13), ζl is the heteroscedasticity parameter, which is a measure of the type of error in the measured variable. When the value of this parameter is not fixed, as such was the case, Aspen Custom Modeler considers ζl = 1 by default.
![]() | (14) |
The atom economy (AE) is defined as the ratio of the summation of the molecular weights of the desired products to that of the reagents utilized.
![]() | (15) |
The E-factor computes the mass of waste generate per unit mass of the products.
![]() | (16) |
The carbon efficiency (CE) regards the amount of carbon that transits from the reactants to the desired end products.
![]() | (17) |
The mass productivity (MP) accounts for the mass of the actual product in relation to the total mass of material utilized in the process.
![]() | (18) |
Finally, the reaction mass efficiency (RME) considers the total actual mass of the products to the mass of reagents used.
![]() | (19) |
It was observed by means of a FBRM probe that the system constituted a dispersion-like liquid–liquid biphasic system from 25 °C to 80 °C, decreasing the droplet size and number as temperature increased, till it was monophasic.
A thermodynamic study determined that the reaction was almost irreversible at 100 °C or higher temperature. A kinetic study served to determine by statistical means that an overall second order potential model, accounting only for the direct reaction, represents adequately the transesterification of ethylene carbonate and glycerol in these conditions.
Finally, a comparative study with other references found in literature regarding the sustainability of the process was conducted; proving that the thermal process herein studied could play an attractive role when taking into account the sustainability of the process, according to common green metric parameters.
EtCarb | Ethylene carbonate |
MEG | Ethylene glycol |
GlyCarb | Glycerol carbonate |
Gly | Glycerol |
AE | Atom economy (eqn (15)) |
AIC | Akaike's information criterion |
C | Concentration of the components at a given time (mol L−1) |
CE | Carbon efficiency (eqn (17)) |
Cp | Specific heat capacity (J mol−1 K−1) |
Eaj/R | Ratio of activation energy and the ideal gas constant (K) |
F | Fischer's F statistical parameter at 95% confidence |
FBRM | Focused beam reflectance measurement |
H | Enthalpy (kJ mol−1) |
HPLC | High-performance liquid chromatography |
K | Thermodynamic constant of equilibrium |
k1…5 | Kinetic constants for the tested models |
kj0 | Preexponential factor of the kinetic constant |
l | Referenced individual variable in eqn (11)–(13). |
M | Initial molar ratio of dimethyl carbonate to glycerol |
MP | Mass productivity (eqn (17)) |
N | Total number of components |
N | Total number of data to which a model is fitted |
P | Number of parameters of a proposed model |
r | Reaction rate (mol L−1 min−1) |
R | Ideal gas constant (J mol−1 K−1) |
RME | Reaction mass efficiency |
RMSE | Residual mean squared error |
S | Entropy (J mol−1 K−1) |
SQR | Sum of quadratic residues |
SS | Stirring speed (rpm) |
T | Temperature (K) |
VE | Variation explained (%) |
x | Molar fraction |
X | Conversion, as defined by eqn (14) |
γ | Activity coefficient |
δ | Variation |
ζ | Heteroscedasticity parameter |
ν | Stoichiometric coefficient of the component i |
ω | Agitation rate (rpm) |
0 | Relative to the start of the reaction, time equals zero |
cat | Relative to the catalyst |
f | Relative to formation (enthalpy) |
i | Relative to component i |
j | Relative to reaction j (j = 1, direct reaction; j = 2, reverse reaction) |
r | Relative to reaction (enthalpy and entropy) |
0 | Relative to standard conditions |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4ra11209a |
This journal is © The Royal Society of Chemistry 2014 |