- Email: [email protected]
The butoxylation of dodecylamine: Reaction mechanism and kinetics
Journal Pre-proofs The Butoxylation of Dodecylamine: Reaction Mechanism and Kinetics P. Müller, R.D.E. Krösschell, W. Winkenwerder, J. van der Schaaf PII: DOI: Reference:
S1385-8947(19)32349-6 https://doi.org/10.1016/j.cej.2019.122939 CEJ 122939
To appear in:
Chemical Engineering Journal
Received Date: Revised Date: Accepted Date:
4 April 2019 19 September 2019 22 September 2019
Please cite this article as: P. Müller, R.D.E. Krösschell, W. Winkenwerder, J. van der Schaaf, The Butoxylation of Dodecylamine: Reaction Mechanism and Kinetics, Chemical Engineering Journal (2019), doi: https://doi.org/ 10.1016/j.cej.2019.122939
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2019 Published by Elsevier B.V.
The Butoxylation of Dodecylamine: Reaction Mechanism and Kinetics
P. Müller1, R.D.E. Krösschell1, W. Winkenwerder2, J. van der Schaaf1,*, 1 Laboratory
of Chemical Reactor Engineering, Department of Chemical Engineering and Chemistry, Eindhoven University of
Technology, P.O. Box 513, 5600MB, the Netherlands 2 Nouryon,
Brewster, New York 10509, United States* Corresponding author: [email protected]
Abstract The reaction mechanism and kinetics of the butoxylation of dodecylamine were investigated experimentally using a semi-batch, fully liquid process at temperatures ranging from 120 ºC to 150 ºC and varying stoichiometric ratios of the reactants. In addition, the catalytic effect of adding 1-3%mol of twelve different functional organic species containing hydroxyl, aldehyde, amine, and/or amide functional groups was studied. It was found that only proton donating groups increase the observed reaction rate. In particular hydroxyl groups in combination with an amine group resulted in a strong acceleration of the reaction. Since the reaction intermediates mono-and dibutoxylated amines have this combination, an autocatalytic reaction mechanism and corresponding rate law are proposed. The kinetic constants were fitted to the experimental data as a function of temperature, following an Arrhenius type of dependency. The results from the model describe the experimental data with 95% accuracy. Moreover, the results show that the butoxylation of dodecylamine, as a model reactant for a short substituted ethylene oxide, with fatty amines, follows the same mechanism and has similar kinetics as epoxide hardening reactions with amines.
1. Introduction Ethoxylated and propoxylated fatty amines are widely used as nonionic surfactants, leading to a steady increase in industrial demand over the past six decades.1,2 These are key components in products of 1
numerous industrial processes, e.g. oil drilling, paper manufacturing, textile as well as fiber manufacturing, paints, dyes and plastics. These products are produced industrially by reaction of amines with 2 moles of alkene oxides (e.g. ethylene or propylene oxide), giving a tertiary amine of the type shown in Figure 1, and further polymerized using a strong basic catalyst,3–5 to provide a polyalkoxilated amine with 5-50 molecules of ethylene oxide and/or propylene oxide attached, which find use as surfactants. Compared to this well-studied polymerization, the initial reaction of fatty amines with two moles of alkoxide is rarely investigated and not fully understood (in this article the term epoxide is used to specifically indicate a three-member ring with an oxygen atom; alkoxide is instead used to indicate all epoxide-containing species, i.e. substituted epoxides). Although the mechanism of the reaction between ethylene oxide and fatty alcohols,6 phenols7 and acids8 has been thoroughly investigated, less attention has been directed to the industrial reaction with fatty amines. Komori et al. were the first in the 1950s to experimentally demonstrate that fatty amines can react with two molar equivalents of ethylene oxide without a catalyst, 9 following an SN2 mechanism for the substitution of both protons of the amine with the alkoxide (Figure 1).3,9 B H H R1
NH
SN2
O
B
R1 R1
O
R2 N H
R2
OH N
R2
R2
HO OH R2
R1 = C8-C18
Figure 1: Nucleophilic attack of a fatty amine to alkoxides, like butylene- or propylene oxide. The overall process involves two successive SN2 reactions with two moles of alkoxide, finally leading to a tertiary amine.
In the above mechanism, the proton transfer steps between the amine and the newly formed hydroxyl group is not fully understood yet. Often a carrier (B) like water or other residues are conceptually introduced to explain this proton transfer.10,11 The reaction is known for being auto-catalytic.1,12–14 The auto-catalytic effect is believed to be due to the destabilization of the ethylene oxide ring by the hydroxyl 2
groups of the products, but this functionality could also act as proton transfer agents. The two effects cannot be clearly separated because of their co-existence. In this article we investigate the reaction between fatty amines and butylene oxide. Kinetic profiles, reaction modeling, and spiking experiments with a range of differently functionalized species are reported, which provided important information about the process and the mechanism of the reaction.
2. Theory The nucleophilic attack of the amine group towards the epoxide ring depends on three main types of effects: steric, polar, and catalytic.6,9,14,15 The substitution pattern, especially of the alkoxides, has steric and polar effects, and a decrease in the reaction rate compared to ethylene oxide is observed for substituted or more polar groups, such as epichlorohydrin15. Furthermore, a substitution on one carbon of the alkoxide ring reduces the rate because only the less substituted methylene carbon participates in the reaction.15–17 Sundaram et al.15 correlated the reaction rate to the degree of substitution in the series ethylene-, propylene- and 1,2-butylene oxide, with several amines, concluding that the different alkoxides follow the same reaction pathway. The catalytic effect is most likely related to the oxide ring opening reactions by a proton donor, such as water, acids or alcohols15,17. The proton destabilizes the ring by bonding with the oxygen, thus reducing the activation energy for the alkoxylation reaction. Unfortunately, such catalytic effect also promotes the undesired polymerization of the alkoxides to polyglycols, commonly observed as side products in these processes.18 This tradeoff between rate acceleration and side product formation of traditional catalysts mostly results in the investigation of non-catalyzed reactions, with a consequent lack of knowledge on the catalytic aspects of the process. Additions to the aforementioned basic SN2 mechanism were proposed. Sanders et al.14 related freely formed bases at low temperatures of 90 ºC to a highly catalytic quaternary ammonium species, formed by reaction with a third equivalent of ethylene oxide. This thermally unstable ion would increase the rate significantly after being formed. This mechanism is however controversial.19–21 The second mechanism discussed in a review by 3
Smith3 begins with the assumption that the epoxide ring must first be destabilized to start the reaction4 (Figure S1). In non-catalyzed reactions, trace amounts of water are thought to act as the catalyst in the initial step of the reaction; subsequently, the reaction products catalyze the ring opening. A termolecular step between the proton donor, alkoxide and amine is proposed as rate limiting step. However, in more recent investigations this thesis is opposed.21,22 In particular, Enikolopiyan et al. were able to prove that the reaction can proceed in the absence of water, using ultra-pure and anhydrous reactants.22 Moreover, elementary termolecular steps are highly unlikely from a kinetic and statistical point of view and are thus highly uncommon. As an alternative to this termolecular reaction, a dipolar intermediate was suggested23 (Figure S1). Sirovski et al. described the reaction of ethoxylation of dodecylamine with a mechanism consisting of an initial SN2 step followed by catalyzed terms with the sum of the products’ hydroxyl groups and water as external catalyst. However, there is no mechanistic study of how these terms were found.13 There is thus a clear gap in the research regarding the mechanism and kinetics of the alkoxylation of fatty amines (including ethylene-, propylene- or butylene oxide), which we are striving to close. In this work, the mechanism and, more specifically, the driving force behind the auto-catalytic behavior was illuminated through investigating the influence on the reaction rate of catalytic amounts of twelve different organic species, containing hydroxyl, aldehyde, nitrile, amine and amide functional groups. These spiking experiments provided invaluable information on the influence of the different functionalities in accelerating or inhibiting the reaction, and their consequent connection to the process and the mechanistic aspects of the reaction.
3. Experimental Section Chemicals. As in Sirovski et al. work,13 dodecylamine (DDA) was chosen as model compound, as it represents a large fraction of most fatty amine mixtures employed industrially, typically ranging from C8 to C18 alkyl chains. Octylamine (OA) was tested as alternative for DDA to demonstrate the generality of 4
the mechanism. Butylene oxide (BO) shows a lower hazard level than ethylene and propylene oxide, and was therefore used as a safer and more convenient model reactant. The chemicals used for the experiments and analytics are listed in Table 1. Among the spiking chemicals, the pure reaction products and methyl dodecylcarbamate are not commercially available and were made in the lab. The final product (bis(hydroxybutyl)dodecylamine, TERT) could be obtained in 99%mol, while the intermediate secondary amine (mono-(hydroxybutyl)dodecylamine, SEC) could be obtained at a concentration of 75%mol from the reaction crude. Dodecyl methyl-carbamate was produced at ambient conditions from dodecylamine and carbon dioxide from air according to the method presented by Aresta and Quaranta24. The purity of the produced carbamates was analyzed with GC to be greater than 90%mol.
Table 1: list of purchased chemicals. The model reaction was dodecylamine (DDA) with 1,2-butylene oxide (BO). Octylamine (OA) was used as alternative to DDA. The listed spiking chemicals were only added in the spiking runs additional to the reaction crude.
Name
Supplier
CAS-number
Purity
Reaction chemicals Dodecylamine (DDA)
Acros Organics
124-22-1
≥95.0%
1,2-Butylene oxide (BO)
EMD Millipore
106-88-7
≥99.0%
106-88-7
≥99.0%
Corporation Octylamine (OA)
Merck KGaA
Analytical chemicals d-Chloroform
Sigma-Aldrich
865-49-6
≥99.96 atom%
Acros Organics
108-88-3
≥99.7%
+0.03%TMS Toluene
Spiking chemicals Methanol
Acros Organics
67-56-1
≥98.0%
Butan-2-ol
Alfa Aesar
78-92-2
≥99.0% 5
2-(Ethylamino)-ethanol
Alfa Aesar
110-73-6
≥98.0%
Diethanolamine
Alfa Aesar
111-42-2
≥99.0%
Dipropylamine
Merck KGaA
142-84-7
≥99.0%
DABCO (1,4-
Sigma-Aldrich
280-57-9
≥99.0%
Salicylaldehyde
Alfa Aesar
90-02-8
≥99.0%
Dodecylamide
TCI Chemicals
1120-16-7
≥96.0%
Dodecylnitrile
TCI Chemicals
2437-25-4
≥96.0%
diazabicyclo[2.2.2]octan)
Methyl dodecylcarbamate
/
/
90%
SEC
/
/
75%
TERT
/
/
99%
Set-up. The set-up is schematically depicted in Figure 2. The reactions were performed in a 158 ml Cyclone 075 stainless steel autoclave (Büchi Glassuster) in which pressure and temperature were monitored at a measurement rate of one second. The heat exchanger was an oil bath RE306 from LaudaBrinkman. BO was added into the reactor through a capillary by using a syringe, while samples were taken through the bottom valve.
6
Figure 2: Experimental set-up for kinetic measurements.
Procedure. DDA was preloaded into the reactor at room temperature. Subsequently, the reactor was heated to reaction temperature and vacuum was applied until a pressure below 20 mbar was reached. After the reaction temperature was reached, BO at room temperature was added in approximately three seconds by using a syringe. In a stepwise manner of 0.5 mole of BO per mole of DDA per step. Before each subsequent addition of BO, it was made sure that the previously added amount of BO was completely reacted by monitoring the pressure in the reactor. Due to the high solubility of BO in DDA the system was treated as a homogenous liquid phase reaction of which the vapor pressure is made up nearly entirely of BO. At full conversion of BO the pressure returned close to vacuum. This procedure was chosen to model the industrial semi-batch process and to allow a fast homogenous mixing of the low volume of BO. The stirring rate was kept constant at 500 rpm. Initial tests of stirring speeds, ranging from 100 to 1000 rpm, showed that no mass or heat transfer limitations were present above 250 rpm at the conditions that gave the highest reaction rate. The same procedure was used for the reaction of OA with BO to demonstrate the applicability of the reaction and the model (vide infra).
7
Catalytic effect of functional groups (spiking experiments). The catalytic activity was tested by measuring the reaction rate of a mixture of 0.25 mole of BO and 1 mole of DDA after adding 1-3%mol of each investigated species. The species used comprise the list of spiking chemicals in Table 1, including SEC, TERT, and methyl-dodecylcarbamate. The reaction rates where normalized with respect to the total added amount of spiking chemicals in moles and compared to the non-catalyzed reaction.
Reactor temperature. In an initial screening, it was determined to use a reaction temperature window of 120 ºC to 150 ºC in steps of 10 ºC. Above 150 ºC, the BO ring opens spontaneously and by-products were formed. Below 120 ºC full conversion could not be achieved in a time of 50 hours. The addition of BO at room temperature to the hot reactor led to an initial drop in temperature at the beginning of each experiment. Assuming a very conservative overall heat transfer coefficient of 26 W/m2K, an area available for heat transfer of 1.6·10-3 m2, and an initial measured temperature difference of 16 K to the desired reaction temperature, the estimated time to reheat the reaction mixture is about 10 minutes to reach 95% of thermal equilibrium (see SI Figure S3). This time span is an order of magnitude smaller than the total reaction time of 120 minutes. Therefore, the reactor is assumed to be isothermal over each run. After the thermal equilibrium was reached, the maximum measured temperature variation was 2 K. Note that in the first 10 minutes no samples were taken, so the analysis is only based on data gained from isothermal conditions.
BO concentration in the liquid phase. The molar fraction of BO in the mixture of amines was always less than 0.33 and BO is the only component with a non-negligible vapor pressure (Pv,DDA= 0.024 bar vs. Pv,BO=7 bar at 140 ºC). After detecting a pressure of maximum 1.59 bar in the reactor (after vacuuming the system) it was clear that the reaction cannot be treated as a gas-liquid system as, for example, the carbonation reaction with CO2 in the work of Cai et al. 25, but as a fully liquid system. The measured vapor pressure value was even lower than the value expected from an ideal mixture (2.32 bar), which
8
clearly supports our treatment of this system as a fully liquid one. The concentration of BO in the liquid phase could be determined using equation (3.1). 𝐶𝑖𝑛𝑖𝑡
𝐶𝐵𝑂 = 𝑃𝐵𝑂,𝑖𝑛𝑖𝑡 ∙ 𝑃𝐵𝑂
(3.1)
The liquid phase concentration of BO (CBO) is assumed directly proportional to the pressure in the reactor. The factor of the initial concentration divided by the initial pressure corresponds to the activity coefficient at this point , which was assumed to be constant during reaction. Even if this would not hold, the direct proportionality to BO partial pressure warrants that the kinetic rates are taken proportional to the activity of BO.
Experimental accuracy. A quantitative measure of the experimental error in the concentration was obtained by comparing the deviations in double and one triple rerun of experiments with their respective means. For the BO pressure, the experimental error was found to be on average 1.9% while for DDA, SEC, and TERT, the averaged standard deviation was 2.1%, 3.2%, and 4.2%, respectively. The relative errors were of sufficiently low levels so that the mass balance over the reactor was always maintained within 5% of deviation.
Density of the mixture. The density was studied with the aim of accurately calculating the reaction mixture volume and the mass balance. For several reaction compositions, the density was determined by heating a measured amount of mass of varying amine species compositions excluding BO, in a glass cylinder and measuring the volume. The experiments were conducted between 323 K and 393 K. The intermediate product SEC could not be purified. Therefore a general approach using the total degree of BO moles added per mole DDA (characterized by n in equation 3.2) was chosen, which is comparable to Rupp et al.26
9
The density of pure DDA at room temperature by the model was 801.7 kg/l, equal to the data reported by the supplier. The following empirical correlation (3.2) was found to accurately describe (compare to Figure 3a) the obtained data as a function of mixture composition and temperature, as shown in Figure 3b. 𝜌𝑚𝑖𝑥 = 𝑎(𝑛) ∙ 𝜃(º𝐶) + 𝑏(𝑛), 𝑎(𝑛) = 0.042𝑛3 ― 0.170𝑛2 + 0.255𝑛 ― 0.983,
(3.2)
𝑏(𝑛) = ―3.5𝑛3 ― 6.5𝑛2 + 62.0𝑛 + 821.4
Figure 3a: parity plot density model.
Figure 3b: Mixture densities as a function of temperature. The reaction coordinate n expresses the degree of butoxylation of dodecylamine, ranging from 0 to 2 stoichiometric mole adducts of BO to DDA.
The parameters a and b were determined by a nonlinear surface fit of the experimental data, using MATLAB’s lsqnonlin with the trust-region-reflective algorithm. The effect of BO in the liquid phase was not taken into account in the model, because of its low volumetric concentration of less than 10%vol.
Analytics. GCMS: The samples were analyzed on a GCMS-QP2010 (Shimadzu) with a DB-5 column using helium as carrier gas. For SEC and TERT the mass spectrum was used to identify the corresponding peaks to the retention times in the chromatogram. The calibration of SEC was done by cross referring to NMR data. The mixtures of DDA and its intermediates tend to interlink via the hydroxyl groups, which 10
was visible in GC as broadened bulk peaks with a longer retention time. Due to this the samples had to be analyzed within at most twelve hours27. NMR: A Bruker 400 MHz was used. Figure 4 shows the final product TERT. For this compound, the following spectrum was recorded 1H NMR (400 MHz, CDCl3) δ 3.62-3.48 (2H, m, 14, 14*), 2.8-2.47 (8H, m), 2.45-2.34 (2H, t), 1.44 (4H, sex), 1.26 (14H, m), 0.97 (6H, t, 16, 16*), 0.88 (3H, t, 1). The shift of protons attached to carbon 1 remains constant for the reactant, product and intermediate, while the protons attached to carbons 14 and 16 showed different shifts for the mono-and di-butoxylated amines SEC and TERT. Therefore, both conversion and selectivity could be calculated. H
1
16*
H 3C
N
H
CH3
14* OH
14 OH
H 3C
16
Figure 4: Fishbone structure of TERT.
Titration: The water content was titrated by the Karl-Fischer method28. Equipment?
4. Results and Discussion Experimental Full conversion was achieved after the addition of 4 times 0.5 equivalents of BO in the temperature range from 120 ºC to 150 ºC. A typical development of the concentration profiles up until full conversion of DDA is presented in Figure 5. BO was assumed to be primarily present in the liquid phase, as explained in the experimental section. The vapor pressure of BO at the reaction temperatures was at all times around 4%mol of the total BO amount that could have been present in the gas phase.
11
Figure 5: Concentration profile of full conversion experiment at 140ºC. The amount of moles of BO in the gas phase was constantly 4% of the total amount BO in the reactor.
The auto-catalytic character of the reaction becomes apparent when comparing the rate of change in DDA concentration in Figure 5 between the first and subsequent additions of BO. The concentration of DDA decreases faster for the later additions of BO, which is a strong indication of the auto-catalytic character. Another interesting feature apparent from Figure 5 is that by adding 0.5 mole of BO to pure DDA first a significant amount of SEC was formed after already 30 minutes. Only after 150 minutes reaction time, significant amounts of TERT were detected. Consequently, the reaction of DDA to SEC is faster than the reaction SEC to TERT. This may be possibly explained by the increase in steric hindrance in SEC, compared to DDA. Ingberman et al.29 state that a primary and a secondary amino group react equally with ethylene and propylene oxide. However, we show here that this might not be the case for the reaction between DDA and BO. While at 140 ºC a total reaction time of 450 minutes was needed to reach full conversion, at 150 ºC the total reaction time was 40% longer (Figure 6).
12
Figure 6: Concentration profile of full conversion experiment at 150 ºC.
In this case full conversion was not reached after four BO additions, and a 5th addition was necessary to achieve full conversion. A comparison of the initial part of the reaction (first addition of BO), showed that the reaction at 150°C is 30 minutes slower than at 140°C. Moreover, up to 150ºC the mole balance was close to 98%mol, and by-product formation was negligible. At 150ºC the formation of side products (e.g. BO oligomers and others), observed in GC-MS (> 350 g/mole, see SI Figure S2) accounted for 4% of the mole balance. Such side products were formed in larger amounts at 160 ºC. The addition of catalytic amounts (1-3%mol) of a range of organic species containing different functional groups were subsequently investigated (spiking experiments). Such investigation would provide insights into which type of functionality is able to promote or inhibit the reaction, also giving insight into the mechanism. The catalytic effect of 12 different species on the reaction rate of BO with DDA at 140 ºC is shown in Figure 7. The reaction rate at 75% conversion of added BO (1st addition) was compared by taking the fractional deviation with the un-spiked experiment (blank). The normalized standard deviation of the blank experiment was ±0.05. 13
Figure 7: Deviation of spiked runs from a blank are shown. The influence of catalytic amounts of each species was tested in an initial step experiment. The relative deviation from a blank “standard’ run at 140 ºC was chosen as base line. Reaction rate promoting species show a positive deviation while reaction rate inhibiting species show a negative deviation.
Interestingly, dodecylamide and dodecylnitrile had a significant inhibiting effect on the reaction rate. These compounds are precursors to primary amines in the industrial process and are therefore common impurities in industrial sources of dodecylamine. Dodecylcarbamate, which is another possible impurity, had very little effect on the reaction rate. It is assumed that this compound decomposes at high temperatures back to the original amine and CO2. DABCO is an example of a strong proton accepting functionality often used as catalyst, which has no significant effect on the rate, just like salicylaldehyde. A remarkable trend is the accelerating effect of all the proton-donating groups, which reflects the catalyzing effect of hydroxyl groups in the reaction. As previously acknowledged, the hydroxyl group likely interacts with the epoxide ring, destabilizing it (Figure 8a). More importantly, however, secondary hydroxyalkylamines such as SEC, diethanolamine, and ethylaminoethanol, which contain both an amine and hydroxyl group showed the strongest effect in accelerating the reaction. Such hydroxy-alkylamines contain a proton donor and the amine at close distance from each other, which could kinetically favor a
14
“concerted” ring activation followed by an amine attack (Figure 8b). This type of intermediate would be the bimolecular analogue of the termolecular transition state proposed by Smith.3 Another explanation could be the stabilization of anionic intermediates formed upon reaction of the amine with BO, which would require the presence of two amine moieties (Figure 8c). Comparable intermediates are known to form during the synthesis of carbamates.30
R
R
O
O
H
H O
O
+
a) Destabilization of epoxide ring by binding to an hydroxyl groups proton OH
O
O
H
N
NH2
2
+
O
NH2 R
R
H 2C
+
H N
R
R
OH
O NH3
+
R
H N
R
b) Stabilizing further the
R
NH3 +
H N
R
c) Ionic complex
hydrogen transfer Figure 8: Possible explanation for enhanced auto-catalytic character of SEC.
In a comparable study focusing on the propoxylation of carbon dioxide catalyzed by alkanolamines, it was seen that di-alkanolamines do not lead to a significant rate increase compared to the amines with only one hydroxyl group31.This finding agrees with our study, where diethanolamine (2 hydroxyl groups) has the same effect as ethylaminoethanol and SEC (1 hydroxyl groups). According to this, TERT would be expected to have the same effect as SEC. However, a much weaker catalytic effect was observed for TERT, which suggests a considerable influence of the steric hindrance on its catalytic activity. 15
The hypothesis of Enikolopiyan et al., that the reaction can proceed in the absence of water,22 was tested by determining the reaction rate after drying the reactants. Water is known as a catalyst32, yet it also decreases the selectivity towards the desired products. BO was dried from a water content of 640 ppm to 200 ppm water, while DDA had a constant water content of 60 ppm. Only a minor effect of +0.2 in Figure 7 was observed for the dried BO, which means that water has, at this low concentration, no significant influence, in line with the experimental results from Enikolopiyan et al.22.
Modelling The experimental data strongly suggests an auto-catalytic reaction where both intermediate- and end product catalyze the reactions. The long reaction time to form the intermediate SEC is in agreement with a secondary, co-dependent reaction order for both reactants, in agreement with the literature33. Using the information gained about the kinetics and catalytic species a full kinetic model was first constructed.
𝐷𝐷𝐴 + 𝐵𝑂 𝐷𝐷𝐴 + 𝐵𝑂 𝑆𝐸𝐶 + 𝐵𝑂
(4.1)
𝑆𝐸𝐶 𝑆𝐸𝐶, 𝑇𝐸𝑅𝑇
𝑆𝐸𝐶, 𝑇𝐸𝑅𝑇
𝑆𝐸𝐶
(4.2a, 4.2b)
𝑇𝐸𝑅𝑇
(4.3a, 4.3b)
With the corresponding rate equations: 𝑟𝑖𝑛𝑖𝑡
= 𝑘𝑖𝑛𝑖𝑡
(4.1)
[𝐷𝐷𝐴] [𝐵𝑂]
𝑟𝑐𝑎𝑡,𝑠𝑒𝑐,1 = 𝑘𝑐𝑎𝑡_𝑠𝑒𝑐,1 [𝐷𝐷𝐴] [𝐵𝑂] [𝑆𝐸𝐶]
(4.2a)
𝑟𝑐𝑎𝑡,𝑠𝑒𝑐,2 = 𝑘𝑐𝑎𝑡_𝑠𝑒𝑐,2 [𝑆𝐸𝐶] [𝐵𝑂] [𝑆𝐸𝐶] `
(4.2b)
𝑟𝑐𝑎𝑡,𝑡𝑒𝑟𝑡,1 = 𝑘𝑐𝑎𝑡_𝑡𝑒𝑟𝑡,1[𝐷𝐷𝐴] [𝐵𝑂] [𝑇𝐸𝑅𝑇]
(4.3a)
𝑟𝑐𝑎𝑡,𝑡𝑒𝑟𝑡,2 = 𝑘𝑐𝑎𝑡_𝑡𝑒𝑟𝑡,2[𝑆𝐸𝐶] [𝐵𝑂] [𝑇𝐸𝑅𝑇]
(4.3b)
16
The non-catalyzed reaction of BO and SEC was excluded from the beginning from this model, as SEC is itself a catalyzing intermediate. For each rate coefficient the pre-exponential factors and activation energies according to Arrhenius law (eq.4.4) were fitted to the experimental data using MATLAB’s nonlinear fitting, using the non-linear least-square fitting procedure with the trust-region-reflective algorithm towards the experimental data. In the fitting routine of Matlab, the mole balance was solved by the ODE15s solver using a relative tolerance of 10-8 (see SI for full mole balance). For each addition the mole balance was updated using new initial concentration values and volumes including the loss in sample volume and addition of BO volume.
𝑘𝑗
𝐸𝐴,𝑗 𝑅𝑇
(― ) =𝐴e 𝑗
𝑗 = 𝑖𝑛𝑖𝑡; 𝑐𝑎𝑡𝑠𝑒𝑐,1,2; 𝑐𝑎𝑡𝑡𝑒𝑟𝑡,1,2
(4.4)
The confidence intervals were calculated via the nlparci function, which returns the 95% confidence intervals for the nonlinear least squares parameter estimates of the non-linear fitting. The coefficient of determination was calculated by equation 4.5:
𝑆𝑆𝑟𝑒𝑠
𝑅2 = 1 ― 𝑆𝑆𝑡𝑜𝑡
(4.5)
Where SSres corresponds to the sum of squares of residuals between fitted and experimental values, and SStot to the sum of total squares, which gives an average of the experimental data variance. In a first step only rate equations (4.1) and (4.2a) where fitted to the first BO addition for each temperature. It was assumed that no TERT was formed in amounts sufficient to catalyze the reaction, as verified by the experiments. A coefficient of 0.975 is reached for this fit. The so defined variables were used in the overall fit with all experimental sets. For the fitting of equations (4.2b), (4.3a) and (4.3b) to all data sets, a determination coefficient of 0.951 was found. The summarized Arrhenius parameters are shown in Table 2. 17
Table 2: pre-exponential factors and activation energy valid from 120ºC to 150ºC for full kinetic model
Pre-exponential factor Ainit
95%-confidence interval
Activation Energy
95%-confidence interval
[10-1 m3 mol-1 s-1]
3.6 ± 0.4
EA,init
[kJ mol-1]
60.0 ± 0.1
Acat_sec,1 [10-5 m6 mol-2 s-1]
1.8 ± 0.1
EA,cat_sec,1 [kJ mol-1]
45.6 ± 0.3
Acat_sec,2 [10-3 m6 mol-2 s-1]
3.5± 0.5
EA,cat_sec,2 [kJ mol-1]
63.7± 0.1
Acat_tert,1 [10-4m6 mol-2 s-1]
4.8± 0.1
EA,cat_tert,1 [kJ mol-1]
52.0± 0.1
Acat_tert,2 [10-2 m6 mol-2 s-1]
1.5± 0.2
EA,cat_tert,2 [kJ mol-1]
64.2± 0.1
An exemplary fitting for the data set at 120 ºC is shown in Figure 9. The experimental fittings of temperature from 130 ºC to 150 ºC are presented in the SI in Figure S4.
Figure 9: Auto-catalytic model fit for experimental set at 120 ºC.
The fitted curves represent reasonably the concentration of the reaction components over the reaction progress. A possible explanation for the residual error between the model and the data is due to the 18
analytic difficulties of measuring the products. SEC and TERT are alcohol amines, which tend to interlink with each other and absorb onto the GC column as mentioned in the experimental section. The kinetic coefficients can be further validated by using the spiking experiments. For the initial run at 140 ºC with additional pure TERT as catalytic agent, a rate constant of 4.67 10 -11m3 mol-1 s-1 for equation 4.3a can be achieved. The full rate constant model gives for equation 4.3a a rate constant of 4.16 10-11m3 mol-1 s-1. Comparing the rate coefficients for the reaction catalyzed by SEC and TERT at 140 ºC, the same ratio of one for each reaction step was found. A ratio of one corresponds to only one hydroxyl group being able to participate as catalyst, independently of the molecule having one (SEC) or two (TERT) hydroxyl groups. This correlation offers the opportunity to simplify the catalyst terms due to their similarity. In addition, a simplified model saves computational time for high intensive simulation for complex reaction systems. Each equation pair 4.2a – 4.3a, and 4.2b – 4.3b, were therefore combined into equation 4.6-4.7 using Kratio = 1. The resulting set is shown below: (4.1)
𝑟𝑖𝑛𝑖𝑡
= 𝑘𝑖𝑛𝑖𝑡
𝑟𝑐𝑎𝑡_1
= 𝑘𝑐𝑎𝑡_1 [𝐷𝐷𝐴] [𝐵𝑂] ([𝑆𝐸𝐶] + 𝐾𝑟𝑎𝑡𝑖𝑜[𝑇𝐸𝑅𝑇])
(4.6)
𝑟𝑐𝑎𝑡_2
= 𝑘𝑐𝑎𝑡_2 [𝑆𝐸𝐶] [𝐵𝑂] ([𝑆𝐸𝐶] + 𝐾𝑟𝑎𝑡𝑖𝑜[𝑇𝐸𝑅𝑇])
(4.7)
[𝐷𝐷𝐴] [𝐵𝑂]
The model presented by Sirovski et al. 13 is similar to the above, but the catalytic activity of TERT is considered double of SEC (KRatio = 2). However, besides being unlikely from a sterical point of view that both hydroxyl groups act at the same time as catalyst, our data further suggest that this is not the case for BO. If it is true for ethylene oxide remains uncertain due to the difference in substitution in our work (BO) compared to Sirvoski’s work with ethylene oxide. Roshan et al. 31 studied the cycloaddition of CO2 to propylene oxide using amines and alcohol amines as catalyst. For their system they also demonstrated that two hydroxyl groups on one molecule of alcohol amine have the same or even lower catalytic activity due to steric hindrance, very similar to our observations.
19
Table 3: pre-exponential factors and activation energy valid from 120 ºC to 150 ºC.
Pre-exponential factor
95%-confidence
Activation Energy
interval
95%-confidence interval
Ainit [10-1 m3 mol-1 s-1]
3.6 ± 0.4
EA,init [kJ mol-1]
60.0 ± 0.1
Acat_1 [10-6 m6 mol-2 s-1]
4.3 ± 0.9
EA,cat_1 [kJ mol-1]
39.3 ± 0.5
Acat_2 [10-4 m6 mol-2 s-1]
1.5 ± 0.5
EA,cat_2 [kJ mol-1]
50.8 ± 1.5
With the coefficients presented in Table 4 for the simplified model an overall determination coefficient of 0.94 can be achieved, only a little lower than for the full kinetic model (value = 0.95 with four parameters more). The fitted curves of the simplified model are given in Figure S5 in the SI. In the spiking experiments with SEC and TERT we found that SEC was twice as active as TERT. Using this in the model as a factor of 0.5 for TERT led to slightly worse results (determination coefficient of 0.93). Possibly, the catalytic activity of TERT is enhanced in presence of SEC.
The parity plot of the simplified model versus the experimental concentrations shown in Figure 10 clearly demonstrate that the model predicts well the concentration profiles of the components.
20
Figure 10: Parity plot of model fit to all experimental data sets at all temperatures
The highest deviations are found for the TERT and SEC concentrations. As discussed before, this is attributed to the limited accuracy of the GC and NMR analysis for estimation of these concentrations.
Figure 11: Auto-catalytic model fit to the experimental butoxylation of OA at 140 ºC.
21
Experiments were also conducted to validate the proposed reaction mechanism for alkyl amines in general. Figure 11 shows the result of experiments with octylamine (OA) instead of DDA, showing that the reaction mechanism, including the auto-catalytic effect and the low initial rate, appears valid for alkyl amines in general. Comparing the two amines, OA was 3% faster but in the order of deviation equal to DDA. The fitted rate constants for this experiment are given in Table 4Table .
Table 4: Kinetic constants given for the reaction of OA with BO at 140 ºC.
Rate constants
95%-confidence interval
kinit [10-9 m3 mol-1 s-1] 1.5 ± 0.3 kcat,1 [10-10m6 mol-2 s-1]
7.0 ± 0.9
kcat,2 [10-10 m6 mol-2 s-1]
4.1 ± 0.7
The accuracy of this fit is lower when looking at the confidence intervals and the coefficient of determination of 0.92 due to the lower amount of experimental points to fit. The model was further validated by comparing the model prediction to an additional spiking run, using a reaction mixture enriched with SEC (2%mol) for the initial step. The result of this can be seen in Figure 12. The model accurately describes the experiment, further confirming the validity of the model.
22
Figure 12: Spiked run modelling for 140 ºC and an addition ratio of 0.5:1 mole of BO: DDA.
5. Conclusions The mechanism and kinetics of the butoxylation of dodecylamine (DDA) have been studied. It was found that a range of proton donating groups such as alcohols can catalyze the reaction. Compounds containing both a hydroxyl and an amine group (alkanolamines) are even better catalysts for this process. Both the intermediate secondary butanolamine (SEC) and the product tertiary di-butanolamine (TERT) therefore have an auto-catalytic effect on the reaction rate. The catalytic effect of TERT is not higher than for SEC, as demonstrated by our spiking experiments and the fitting model. We believe this is due to the increased steric hindrance of TERT compared to SEC. Water could be excluded from the kinetic model because it had no significant catalyzing effect with different ppm level concentrations. In addition, spiking experiments demonstrated that a significant inhibition of the reaction rate occurs upon addition of dodecylnitrile and dodecylamide, which are the two most common impurities in the industrial sources of DDA.
23
Two different rate laws were derived that incorporate the auto-catalytic effect of the proton-donating groups of the intermediate and final products. The first includes separate terms for the effect of SEC and TERT in the reaction, while the second represents a simplified, yet almost as accurate model, where SEC and TERT are lumped together. It was found that this rate law satisfactorily describes the experimental data, assuming an Arrhenius type of dependency of the kinetic constants on temperature.
6. Acknowledgements This research was carried out within the HighSinc program – a joint development between Nouryon Specialty Chemicals and the Department of Chemical Engineering and Chemistry from Eindhoven University of Technology.
Abbreviations
BO, butylene oxide; DDA, dodecylamine; OA, octylamine; SEC, mono-(hydroxybutyl) dodecylamine; TERT, bis-( hydroxybutyl) dodecylamine, GC, gas chromatography; MS, mass spectrometer, FID, flame ionization detector; CSTR, continuous stirred tank reactor
References (1)
Schönfeldt, N. Surface Active Ethylene Oxide Adducts; Elsevier, 1969.
(2)
Fatty Amines Market Analysis 2018-2025 https://www.grandviewresearch.com/industry-analysis/fatty-amines-market (accessed Apr 18, 2018).
(3)
Smith, I. T. The Mechanism of the Crosslinking of Epoxide Resins by Amines. Polymer 1961, 2, 95–108.
(4)
Shechter, L.; Wynstra, J.; Kurkjy, R. P. Glycidyl Ether Reactions with Amines. Ind. Eng. Chem. 1956, 48 (1), 94–97.
24
(5)
Holubka, J. W.; Bach, R. D.; Andres, J. L. Theoretical Study of the Reactions of Ethylene Oxide and Ammonia: A Model Study of the Epoxy Adhesive Curing Mechanism. Macromolecules 1992, 25 (3), 1189–1192. https://doi.org/10.1021/ma00029a028.
(6)
Di Serio, M.; Vairo, G.; Iengo, P.; Felippone, F.; Santacesaria, E. Kinetics of Ethoxylation and Propoxylation of 1- and 2Octanol Catalyzed by KOH. Ind. Eng. Chem. Res. 1996, 35 (11), 3848–3853. https://doi.org/10.1021/ie960200c.
(7)
Chiu, Y. N.; Naser, J.; Easton, A.; Ngian, K. F.; Pratt, K. C. Kinetics of a Catalyzed Semi-Batch Ethoxylation of Nonylphenol. Chem. Eng. Sci. 2010, 65 (3), 1167–1172. https://doi.org/10.1016/j.ces.2009.09.080.
(8)
O’Lenick, A. Group Selectivity of Ethoxylation of Hydroxy Acids. Chemist 2002, 79, 9–16.
(9)
Komori, S.; Karaki, T. The Reaction between Alkylamine and Ethylene Oxide without Any Catalyst. J. Soc. Chem. Ind. Jpn. 1959, 62 (4), 538–542. https://doi.org/10.1246/nikkashi1898.62.4_538.
(10)
Rupp, M.; Ruback, W.; Klemm, E. Alcohol Ethoxylation Kinetics: Proton Transfer Influence on Product Distribution in Microchannels. Chem. Eng. Process. Process Intensif. 2013, 74, 187–192. https://doi.org/10.1016/j.cep.2013.09.006.
(11)
Wu, Y.; Liu, Q.; Su, X.; Mi, Z. Effect of Solvents on Propylene Epoxidation over TS-1 Catalyst. Front. Chem. China 2008, 3 (1), 112–117. https://doi.org/10.1007/s11458-008-0007-2.
(12)
Brown, W.; Foote, C.; Iverson, B.; Anslyn, E. Organic Chemistry; Cengage Learning, 2008.
(13)
Sirovski, F.; Mulyashov, S.; Shvets, V. Large-Scale Fatty Amine Ethoxylation Reactor: A Dynamic Model. Chem. Eng. J. 2006, 117 (3), 197–203. https://doi.org/10.1016/j.cej.2005.12.005.
(14)
Sanders, H. L.; Braunwarth, J. B.; McConnell, R. B.; Swenson, R. A. Ethoxylation of Fatty Amines. J. Am. Oil Chem. Soc. 1969, 46 (3), 167–170.
(15)
Sundaram, P. K.; Sharma, M. M. Kinetics of Reactions of Amines with Alkene Oxides. Bull. Chem. Soc. Jpn. 1969, 42 (11), 3141–3147.
(16)
Kothari, P. J.; Sharma, M. M. Kinetics of Reaction between CS2 and Amines. Chem. Eng. Sci. 1966, 21 (5), 391–396. https://doi.org/10.1016/0009-2509(66)85049-2.
(17)
Gaertner, V. R. Alkyl-2, 3-Epoxypropylamines: Cyclodimerization and Related Eight-Membered Ring Closures. Tetrahedron 1967, 23 (5), 2123–2136.
(18)
Lang, R. F.; Parra-Diaz, D.; Jacobs, D. Analysis of Ethoxylated Fatty Amines. Comparison of Methods for the Determination of Molecular Weight. J. Surfactants Deterg. 1999, 2 (4), 503–513.
(19)
Laird, R. M.; Parker, R. E. The Mechanism of Epoxide Reactions. Part XII. Reactions of Ethylene Oxide with Alcohols in the Presence of Sodium Alkoxides and of Tertiary Amines. J. Chem. Soc. B Phys. Org. 1969, 1062–1068.
(20)
Lovett, E. G. Zwitterionic Quaternary Ammonium Alkoxides: Organic Strong Bases. J. Org. Chem. 1991, 56 (8), 2755– 2758. https://doi.org/10.1021/jo00008a034.
25
(21)
Trejbal, J.; Petrisko, M.; Pašek, J. Reactions of Ethoxylates of Ethylenediamine and Piperazin over ZSM-5 Zeolithe. Pet. Coal 2012, 54 (4), 340–347.
(22)
Enikolopiyan, N. S. New Aspects of the Nucleophilic Opening of Epoxide Rings. Pure Appl. Chem. 1976, 48 (3), 317– 328. https://doi.org/10.1351/pac197648030317.
(23)
Tiltscher, V. H. Kinetik Und Mechanismus Der Anlagerung von Propylenoxid an Diäthylamin Unter Ausschluß Protonenaktiver Verbindungen. Angew. Makromol. Chem. 1972, 25 (1), 1–14. https://doi.org/10.1002/apmc.1972.050250101.
(24)
Aresta, M.; Quaranta, E. Mechanistic Studies on the Role of Carbon Dioxide in the Synthesis of Methylcarbamates from Amines and Dimethylcarbonate in the Presence of CO2. Tetrahedron 1991, 47 (45), 9489–9502. https://doi.org/10.1016/S0040-4020(01)80894-2.
(25)
Cai, X.; Zheng, J. L.; Wärnå, J.; Salmi, T.; Taouk, B.; Leveneur, S. Influence of Gas-Liquid Mass Transfer on Kinetic Modeling: Carbonation of Epoxidized Vegetable Oils. Chem. Eng. J. 2017, 313, 1168–1183. https://doi.org/10.1016/j.cej.2016.11.012.
(26)
Rupp, M.; Ruback, W.; Klemm, E. Octanol Ethoxylation in Microchannels. Chem. Eng. Process. Process Intensif. 2013, 74, 19–26. https://doi.org/10.1016/j.cep.2013.09.012.
(27)
Pinchas, S.; Ben-Ishai, D. The Carbonyl Absorption of Carbamates and 2-Oxazolidones in the Infrared Region. J. Am. Chem. Soc. 1957, 79 (15), 4099–4104. https://doi.org/10.1021/ja01572a035.
(28)
Scholz, E. Karl Fischer Titration: Determination of Water; Chemical Laboratory Practice; Springer-Verlag: Berlin Heidelberg, 1984.
(29)
Ingberman; Walton. Low Toxicity Aliphatic Amines as Crosslinking Agents for Polyepoxy Resins. J. Polym. Sci. 1957, 28 (117).
(30)
McCann, N.; Phan, D.; Wang, X.; Conway, W.; Burns, R.; Attalla, M.; Puxty, G.; Maeder, M. Kinetics and Mechanism of Carbamate Formation from CO2(Aq), Carbonate Species, and Monoethanolamine in Aqueous Solution. J. Phys. Chem. A 2009, 113 (17), 5022–5029. https://doi.org/10.1021/jp810564z.
(31)
Roshan, K. R.; Kim, B. M.; Kathalikkattil, A. C.; Tharun, J.; Won, Y. S.; Park, D. W. The Unprecedented Catalytic Activity of Alkanolamine CO2 Scrubbers in the Cycloaddition of CO2 and Oxiranes: A DFT Endorsed Study. Chem Commun 2014, 50 (89), 13664–13667. https://doi.org/10.1039/C4CC04195J.
(32)
Azizi, N.; Saidi, M. R. Highly Chemoselective Addition of Amines to Epoxides in Water. Org. Lett. 2005, 7 (17), 3649– 3651. https://doi.org/10.1021/ol051220q.
(33)
Os, N. M. van. Nonionic Surfactants: Organic Chemistry; CRC Press, 1997.
26
Highlights
Experimental study on butoxylation of fatty primary amines Reaction mechanism elucidated leading to an accurate kinetic model Auto-catalytic reaction is the driving force for high conversions Reaction rate strongly enhanced by bidentate action of the alcohol and amine group
27
S1385-8947(19)32349-6 https://doi.org/10.1016/j.cej.2019.122939 CEJ 122939
To appear in:
Chemical Engineering Journal
Received Date: Revised Date: Accepted Date:
4 April 2019 19 September 2019 22 September 2019
Please cite this article as: P. Müller, R.D.E. Krösschell, W. Winkenwerder, J. van der Schaaf, The Butoxylation of Dodecylamine: Reaction Mechanism and Kinetics, Chemical Engineering Journal (2019), doi: https://doi.org/ 10.1016/j.cej.2019.122939
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2019 Published by Elsevier B.V.
The Butoxylation of Dodecylamine: Reaction Mechanism and Kinetics
P. Müller1, R.D.E. Krösschell1, W. Winkenwerder2, J. van der Schaaf1,*, 1 Laboratory
of Chemical Reactor Engineering, Department of Chemical Engineering and Chemistry, Eindhoven University of
Technology, P.O. Box 513, 5600MB, the Netherlands 2 Nouryon,
Brewster, New York 10509, United States* Corresponding author: [email protected]
Abstract The reaction mechanism and kinetics of the butoxylation of dodecylamine were investigated experimentally using a semi-batch, fully liquid process at temperatures ranging from 120 ºC to 150 ºC and varying stoichiometric ratios of the reactants. In addition, the catalytic effect of adding 1-3%mol of twelve different functional organic species containing hydroxyl, aldehyde, amine, and/or amide functional groups was studied. It was found that only proton donating groups increase the observed reaction rate. In particular hydroxyl groups in combination with an amine group resulted in a strong acceleration of the reaction. Since the reaction intermediates mono-and dibutoxylated amines have this combination, an autocatalytic reaction mechanism and corresponding rate law are proposed. The kinetic constants were fitted to the experimental data as a function of temperature, following an Arrhenius type of dependency. The results from the model describe the experimental data with 95% accuracy. Moreover, the results show that the butoxylation of dodecylamine, as a model reactant for a short substituted ethylene oxide, with fatty amines, follows the same mechanism and has similar kinetics as epoxide hardening reactions with amines.
1. Introduction Ethoxylated and propoxylated fatty amines are widely used as nonionic surfactants, leading to a steady increase in industrial demand over the past six decades.1,2 These are key components in products of 1
numerous industrial processes, e.g. oil drilling, paper manufacturing, textile as well as fiber manufacturing, paints, dyes and plastics. These products are produced industrially by reaction of amines with 2 moles of alkene oxides (e.g. ethylene or propylene oxide), giving a tertiary amine of the type shown in Figure 1, and further polymerized using a strong basic catalyst,3–5 to provide a polyalkoxilated amine with 5-50 molecules of ethylene oxide and/or propylene oxide attached, which find use as surfactants. Compared to this well-studied polymerization, the initial reaction of fatty amines with two moles of alkoxide is rarely investigated and not fully understood (in this article the term epoxide is used to specifically indicate a three-member ring with an oxygen atom; alkoxide is instead used to indicate all epoxide-containing species, i.e. substituted epoxides). Although the mechanism of the reaction between ethylene oxide and fatty alcohols,6 phenols7 and acids8 has been thoroughly investigated, less attention has been directed to the industrial reaction with fatty amines. Komori et al. were the first in the 1950s to experimentally demonstrate that fatty amines can react with two molar equivalents of ethylene oxide without a catalyst, 9 following an SN2 mechanism for the substitution of both protons of the amine with the alkoxide (Figure 1).3,9 B H H R1
NH
SN2
O
B
R1 R1
O
R2 N H
R2
OH N
R2
R2
HO OH R2
R1 = C8-C18
Figure 1: Nucleophilic attack of a fatty amine to alkoxides, like butylene- or propylene oxide. The overall process involves two successive SN2 reactions with two moles of alkoxide, finally leading to a tertiary amine.
In the above mechanism, the proton transfer steps between the amine and the newly formed hydroxyl group is not fully understood yet. Often a carrier (B) like water or other residues are conceptually introduced to explain this proton transfer.10,11 The reaction is known for being auto-catalytic.1,12–14 The auto-catalytic effect is believed to be due to the destabilization of the ethylene oxide ring by the hydroxyl 2
groups of the products, but this functionality could also act as proton transfer agents. The two effects cannot be clearly separated because of their co-existence. In this article we investigate the reaction between fatty amines and butylene oxide. Kinetic profiles, reaction modeling, and spiking experiments with a range of differently functionalized species are reported, which provided important information about the process and the mechanism of the reaction.
2. Theory The nucleophilic attack of the amine group towards the epoxide ring depends on three main types of effects: steric, polar, and catalytic.6,9,14,15 The substitution pattern, especially of the alkoxides, has steric and polar effects, and a decrease in the reaction rate compared to ethylene oxide is observed for substituted or more polar groups, such as epichlorohydrin15. Furthermore, a substitution on one carbon of the alkoxide ring reduces the rate because only the less substituted methylene carbon participates in the reaction.15–17 Sundaram et al.15 correlated the reaction rate to the degree of substitution in the series ethylene-, propylene- and 1,2-butylene oxide, with several amines, concluding that the different alkoxides follow the same reaction pathway. The catalytic effect is most likely related to the oxide ring opening reactions by a proton donor, such as water, acids or alcohols15,17. The proton destabilizes the ring by bonding with the oxygen, thus reducing the activation energy for the alkoxylation reaction. Unfortunately, such catalytic effect also promotes the undesired polymerization of the alkoxides to polyglycols, commonly observed as side products in these processes.18 This tradeoff between rate acceleration and side product formation of traditional catalysts mostly results in the investigation of non-catalyzed reactions, with a consequent lack of knowledge on the catalytic aspects of the process. Additions to the aforementioned basic SN2 mechanism were proposed. Sanders et al.14 related freely formed bases at low temperatures of 90 ºC to a highly catalytic quaternary ammonium species, formed by reaction with a third equivalent of ethylene oxide. This thermally unstable ion would increase the rate significantly after being formed. This mechanism is however controversial.19–21 The second mechanism discussed in a review by 3
Smith3 begins with the assumption that the epoxide ring must first be destabilized to start the reaction4 (Figure S1). In non-catalyzed reactions, trace amounts of water are thought to act as the catalyst in the initial step of the reaction; subsequently, the reaction products catalyze the ring opening. A termolecular step between the proton donor, alkoxide and amine is proposed as rate limiting step. However, in more recent investigations this thesis is opposed.21,22 In particular, Enikolopiyan et al. were able to prove that the reaction can proceed in the absence of water, using ultra-pure and anhydrous reactants.22 Moreover, elementary termolecular steps are highly unlikely from a kinetic and statistical point of view and are thus highly uncommon. As an alternative to this termolecular reaction, a dipolar intermediate was suggested23 (Figure S1). Sirovski et al. described the reaction of ethoxylation of dodecylamine with a mechanism consisting of an initial SN2 step followed by catalyzed terms with the sum of the products’ hydroxyl groups and water as external catalyst. However, there is no mechanistic study of how these terms were found.13 There is thus a clear gap in the research regarding the mechanism and kinetics of the alkoxylation of fatty amines (including ethylene-, propylene- or butylene oxide), which we are striving to close. In this work, the mechanism and, more specifically, the driving force behind the auto-catalytic behavior was illuminated through investigating the influence on the reaction rate of catalytic amounts of twelve different organic species, containing hydroxyl, aldehyde, nitrile, amine and amide functional groups. These spiking experiments provided invaluable information on the influence of the different functionalities in accelerating or inhibiting the reaction, and their consequent connection to the process and the mechanistic aspects of the reaction.
3. Experimental Section Chemicals. As in Sirovski et al. work,13 dodecylamine (DDA) was chosen as model compound, as it represents a large fraction of most fatty amine mixtures employed industrially, typically ranging from C8 to C18 alkyl chains. Octylamine (OA) was tested as alternative for DDA to demonstrate the generality of 4
the mechanism. Butylene oxide (BO) shows a lower hazard level than ethylene and propylene oxide, and was therefore used as a safer and more convenient model reactant. The chemicals used for the experiments and analytics are listed in Table 1. Among the spiking chemicals, the pure reaction products and methyl dodecylcarbamate are not commercially available and were made in the lab. The final product (bis(hydroxybutyl)dodecylamine, TERT) could be obtained in 99%mol, while the intermediate secondary amine (mono-(hydroxybutyl)dodecylamine, SEC) could be obtained at a concentration of 75%mol from the reaction crude. Dodecyl methyl-carbamate was produced at ambient conditions from dodecylamine and carbon dioxide from air according to the method presented by Aresta and Quaranta24. The purity of the produced carbamates was analyzed with GC to be greater than 90%mol.
Table 1: list of purchased chemicals. The model reaction was dodecylamine (DDA) with 1,2-butylene oxide (BO). Octylamine (OA) was used as alternative to DDA. The listed spiking chemicals were only added in the spiking runs additional to the reaction crude.
Name
Supplier
CAS-number
Purity
Reaction chemicals Dodecylamine (DDA)
Acros Organics
124-22-1
≥95.0%
1,2-Butylene oxide (BO)
EMD Millipore
106-88-7
≥99.0%
106-88-7
≥99.0%
Corporation Octylamine (OA)
Merck KGaA
Analytical chemicals d-Chloroform
Sigma-Aldrich
865-49-6
≥99.96 atom%
Acros Organics
108-88-3
≥99.7%
+0.03%TMS Toluene
Spiking chemicals Methanol
Acros Organics
67-56-1
≥98.0%
Butan-2-ol
Alfa Aesar
78-92-2
≥99.0% 5
2-(Ethylamino)-ethanol
Alfa Aesar
110-73-6
≥98.0%
Diethanolamine
Alfa Aesar
111-42-2
≥99.0%
Dipropylamine
Merck KGaA
142-84-7
≥99.0%
DABCO (1,4-
Sigma-Aldrich
280-57-9
≥99.0%
Salicylaldehyde
Alfa Aesar
90-02-8
≥99.0%
Dodecylamide
TCI Chemicals
1120-16-7
≥96.0%
Dodecylnitrile
TCI Chemicals
2437-25-4
≥96.0%
diazabicyclo[2.2.2]octan)
Methyl dodecylcarbamate
/
/
90%
SEC
/
/
75%
TERT
/
/
99%
Set-up. The set-up is schematically depicted in Figure 2. The reactions were performed in a 158 ml Cyclone 075 stainless steel autoclave (Büchi Glassuster) in which pressure and temperature were monitored at a measurement rate of one second. The heat exchanger was an oil bath RE306 from LaudaBrinkman. BO was added into the reactor through a capillary by using a syringe, while samples were taken through the bottom valve.
6
Figure 2: Experimental set-up for kinetic measurements.
Procedure. DDA was preloaded into the reactor at room temperature. Subsequently, the reactor was heated to reaction temperature and vacuum was applied until a pressure below 20 mbar was reached. After the reaction temperature was reached, BO at room temperature was added in approximately three seconds by using a syringe. In a stepwise manner of 0.5 mole of BO per mole of DDA per step. Before each subsequent addition of BO, it was made sure that the previously added amount of BO was completely reacted by monitoring the pressure in the reactor. Due to the high solubility of BO in DDA the system was treated as a homogenous liquid phase reaction of which the vapor pressure is made up nearly entirely of BO. At full conversion of BO the pressure returned close to vacuum. This procedure was chosen to model the industrial semi-batch process and to allow a fast homogenous mixing of the low volume of BO. The stirring rate was kept constant at 500 rpm. Initial tests of stirring speeds, ranging from 100 to 1000 rpm, showed that no mass or heat transfer limitations were present above 250 rpm at the conditions that gave the highest reaction rate. The same procedure was used for the reaction of OA with BO to demonstrate the applicability of the reaction and the model (vide infra).
7
Catalytic effect of functional groups (spiking experiments). The catalytic activity was tested by measuring the reaction rate of a mixture of 0.25 mole of BO and 1 mole of DDA after adding 1-3%mol of each investigated species. The species used comprise the list of spiking chemicals in Table 1, including SEC, TERT, and methyl-dodecylcarbamate. The reaction rates where normalized with respect to the total added amount of spiking chemicals in moles and compared to the non-catalyzed reaction.
Reactor temperature. In an initial screening, it was determined to use a reaction temperature window of 120 ºC to 150 ºC in steps of 10 ºC. Above 150 ºC, the BO ring opens spontaneously and by-products were formed. Below 120 ºC full conversion could not be achieved in a time of 50 hours. The addition of BO at room temperature to the hot reactor led to an initial drop in temperature at the beginning of each experiment. Assuming a very conservative overall heat transfer coefficient of 26 W/m2K, an area available for heat transfer of 1.6·10-3 m2, and an initial measured temperature difference of 16 K to the desired reaction temperature, the estimated time to reheat the reaction mixture is about 10 minutes to reach 95% of thermal equilibrium (see SI Figure S3). This time span is an order of magnitude smaller than the total reaction time of 120 minutes. Therefore, the reactor is assumed to be isothermal over each run. After the thermal equilibrium was reached, the maximum measured temperature variation was 2 K. Note that in the first 10 minutes no samples were taken, so the analysis is only based on data gained from isothermal conditions.
BO concentration in the liquid phase. The molar fraction of BO in the mixture of amines was always less than 0.33 and BO is the only component with a non-negligible vapor pressure (Pv,DDA= 0.024 bar vs. Pv,BO=7 bar at 140 ºC). After detecting a pressure of maximum 1.59 bar in the reactor (after vacuuming the system) it was clear that the reaction cannot be treated as a gas-liquid system as, for example, the carbonation reaction with CO2 in the work of Cai et al. 25, but as a fully liquid system. The measured vapor pressure value was even lower than the value expected from an ideal mixture (2.32 bar), which
8
clearly supports our treatment of this system as a fully liquid one. The concentration of BO in the liquid phase could be determined using equation (3.1). 𝐶𝑖𝑛𝑖𝑡
𝐶𝐵𝑂 = 𝑃𝐵𝑂,𝑖𝑛𝑖𝑡 ∙ 𝑃𝐵𝑂
(3.1)
The liquid phase concentration of BO (CBO) is assumed directly proportional to the pressure in the reactor. The factor of the initial concentration divided by the initial pressure corresponds to the activity coefficient at this point , which was assumed to be constant during reaction. Even if this would not hold, the direct proportionality to BO partial pressure warrants that the kinetic rates are taken proportional to the activity of BO.
Experimental accuracy. A quantitative measure of the experimental error in the concentration was obtained by comparing the deviations in double and one triple rerun of experiments with their respective means. For the BO pressure, the experimental error was found to be on average 1.9% while for DDA, SEC, and TERT, the averaged standard deviation was 2.1%, 3.2%, and 4.2%, respectively. The relative errors were of sufficiently low levels so that the mass balance over the reactor was always maintained within 5% of deviation.
Density of the mixture. The density was studied with the aim of accurately calculating the reaction mixture volume and the mass balance. For several reaction compositions, the density was determined by heating a measured amount of mass of varying amine species compositions excluding BO, in a glass cylinder and measuring the volume. The experiments were conducted between 323 K and 393 K. The intermediate product SEC could not be purified. Therefore a general approach using the total degree of BO moles added per mole DDA (characterized by n in equation 3.2) was chosen, which is comparable to Rupp et al.26
9
The density of pure DDA at room temperature by the model was 801.7 kg/l, equal to the data reported by the supplier. The following empirical correlation (3.2) was found to accurately describe (compare to Figure 3a) the obtained data as a function of mixture composition and temperature, as shown in Figure 3b. 𝜌𝑚𝑖𝑥 = 𝑎(𝑛) ∙ 𝜃(º𝐶) + 𝑏(𝑛), 𝑎(𝑛) = 0.042𝑛3 ― 0.170𝑛2 + 0.255𝑛 ― 0.983,
(3.2)
𝑏(𝑛) = ―3.5𝑛3 ― 6.5𝑛2 + 62.0𝑛 + 821.4
Figure 3a: parity plot density model.
Figure 3b: Mixture densities as a function of temperature. The reaction coordinate n expresses the degree of butoxylation of dodecylamine, ranging from 0 to 2 stoichiometric mole adducts of BO to DDA.
The parameters a and b were determined by a nonlinear surface fit of the experimental data, using MATLAB’s lsqnonlin with the trust-region-reflective algorithm. The effect of BO in the liquid phase was not taken into account in the model, because of its low volumetric concentration of less than 10%vol.
Analytics. GCMS: The samples were analyzed on a GCMS-QP2010 (Shimadzu) with a DB-5 column using helium as carrier gas. For SEC and TERT the mass spectrum was used to identify the corresponding peaks to the retention times in the chromatogram. The calibration of SEC was done by cross referring to NMR data. The mixtures of DDA and its intermediates tend to interlink via the hydroxyl groups, which 10
was visible in GC as broadened bulk peaks with a longer retention time. Due to this the samples had to be analyzed within at most twelve hours27. NMR: A Bruker 400 MHz was used. Figure 4 shows the final product TERT. For this compound, the following spectrum was recorded 1H NMR (400 MHz, CDCl3) δ 3.62-3.48 (2H, m, 14, 14*), 2.8-2.47 (8H, m), 2.45-2.34 (2H, t), 1.44 (4H, sex), 1.26 (14H, m), 0.97 (6H, t, 16, 16*), 0.88 (3H, t, 1). The shift of protons attached to carbon 1 remains constant for the reactant, product and intermediate, while the protons attached to carbons 14 and 16 showed different shifts for the mono-and di-butoxylated amines SEC and TERT. Therefore, both conversion and selectivity could be calculated. H
1
16*
H 3C
N
H
CH3
14* OH
14 OH
H 3C
16
Figure 4: Fishbone structure of TERT.
Titration: The water content was titrated by the Karl-Fischer method28. Equipment?
4. Results and Discussion Experimental Full conversion was achieved after the addition of 4 times 0.5 equivalents of BO in the temperature range from 120 ºC to 150 ºC. A typical development of the concentration profiles up until full conversion of DDA is presented in Figure 5. BO was assumed to be primarily present in the liquid phase, as explained in the experimental section. The vapor pressure of BO at the reaction temperatures was at all times around 4%mol of the total BO amount that could have been present in the gas phase.
11
Figure 5: Concentration profile of full conversion experiment at 140ºC. The amount of moles of BO in the gas phase was constantly 4% of the total amount BO in the reactor.
The auto-catalytic character of the reaction becomes apparent when comparing the rate of change in DDA concentration in Figure 5 between the first and subsequent additions of BO. The concentration of DDA decreases faster for the later additions of BO, which is a strong indication of the auto-catalytic character. Another interesting feature apparent from Figure 5 is that by adding 0.5 mole of BO to pure DDA first a significant amount of SEC was formed after already 30 minutes. Only after 150 minutes reaction time, significant amounts of TERT were detected. Consequently, the reaction of DDA to SEC is faster than the reaction SEC to TERT. This may be possibly explained by the increase in steric hindrance in SEC, compared to DDA. Ingberman et al.29 state that a primary and a secondary amino group react equally with ethylene and propylene oxide. However, we show here that this might not be the case for the reaction between DDA and BO. While at 140 ºC a total reaction time of 450 minutes was needed to reach full conversion, at 150 ºC the total reaction time was 40% longer (Figure 6).
12
Figure 6: Concentration profile of full conversion experiment at 150 ºC.
In this case full conversion was not reached after four BO additions, and a 5th addition was necessary to achieve full conversion. A comparison of the initial part of the reaction (first addition of BO), showed that the reaction at 150°C is 30 minutes slower than at 140°C. Moreover, up to 150ºC the mole balance was close to 98%mol, and by-product formation was negligible. At 150ºC the formation of side products (e.g. BO oligomers and others), observed in GC-MS (> 350 g/mole, see SI Figure S2) accounted for 4% of the mole balance. Such side products were formed in larger amounts at 160 ºC. The addition of catalytic amounts (1-3%mol) of a range of organic species containing different functional groups were subsequently investigated (spiking experiments). Such investigation would provide insights into which type of functionality is able to promote or inhibit the reaction, also giving insight into the mechanism. The catalytic effect of 12 different species on the reaction rate of BO with DDA at 140 ºC is shown in Figure 7. The reaction rate at 75% conversion of added BO (1st addition) was compared by taking the fractional deviation with the un-spiked experiment (blank). The normalized standard deviation of the blank experiment was ±0.05. 13
Figure 7: Deviation of spiked runs from a blank are shown. The influence of catalytic amounts of each species was tested in an initial step experiment. The relative deviation from a blank “standard’ run at 140 ºC was chosen as base line. Reaction rate promoting species show a positive deviation while reaction rate inhibiting species show a negative deviation.
Interestingly, dodecylamide and dodecylnitrile had a significant inhibiting effect on the reaction rate. These compounds are precursors to primary amines in the industrial process and are therefore common impurities in industrial sources of dodecylamine. Dodecylcarbamate, which is another possible impurity, had very little effect on the reaction rate. It is assumed that this compound decomposes at high temperatures back to the original amine and CO2. DABCO is an example of a strong proton accepting functionality often used as catalyst, which has no significant effect on the rate, just like salicylaldehyde. A remarkable trend is the accelerating effect of all the proton-donating groups, which reflects the catalyzing effect of hydroxyl groups in the reaction. As previously acknowledged, the hydroxyl group likely interacts with the epoxide ring, destabilizing it (Figure 8a). More importantly, however, secondary hydroxyalkylamines such as SEC, diethanolamine, and ethylaminoethanol, which contain both an amine and hydroxyl group showed the strongest effect in accelerating the reaction. Such hydroxy-alkylamines contain a proton donor and the amine at close distance from each other, which could kinetically favor a
14
“concerted” ring activation followed by an amine attack (Figure 8b). This type of intermediate would be the bimolecular analogue of the termolecular transition state proposed by Smith.3 Another explanation could be the stabilization of anionic intermediates formed upon reaction of the amine with BO, which would require the presence of two amine moieties (Figure 8c). Comparable intermediates are known to form during the synthesis of carbamates.30
R
R
O
O
H
H O
O
+
a) Destabilization of epoxide ring by binding to an hydroxyl groups proton OH
O
O
H
N
NH2
2
+
O
NH2 R
R
H 2C
+
H N
R
R
OH
O NH3
+
R
H N
R
b) Stabilizing further the
R
NH3 +
H N
R
c) Ionic complex
hydrogen transfer Figure 8: Possible explanation for enhanced auto-catalytic character of SEC.
In a comparable study focusing on the propoxylation of carbon dioxide catalyzed by alkanolamines, it was seen that di-alkanolamines do not lead to a significant rate increase compared to the amines with only one hydroxyl group31.This finding agrees with our study, where diethanolamine (2 hydroxyl groups) has the same effect as ethylaminoethanol and SEC (1 hydroxyl groups). According to this, TERT would be expected to have the same effect as SEC. However, a much weaker catalytic effect was observed for TERT, which suggests a considerable influence of the steric hindrance on its catalytic activity. 15
The hypothesis of Enikolopiyan et al., that the reaction can proceed in the absence of water,22 was tested by determining the reaction rate after drying the reactants. Water is known as a catalyst32, yet it also decreases the selectivity towards the desired products. BO was dried from a water content of 640 ppm to 200 ppm water, while DDA had a constant water content of 60 ppm. Only a minor effect of +0.2 in Figure 7 was observed for the dried BO, which means that water has, at this low concentration, no significant influence, in line with the experimental results from Enikolopiyan et al.22.
Modelling The experimental data strongly suggests an auto-catalytic reaction where both intermediate- and end product catalyze the reactions. The long reaction time to form the intermediate SEC is in agreement with a secondary, co-dependent reaction order for both reactants, in agreement with the literature33. Using the information gained about the kinetics and catalytic species a full kinetic model was first constructed.
𝐷𝐷𝐴 + 𝐵𝑂 𝐷𝐷𝐴 + 𝐵𝑂 𝑆𝐸𝐶 + 𝐵𝑂
(4.1)
𝑆𝐸𝐶 𝑆𝐸𝐶, 𝑇𝐸𝑅𝑇
𝑆𝐸𝐶, 𝑇𝐸𝑅𝑇
𝑆𝐸𝐶
(4.2a, 4.2b)
𝑇𝐸𝑅𝑇
(4.3a, 4.3b)
With the corresponding rate equations: 𝑟𝑖𝑛𝑖𝑡
= 𝑘𝑖𝑛𝑖𝑡
(4.1)
[𝐷𝐷𝐴] [𝐵𝑂]
𝑟𝑐𝑎𝑡,𝑠𝑒𝑐,1 = 𝑘𝑐𝑎𝑡_𝑠𝑒𝑐,1 [𝐷𝐷𝐴] [𝐵𝑂] [𝑆𝐸𝐶]
(4.2a)
𝑟𝑐𝑎𝑡,𝑠𝑒𝑐,2 = 𝑘𝑐𝑎𝑡_𝑠𝑒𝑐,2 [𝑆𝐸𝐶] [𝐵𝑂] [𝑆𝐸𝐶] `
(4.2b)
𝑟𝑐𝑎𝑡,𝑡𝑒𝑟𝑡,1 = 𝑘𝑐𝑎𝑡_𝑡𝑒𝑟𝑡,1[𝐷𝐷𝐴] [𝐵𝑂] [𝑇𝐸𝑅𝑇]
(4.3a)
𝑟𝑐𝑎𝑡,𝑡𝑒𝑟𝑡,2 = 𝑘𝑐𝑎𝑡_𝑡𝑒𝑟𝑡,2[𝑆𝐸𝐶] [𝐵𝑂] [𝑇𝐸𝑅𝑇]
(4.3b)
16
The non-catalyzed reaction of BO and SEC was excluded from the beginning from this model, as SEC is itself a catalyzing intermediate. For each rate coefficient the pre-exponential factors and activation energies according to Arrhenius law (eq.4.4) were fitted to the experimental data using MATLAB’s nonlinear fitting, using the non-linear least-square fitting procedure with the trust-region-reflective algorithm towards the experimental data. In the fitting routine of Matlab, the mole balance was solved by the ODE15s solver using a relative tolerance of 10-8 (see SI for full mole balance). For each addition the mole balance was updated using new initial concentration values and volumes including the loss in sample volume and addition of BO volume.
𝑘𝑗
𝐸𝐴,𝑗 𝑅𝑇
(― ) =𝐴e 𝑗
𝑗 = 𝑖𝑛𝑖𝑡; 𝑐𝑎𝑡𝑠𝑒𝑐,1,2; 𝑐𝑎𝑡𝑡𝑒𝑟𝑡,1,2
(4.4)
The confidence intervals were calculated via the nlparci function, which returns the 95% confidence intervals for the nonlinear least squares parameter estimates of the non-linear fitting. The coefficient of determination was calculated by equation 4.5:
𝑆𝑆𝑟𝑒𝑠
𝑅2 = 1 ― 𝑆𝑆𝑡𝑜𝑡
(4.5)
Where SSres corresponds to the sum of squares of residuals between fitted and experimental values, and SStot to the sum of total squares, which gives an average of the experimental data variance. In a first step only rate equations (4.1) and (4.2a) where fitted to the first BO addition for each temperature. It was assumed that no TERT was formed in amounts sufficient to catalyze the reaction, as verified by the experiments. A coefficient of 0.975 is reached for this fit. The so defined variables were used in the overall fit with all experimental sets. For the fitting of equations (4.2b), (4.3a) and (4.3b) to all data sets, a determination coefficient of 0.951 was found. The summarized Arrhenius parameters are shown in Table 2. 17
Table 2: pre-exponential factors and activation energy valid from 120ºC to 150ºC for full kinetic model
Pre-exponential factor Ainit
95%-confidence interval
Activation Energy
95%-confidence interval
[10-1 m3 mol-1 s-1]
3.6 ± 0.4
EA,init
[kJ mol-1]
60.0 ± 0.1
Acat_sec,1 [10-5 m6 mol-2 s-1]
1.8 ± 0.1
EA,cat_sec,1 [kJ mol-1]
45.6 ± 0.3
Acat_sec,2 [10-3 m6 mol-2 s-1]
3.5± 0.5
EA,cat_sec,2 [kJ mol-1]
63.7± 0.1
Acat_tert,1 [10-4m6 mol-2 s-1]
4.8± 0.1
EA,cat_tert,1 [kJ mol-1]
52.0± 0.1
Acat_tert,2 [10-2 m6 mol-2 s-1]
1.5± 0.2
EA,cat_tert,2 [kJ mol-1]
64.2± 0.1
An exemplary fitting for the data set at 120 ºC is shown in Figure 9. The experimental fittings of temperature from 130 ºC to 150 ºC are presented in the SI in Figure S4.
Figure 9: Auto-catalytic model fit for experimental set at 120 ºC.
The fitted curves represent reasonably the concentration of the reaction components over the reaction progress. A possible explanation for the residual error between the model and the data is due to the 18
analytic difficulties of measuring the products. SEC and TERT are alcohol amines, which tend to interlink with each other and absorb onto the GC column as mentioned in the experimental section. The kinetic coefficients can be further validated by using the spiking experiments. For the initial run at 140 ºC with additional pure TERT as catalytic agent, a rate constant of 4.67 10 -11m3 mol-1 s-1 for equation 4.3a can be achieved. The full rate constant model gives for equation 4.3a a rate constant of 4.16 10-11m3 mol-1 s-1. Comparing the rate coefficients for the reaction catalyzed by SEC and TERT at 140 ºC, the same ratio of one for each reaction step was found. A ratio of one corresponds to only one hydroxyl group being able to participate as catalyst, independently of the molecule having one (SEC) or two (TERT) hydroxyl groups. This correlation offers the opportunity to simplify the catalyst terms due to their similarity. In addition, a simplified model saves computational time for high intensive simulation for complex reaction systems. Each equation pair 4.2a – 4.3a, and 4.2b – 4.3b, were therefore combined into equation 4.6-4.7 using Kratio = 1. The resulting set is shown below: (4.1)
𝑟𝑖𝑛𝑖𝑡
= 𝑘𝑖𝑛𝑖𝑡
𝑟𝑐𝑎𝑡_1
= 𝑘𝑐𝑎𝑡_1 [𝐷𝐷𝐴] [𝐵𝑂] ([𝑆𝐸𝐶] + 𝐾𝑟𝑎𝑡𝑖𝑜[𝑇𝐸𝑅𝑇])
(4.6)
𝑟𝑐𝑎𝑡_2
= 𝑘𝑐𝑎𝑡_2 [𝑆𝐸𝐶] [𝐵𝑂] ([𝑆𝐸𝐶] + 𝐾𝑟𝑎𝑡𝑖𝑜[𝑇𝐸𝑅𝑇])
(4.7)
[𝐷𝐷𝐴] [𝐵𝑂]
The model presented by Sirovski et al. 13 is similar to the above, but the catalytic activity of TERT is considered double of SEC (KRatio = 2). However, besides being unlikely from a sterical point of view that both hydroxyl groups act at the same time as catalyst, our data further suggest that this is not the case for BO. If it is true for ethylene oxide remains uncertain due to the difference in substitution in our work (BO) compared to Sirvoski’s work with ethylene oxide. Roshan et al. 31 studied the cycloaddition of CO2 to propylene oxide using amines and alcohol amines as catalyst. For their system they also demonstrated that two hydroxyl groups on one molecule of alcohol amine have the same or even lower catalytic activity due to steric hindrance, very similar to our observations.
19
Table 3: pre-exponential factors and activation energy valid from 120 ºC to 150 ºC.
Pre-exponential factor
95%-confidence
Activation Energy
interval
95%-confidence interval
Ainit [10-1 m3 mol-1 s-1]
3.6 ± 0.4
EA,init [kJ mol-1]
60.0 ± 0.1
Acat_1 [10-6 m6 mol-2 s-1]
4.3 ± 0.9
EA,cat_1 [kJ mol-1]
39.3 ± 0.5
Acat_2 [10-4 m6 mol-2 s-1]
1.5 ± 0.5
EA,cat_2 [kJ mol-1]
50.8 ± 1.5
With the coefficients presented in Table 4 for the simplified model an overall determination coefficient of 0.94 can be achieved, only a little lower than for the full kinetic model (value = 0.95 with four parameters more). The fitted curves of the simplified model are given in Figure S5 in the SI. In the spiking experiments with SEC and TERT we found that SEC was twice as active as TERT. Using this in the model as a factor of 0.5 for TERT led to slightly worse results (determination coefficient of 0.93). Possibly, the catalytic activity of TERT is enhanced in presence of SEC.
The parity plot of the simplified model versus the experimental concentrations shown in Figure 10 clearly demonstrate that the model predicts well the concentration profiles of the components.
20
Figure 10: Parity plot of model fit to all experimental data sets at all temperatures
The highest deviations are found for the TERT and SEC concentrations. As discussed before, this is attributed to the limited accuracy of the GC and NMR analysis for estimation of these concentrations.
Figure 11: Auto-catalytic model fit to the experimental butoxylation of OA at 140 ºC.
21
Experiments were also conducted to validate the proposed reaction mechanism for alkyl amines in general. Figure 11 shows the result of experiments with octylamine (OA) instead of DDA, showing that the reaction mechanism, including the auto-catalytic effect and the low initial rate, appears valid for alkyl amines in general. Comparing the two amines, OA was 3% faster but in the order of deviation equal to DDA. The fitted rate constants for this experiment are given in Table 4Table .
Table 4: Kinetic constants given for the reaction of OA with BO at 140 ºC.
Rate constants
95%-confidence interval
kinit [10-9 m3 mol-1 s-1] 1.5 ± 0.3 kcat,1 [10-10m6 mol-2 s-1]
7.0 ± 0.9
kcat,2 [10-10 m6 mol-2 s-1]
4.1 ± 0.7
The accuracy of this fit is lower when looking at the confidence intervals and the coefficient of determination of 0.92 due to the lower amount of experimental points to fit. The model was further validated by comparing the model prediction to an additional spiking run, using a reaction mixture enriched with SEC (2%mol) for the initial step. The result of this can be seen in Figure 12. The model accurately describes the experiment, further confirming the validity of the model.
22
Figure 12: Spiked run modelling for 140 ºC and an addition ratio of 0.5:1 mole of BO: DDA.
5. Conclusions The mechanism and kinetics of the butoxylation of dodecylamine (DDA) have been studied. It was found that a range of proton donating groups such as alcohols can catalyze the reaction. Compounds containing both a hydroxyl and an amine group (alkanolamines) are even better catalysts for this process. Both the intermediate secondary butanolamine (SEC) and the product tertiary di-butanolamine (TERT) therefore have an auto-catalytic effect on the reaction rate. The catalytic effect of TERT is not higher than for SEC, as demonstrated by our spiking experiments and the fitting model. We believe this is due to the increased steric hindrance of TERT compared to SEC. Water could be excluded from the kinetic model because it had no significant catalyzing effect with different ppm level concentrations. In addition, spiking experiments demonstrated that a significant inhibition of the reaction rate occurs upon addition of dodecylnitrile and dodecylamide, which are the two most common impurities in the industrial sources of DDA.
23
Two different rate laws were derived that incorporate the auto-catalytic effect of the proton-donating groups of the intermediate and final products. The first includes separate terms for the effect of SEC and TERT in the reaction, while the second represents a simplified, yet almost as accurate model, where SEC and TERT are lumped together. It was found that this rate law satisfactorily describes the experimental data, assuming an Arrhenius type of dependency of the kinetic constants on temperature.
6. Acknowledgements This research was carried out within the HighSinc program – a joint development between Nouryon Specialty Chemicals and the Department of Chemical Engineering and Chemistry from Eindhoven University of Technology.
Abbreviations
BO, butylene oxide; DDA, dodecylamine; OA, octylamine; SEC, mono-(hydroxybutyl) dodecylamine; TERT, bis-( hydroxybutyl) dodecylamine, GC, gas chromatography; MS, mass spectrometer, FID, flame ionization detector; CSTR, continuous stirred tank reactor
References (1)
Schönfeldt, N. Surface Active Ethylene Oxide Adducts; Elsevier, 1969.
(2)
Fatty Amines Market Analysis 2018-2025 https://www.grandviewresearch.com/industry-analysis/fatty-amines-market (accessed Apr 18, 2018).
(3)
Smith, I. T. The Mechanism of the Crosslinking of Epoxide Resins by Amines. Polymer 1961, 2, 95–108.
(4)
Shechter, L.; Wynstra, J.; Kurkjy, R. P. Glycidyl Ether Reactions with Amines. Ind. Eng. Chem. 1956, 48 (1), 94–97.
24
(5)
Holubka, J. W.; Bach, R. D.; Andres, J. L. Theoretical Study of the Reactions of Ethylene Oxide and Ammonia: A Model Study of the Epoxy Adhesive Curing Mechanism. Macromolecules 1992, 25 (3), 1189–1192. https://doi.org/10.1021/ma00029a028.
(6)
Di Serio, M.; Vairo, G.; Iengo, P.; Felippone, F.; Santacesaria, E. Kinetics of Ethoxylation and Propoxylation of 1- and 2Octanol Catalyzed by KOH. Ind. Eng. Chem. Res. 1996, 35 (11), 3848–3853. https://doi.org/10.1021/ie960200c.
(7)
Chiu, Y. N.; Naser, J.; Easton, A.; Ngian, K. F.; Pratt, K. C. Kinetics of a Catalyzed Semi-Batch Ethoxylation of Nonylphenol. Chem. Eng. Sci. 2010, 65 (3), 1167–1172. https://doi.org/10.1016/j.ces.2009.09.080.
(8)
O’Lenick, A. Group Selectivity of Ethoxylation of Hydroxy Acids. Chemist 2002, 79, 9–16.
(9)
Komori, S.; Karaki, T. The Reaction between Alkylamine and Ethylene Oxide without Any Catalyst. J. Soc. Chem. Ind. Jpn. 1959, 62 (4), 538–542. https://doi.org/10.1246/nikkashi1898.62.4_538.
(10)
Rupp, M.; Ruback, W.; Klemm, E. Alcohol Ethoxylation Kinetics: Proton Transfer Influence on Product Distribution in Microchannels. Chem. Eng. Process. Process Intensif. 2013, 74, 187–192. https://doi.org/10.1016/j.cep.2013.09.006.
(11)
Wu, Y.; Liu, Q.; Su, X.; Mi, Z. Effect of Solvents on Propylene Epoxidation over TS-1 Catalyst. Front. Chem. China 2008, 3 (1), 112–117. https://doi.org/10.1007/s11458-008-0007-2.
(12)
Brown, W.; Foote, C.; Iverson, B.; Anslyn, E. Organic Chemistry; Cengage Learning, 2008.
(13)
Sirovski, F.; Mulyashov, S.; Shvets, V. Large-Scale Fatty Amine Ethoxylation Reactor: A Dynamic Model. Chem. Eng. J. 2006, 117 (3), 197–203. https://doi.org/10.1016/j.cej.2005.12.005.
(14)
Sanders, H. L.; Braunwarth, J. B.; McConnell, R. B.; Swenson, R. A. Ethoxylation of Fatty Amines. J. Am. Oil Chem. Soc. 1969, 46 (3), 167–170.
(15)
Sundaram, P. K.; Sharma, M. M. Kinetics of Reactions of Amines with Alkene Oxides. Bull. Chem. Soc. Jpn. 1969, 42 (11), 3141–3147.
(16)
Kothari, P. J.; Sharma, M. M. Kinetics of Reaction between CS2 and Amines. Chem. Eng. Sci. 1966, 21 (5), 391–396. https://doi.org/10.1016/0009-2509(66)85049-2.
(17)
Gaertner, V. R. Alkyl-2, 3-Epoxypropylamines: Cyclodimerization and Related Eight-Membered Ring Closures. Tetrahedron 1967, 23 (5), 2123–2136.
(18)
Lang, R. F.; Parra-Diaz, D.; Jacobs, D. Analysis of Ethoxylated Fatty Amines. Comparison of Methods for the Determination of Molecular Weight. J. Surfactants Deterg. 1999, 2 (4), 503–513.
(19)
Laird, R. M.; Parker, R. E. The Mechanism of Epoxide Reactions. Part XII. Reactions of Ethylene Oxide with Alcohols in the Presence of Sodium Alkoxides and of Tertiary Amines. J. Chem. Soc. B Phys. Org. 1969, 1062–1068.
(20)
Lovett, E. G. Zwitterionic Quaternary Ammonium Alkoxides: Organic Strong Bases. J. Org. Chem. 1991, 56 (8), 2755– 2758. https://doi.org/10.1021/jo00008a034.
25
(21)
Trejbal, J.; Petrisko, M.; Pašek, J. Reactions of Ethoxylates of Ethylenediamine and Piperazin over ZSM-5 Zeolithe. Pet. Coal 2012, 54 (4), 340–347.
(22)
Enikolopiyan, N. S. New Aspects of the Nucleophilic Opening of Epoxide Rings. Pure Appl. Chem. 1976, 48 (3), 317– 328. https://doi.org/10.1351/pac197648030317.
(23)
Tiltscher, V. H. Kinetik Und Mechanismus Der Anlagerung von Propylenoxid an Diäthylamin Unter Ausschluß Protonenaktiver Verbindungen. Angew. Makromol. Chem. 1972, 25 (1), 1–14. https://doi.org/10.1002/apmc.1972.050250101.
(24)
Aresta, M.; Quaranta, E. Mechanistic Studies on the Role of Carbon Dioxide in the Synthesis of Methylcarbamates from Amines and Dimethylcarbonate in the Presence of CO2. Tetrahedron 1991, 47 (45), 9489–9502. https://doi.org/10.1016/S0040-4020(01)80894-2.
(25)
Cai, X.; Zheng, J. L.; Wärnå, J.; Salmi, T.; Taouk, B.; Leveneur, S. Influence of Gas-Liquid Mass Transfer on Kinetic Modeling: Carbonation of Epoxidized Vegetable Oils. Chem. Eng. J. 2017, 313, 1168–1183. https://doi.org/10.1016/j.cej.2016.11.012.
(26)
Rupp, M.; Ruback, W.; Klemm, E. Octanol Ethoxylation in Microchannels. Chem. Eng. Process. Process Intensif. 2013, 74, 19–26. https://doi.org/10.1016/j.cep.2013.09.012.
(27)
Pinchas, S.; Ben-Ishai, D. The Carbonyl Absorption of Carbamates and 2-Oxazolidones in the Infrared Region. J. Am. Chem. Soc. 1957, 79 (15), 4099–4104. https://doi.org/10.1021/ja01572a035.
(28)
Scholz, E. Karl Fischer Titration: Determination of Water; Chemical Laboratory Practice; Springer-Verlag: Berlin Heidelberg, 1984.
(29)
Ingberman; Walton. Low Toxicity Aliphatic Amines as Crosslinking Agents for Polyepoxy Resins. J. Polym. Sci. 1957, 28 (117).
(30)
McCann, N.; Phan, D.; Wang, X.; Conway, W.; Burns, R.; Attalla, M.; Puxty, G.; Maeder, M. Kinetics and Mechanism of Carbamate Formation from CO2(Aq), Carbonate Species, and Monoethanolamine in Aqueous Solution. J. Phys. Chem. A 2009, 113 (17), 5022–5029. https://doi.org/10.1021/jp810564z.
(31)
Roshan, K. R.; Kim, B. M.; Kathalikkattil, A. C.; Tharun, J.; Won, Y. S.; Park, D. W. The Unprecedented Catalytic Activity of Alkanolamine CO2 Scrubbers in the Cycloaddition of CO2 and Oxiranes: A DFT Endorsed Study. Chem Commun 2014, 50 (89), 13664–13667. https://doi.org/10.1039/C4CC04195J.
(32)
Azizi, N.; Saidi, M. R. Highly Chemoselective Addition of Amines to Epoxides in Water. Org. Lett. 2005, 7 (17), 3649– 3651. https://doi.org/10.1021/ol051220q.
(33)
Os, N. M. van. Nonionic Surfactants: Organic Chemistry; CRC Press, 1997.
26
Highlights
Experimental study on butoxylation of fatty primary amines Reaction mechanism elucidated leading to an accurate kinetic model Auto-catalytic reaction is the driving force for high conversions Reaction rate strongly enhanced by bidentate action of the alcohol and amine group
27