Fatty nitriles have lately become of interest in the framework of biofuels and for the valorization of the oil part of biomass to form fine chemicals or polymers. The production of long-chain fatty nitriles by the direct reaction of acids with NH3 has not been extensively studied, although several catalysts have been developed and published as patents. The characterization of this reaction with and without catalyst is, to the best of our knowledge, performed for the first time in this study. Several catalysts with various acid–base features were tested, and the best catalysts at 250 °C (Zn- and In-based catalysts) were further studied. Catalytically active forms and models are proposed for the Zn- and In-based catalysts, and the kinetic parameters for the amide to nitrile reaction are evaluated.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Nitriles are platform molecules that are useful in medicine and polymer chemistry. In the framework of renewable energy resources, the valorization of nonedible biomass into biofuels and fine chemicals has become a major field of research, and the conversion of the oil part of biomass, that is, triglycerides, into fatty esters and nitriles has been envisaged for biofuel production,1–3 in which case the alteration of the chain during the process is not a problem. The valorization of nitriles as fine chemicals, however, raises the necessity of controlling the nature of the chain, especially regarding unsaturation. Indeed, the nature of commonly used catalysts and the high operating temperature of nitrilation processes are the source of isomerization and several side-reactions, such as the Piria or Diels–Alder reactions or peroxidation in the α position of the unsaturation (Figure 1). Thus, to produce high-added-value nitriles with control over the fatty chain, it is necessary to reduce the operating temperature. In addition, this will allow the processing of shorter-chain material and reduce the energy cost of such a process, which is of environmental and economic interest.
The nitrilation of fatty acids by direct reaction with NH3 is of interest to industrial chemists, and several patents have been registered since the middle of the 20th century; however, the number of patents increased significantly in the last decade because of the increasing interest in biofuels. Among the patent literature, three processes can be distinguished: (1) the batch process, in which the acid reactant is in the liquid phase and the nitrile remains in the reactor, (2) the gas-phase continuous process, in which the acid is vaporized prior to introduction into a catalytic bed, through which it passes together with NH3, and (3) the liquid-phase continuous process, which requires very specific thermodynamic conditions (the operating temperature must be lower than the nitrile boiling point). Whereas the gas-phase process consumes a considerable amount of energy for the vaporization of the acid, which can lead to the alteration of the carbon chain in the evaporation chamber before it reaches the catalytic bed, the batch liquid-phase process needs a few hours of reaction at high temperature, during which side-reactions are likely to occur, especially with unsaturated chains. Thus the gas-phase process is usually more adapted to short carbon chains (C≤12) and unsaturated carbon chains as the contact time with the catalyst at high temperature is reduced, whereas the liquid-phase process is more adapted to long carbon chains (C≥12) and especially saturated chains for thermodynamic reasons. The scope of this study is the transformation of long-chain fatty acids (C≥14 and the main fraction of C18:1), and the liquid-phase batch process was chosen in an attempt to decrease the operating temperature of the process relative to that of the state-of-the-art.
The first know-how concerning the liquid-phase conversion of long-chain fatty acids is in a patent published in 1941,4 in which bauxite, titanium dioxide, or thorium oxide are presented as good catalysts for nitrilation at 300–400 °C under a high pressure of NH3 for several long-chain and unsaturated fatty acids. A patent from 1950 proposed Co compounds as catalysts and gave an example of the nitrilation of oleic acid at 270 °C.5 In 1961, another patent reported the development of titanium alcoholate catalysts for the conversion of unsaturated acids at 288 °C.6 In 1983–84, Fe-based catalysts were reported, which converted C12–C18 acids at 250 °C over 12–15 h with 0.1 stoichiometric equivalents of NH3 per hour with the observation of a decrease of the kinetics with temperature in the case of lauric acid conversion.7 A series of patents was then published on improvements to this process;8 as the state-of-the-art catalysts such as zinc oxide, Co salts, Fe compounds, and kaolin are soluble in the reaction medium and as their separation from the medium decreases the effective yield or increases the waste, insoluble catalysts based on TiO2 and Nb2O5 have been developed, and results with stearic acid and methyl ester at 260 °C have been published. In 2007–2009, patents were published on the conversion of technical mixtures of unsaturated chains (hardened or distilled tallow fatty acids) at 300 °C using p-toluenesulfonic acid with salts of Zn, Al, Co, or Ti. In summary, the first efficient acid–base catalysts developed industrially were soluble in the reaction medium and innovation in the last decade was to develop insoluble catalysts that were improved with surface-acid treatment; however, the most recent patents lead us to observe that, to convert unsaturated and long-chain acids into nitriles at a reasonable temperature, soluble catalysts still dominate.
The conversion of fatty acids or amides into nitriles has also been reported in the academic literature, essentially under nonindustrial conditions, by using low concentrations of the reactants in solvents such as toluene, tetrahydrofuran, or mesitylene, and using dehydrating agents in stoichiometric amounts, such as the recent studies by Enthaler et al. that used Zn, Fe, or Cu salts as precursors for catalysts that converted aromatic substituted primary amides into nitriles at 70–100 °C in 24 h.9 Only a few studies have investigated nitrilation by direct reaction with NH3 under industrial conditions and, to the best of our knowledge, only in the gas phase. Such is the case of Mitchell and Reid,10 who published a study on the nitrilation of C2–C7 acids at 500 °C on alumina, which resulted in good yields. In addition, a test on palmitic acid (C16) did not result in any nitrile. Bizhanov et al.11 reported a kinetic study on the nitrilation of mixtures of C10–C22 aliphatic acids by direct reaction with NH3 in the gas phase at 300 °C. By using an alumina catalyst, they obtained almost full conversion in 40 min contact time and observed kinetic zero laws for the formation of amide and nitrile of 123.3 and 97.7 mmol s−1, respectively.
In this study, our aims were to develop amphoteric oxide catalysts to improve the catalytic nitrilation of technical oleic acid at 250 °C, to evaluate the influence of the system parameters on efficiency, and to evaluate the interaction of working catalysts with the reactants.
The production of nitrile without catalyst will be presented first, then the impact and the nature of catalysts will be evaluated; finally, two efficient catalysts (based on Zn and In) will be more specifically discussed, and their reciprocal synergy effect will be examined.
Results and Discussion
Several catalysts have been developed for this reaction; however, the patent literature only discloses final yields and tests are not standardized. Discussion on the analytical means deployed here and on the whereabouts of isomerization and its evaluation in this reaction can be found in the Experimental Section. Nitrilation without catalyst is described first, then the effect of catalysts of different natures will be presented, and finally Zn- and In-based catalysis will be studied more deeply.
Nitrilation without catalyst
Blank experiments performed under standard conditions with variation of the operating temperature provide evidence on the actual conversion of acid and NH3 into nitrile without a catalyst but with considerably slower kinetics of amide conversion. As can be seen from Figure 2, the amount of amide almost stops decreasing after a certain time and stays at 14–16 mol % at 300 °C, starting from an acid or amide medium, respectively, which shows a form of equilibrium between the nitrile and amide. In the late stages of the reaction, the medium is mainly composed of nitrile and every acid or amide molecule that becomes dehydrated produces a water molecule that then travels towards the surface and encounters several nitrile molecules with a nonzero probability of reacting with it to produce amide. At this stage, it appears as if equilibrium is found in this medium at 300 °C stirred at 150 rpm with a 5:15:80 repartition of ammonium carboxylate, amide, and nitrile, respectively. This hypothesis can lead to an estimation of the ratio of both of the thermodynamic constants of the dehydration reactions of 1.78, considering that the water content is somehow stable. Consequently, at 300 °C operating temperature, one can estimate the difference of the free enthalpies of reaction between both stages of conversion (salt to amide and amide to nitrile) to be approximately 2.7 kJ mol−1. The amounts of fatty compounds if starting with a batch of amide at 300 °C without catalyst or NH3 flow are shown in Figure 3. If we consider the situation in the absence of NH3 bubbling, the ammonium salt is probably converted into the acid, which prevents any further reaction of acid into amide and concentrations appear to reach a stable regime. If we consider the simple model shown in Figure 4 for this kinetically determined system, these products provide an evaluation of the ratio of kinetic constants k2/k′−1 of approximately 1.7. We compare the kinetics of dehydration versus hydration of amide with the hypotheses that the carboxylate salt is unstable under these conditions and that no diffusion limitation of water or NH3 takes place. The same experiment at an operating temperature of 250 °C reveals significantly lower kinetics, but the eventual thresholds will not be reached in the timescale of our experiments.
However, experiments that start with a nitrile and water feed performed with and without catalyst at 250 °C display no amide production, which gives low credit to this equilibrium scenario unless the active catalyst cannot be formed under these conditions.
In the case of amide batch experiments without catalyst, if we consider that the acid content is kept low enough by the continuous NH3 flow and that the nitrile production is almost unaffected by this back reaction, the activation energy of the amide-to-nitrile reaction can be evaluated to be in the order of 109 kJ mol−1 by an Arrhenius plot. The experiments that start with amide and 0.23 mol % zinc oxide at 300, 270, and 250 °C are displayed in Figure 5, and, with the same assumptions, the activation energy can be evaluated to be in the order of 85 kJ mol−1, which features a lower activation energy because of the catalytic pathway.
The patent literature provides several examples of efficient catalysts (e.g., ZnO, Fe compounds, kaolinite, Nb2O5). Some of these were tested on the lab-scale pilot system, among which some were observed to dissolve in the medium with the background idea that the dehydration reactions must be catalyzed by acid–base features. A first series of catalysts was tested with 0.065 wt % of introduced material: magnesium, zinc, aluminum, gallium, indium, tin, and niobium oxides, tungstated zirconia, and ceria-supported vanadia.
As can be observed from Figure 6, the consumption of acid is not significantly different throughout the tested catalysts and is similar with and without catalyst at 300 and 250 °C. The acid consumption is limited by the NH3 feed rate (dotted line), which is imposed by the system and constant throughout the tested catalysts; additional comments that concern NH3 are given in the Experimental Section. The conversion of acid into amide appears to be not or only slightly affected by the presence of catalyst, and the slight disparity can either be because of a slight variation of the NH3 feed or to a different amount of amide accumulation (backward reaction). Table 1 gathers values of the amide content after 8 h reaction of the acid batch at 300 and 250 °C. The amide accumulation differs importantly between the catalysts. At 300 °C, tin dioxide and the very acidic zirconia-supported tungstate display as much amide accumulation as the reaction without catalyst, whereas basic magnesium oxide and the amphoteric oxides (Zn, Ga, In, and Al) display a similarly low amide content after 8 h. At 250 °C, every acidic (niobia), basic (magnesia), and amphoteric (e.g., Ga2O3, Al2O3) catalyst displays a high amide accumulation, similar to the reaction without catalyst. The kinetics of amide conversion into nitrile is influenced by the presence of a catalyst; however, the acid–base surface properties do not correlate with this. Notably, the catalysts do not behave similarly under these harsh conditions: some dissolve and others stay solid.
In this test, an insoluble catalyst such as niobium oxide8 displays less efficient amide conversion than zinc oxide, which was observed to dissolve in such a medium.8a Indium oxide also displays efficient amide conversion at 250 °C, but oxides of the same column (IIIa) with similar amphoteric surface features do not. Gallium and aluminum nitrates were tested with the same molar amount of metal as in the tests with oxides. Their kinetics are similar to each other and to the corresponding oxides at 300 or 250 °C starting from acid or amide, and the amide kinetics are almost the same as those without catalyst.
Having focused on the two catalysts that perform best at 250 °C (zinc and indium oxides), tests were performed that started from acid or amide, and the results were compared for the oxide or nitrate salt catalyst precursors.
First, zinc oxide and nitrate led to a significant enhancement of the kinetics of the conversion of amide into nitrile, and for similar molar amounts of Zn, both precursors led to similar kinetics. It was also observed that soon after the introduction of the solid precursors, no particle suspension could be seen. Second, indium oxide and nitrate led to a variable enhancement of the kinetics. For similar molar amounts of introduced In species, nitrate salts perform better than oxides. Inductively coupled plasma optical emission spectroscopy (ICP-OES) performed on the decanted kinetic samples provide information on the phenomena that occur: depending on the starting reactant (acid or amide), dissolution or leaching happens differently. Indium nitrate is already completely dissolved in the fatty medium after 1 h and remains at a stable concentration, whereas the oxide is leached slowly and considerably more in the acid than in the amide medium (stable over 8 h at 33 % relative dissolution in the acid, slowly increasing by 0.4 % h−1, and finishes at approximately 4 % relative dissolution in the amide; Figure 7, top). The kinetics of nitrile production appears to be correlated to the dissolved amount of In cations in the medium (Figure 7, bottom.)
Although zinc oxide, kaolinite, and Fe compounds are soluble in the studied medium, without precision of the temperature at which this observation was made,8a no mechanism or homogeneous catalytically active form was proposed. The dissolution of Zn has also been reported by Zhao et al. in a similar medium that consisted of urea and alcohol at 180 °C in an autoclave; a hypothetical mechanism is provided with a catalytically active form that consists of a Zn2+ cation coordinated by two NH3 and two NCO ligands.13 In the nitrilation batch set-up, the species present are NH3, ammonium carboxylate, water, hydroxyl, amide, and nitrile, the temperature is significantly higher (250–300 °C) and no solvent is used. Then, if a homogeneous catalytically active form is to be created, it may be coordinated with NH3 and the most nucleophilic molecules (carboxylate or hydroxyl groups). As the medium was at high temperature and opaque, no technique was found to directly observe the homogeneous catalytically active form in situ.
Methods exist to determine if the catalytically active form is heterogeneous or leached and then homogeneous—the most accepted involves filtration of the medium at the operating temperature14—however, all of these methods were impossible to perform in the special case that we have studied because of the harsh conditions of causticity, the opacity of the medium, the high temperature, and the viscosity. Thus to give more credit to homogeneous catalysis of Zn, a series of experiments was performed at 250 °C with several Zn species: commercial zinc oxide, nanoscale zinc oxide, synthesized zinc oxide,15, 16 zinc hydroxide, zinc nitrate, zinc stearate, and zinc methane sulfonate, with equal molar amounts of Zn, and the observed kinetics are almost identical. From this we can estimate that the catalytically active form is created almost as easily and in similar amounts for each of the tested Zn species.
A more complete study is discussed in the next section on both Zn- and In-based catalysts, which give the best results with a 0.065 wt % catalyst loading. Amide is chosen as a starting reactant to get rid of the NH3-feed limitation and to let the actual kinetics be observed. The acid will mainly be present in its ammonium carboxylate form.
Homogeneous Zn catalysis
A series of experiments that used Zn concentrations of 0.023–8.0 mol % was performed at 250 °C for 8 h, starting from a batch of amide and providing a feed of the oxide precursor, the dissolution of which was monitored by decantation and ICP-OES analysis. It was observed that even at high Zn loading, the totality of the precursor is dissolved in the starting hours. However, for a high loading and only after a few hours, we can observe a continuous decrease of the Zn concentration that correlates with the evolution of the medium: if the concentrations of the acid and amide are too low, the Zn appears to precipitate (with the appearance of a white suspended powder), which probably leads to a decrease of the catalytic efficiency. The evolution of Zn solubility with acid and amide content can be observed in Figures 8 and 9, in which the results of experiments performed with a high Zn loading (1.6 and 8.0 mol %) are displayed, and the acid and amide content are compared to the amount of introduced and dissolved Zn. As long as the acid content is twice as large as the Zn content, full dissolution appears to be obtained, but for lower acid contents Zn starts to precipitate and is squashed by the acid-consumption curve (this may also explain why no reaction occurs if nitrile is mixed with water as the active catalyst is not generated). The analyses that yield the acid and amide contents take into account the free molecules as well as any coordinated ones. Then, as the contents of acid and Zn have a ratio of two, we can assume that two carboxylate molecules coordinate a Zn2+ cation. This complex also appears quite unstable as it lets the reaction consume its ligands rather than keeping its structure. The experiment with 8 mol % Zn was discarded from the database for kinetic reasons as its Zn content varies too much too early (≈2 h) in the experiment.
Figure 10 presents the nitrile concentration of the medium for this series of experiments on Zn, in which the kinetics appears to increase with the amount of dissolved Zn. Figure 11 presents the carboxylate salt concentration for this same series, and the significant evolution of the time constants of production and consumption can be appreciated.
Next, experiments were performed to understand the mechanism and the catalytically active form. Experiments on the hydration of nitrile with and without Zn catalyst at 250 °C were performed. No formation of amide was observed, and the reaction was displaced by the instability of water inside the medium at such a temperature. However, ICP-OES was performed, and the amount of dissolved Zn did not reach more than 0.026 mol %; therefore, the inefficiency of the backward reaction could be because of the low amount of catalytically active form, which needs more carboxylate molecules. Salting-out experiments (with the addition of sodium chloride up to 15 mol %) were also performed, with which we intended to influence the accessible water content of the medium. No poisoning of the Zn-based catalytically active form happened, but also no significant change in the two hydration reactions was observed. Furthermore, the nitrile hydration reaction kinetic constant is considered too small to figure in the equation system.
In the series of experiments that start from amide without NH3, the amounts of nitrile and acid or ammonium carboxylate are stable and significant; thus it can be assumed that the reaction of nitrile with acid towards N,N-diamide can be neglected.17
If we consider the influence of NH3, it is necessary to provide the system with enough NH3 because the backward reaction of amide to acid is a source of NH3 loss. In addition, if we start from the acid, at least a stoichiometric amount of NH3 has to be fed into the system, and the kinetics is strongly influenced by the NH3 flow-rate as can be seen in Figure 12 (top). A series of experiments that used NH3 flow rates of 37.5, 56.25, 75, or 425 mL min−1 and a Zn concentration of 0.23 mol % was performed at 250 °C for 8 h starting from 85 g of amide with a feed of the oxide precursor. As the evolution of the acid content is similar for each experiment, the characteristic time of amide conversion can be observed in Figure 12, bottom, as a function of the NH3 flow rate. A threshold appears above 75 mL min−1, 0.6 molar equivalents per hour, which seems to characterize the NH3 diffusion limitation domain.
Thus to prevent the rate-determining step from being NH3 diffusion, the flow rate cannot exceed this value; however, it is also interesting to keep the flow rate high enough to prevent the observation of too much back reaction towards the acid. The kinetics observed at 250 °C on batch experiments of 85 and 170 g of amide with the same agitation and the same relative NH3 flow rate at 0.3 molar equivalents per hour are very similar. As seen earlier, this domain of NH3 flow rate appears to suffer no diffusion limitation. For the experiments performed on a batch of amide, a flow rate of 0.6 molar equivalents per hour, that is, 75 mL min−1 for 85 g of reactant, was chosen, which ensures no NH3 diffusion limitation while keeping the acid accumulation low.
For the series of experiments with a variable Zn content in the medium, in the timeframe of acid content stability, the global kinetic order for the amide content towards the nitrile production rate can be evaluated. As can be observed from Figure 13, the apparent reaction order towards the amide content appears to be close to 1, which would correspond to a rate-determining step that includes the direct reaction with one molecule of amide, and the corresponding constant k2 depends on the catalyst content. If we consider the acid and amide kinetics, although differences appear with variation of the catalyst content, especially around the time of maximum acid content, which is delayed sooner with an increasing amount of catalyst, the mechanism needs to be modeled.
Owing to the lack of knowledge on the exact composition of the catalytically active form and on the reactions on which its influence is significant, the role of the catalyst was first considered implicitly in the kinetic constants of the following mechanism [Eqs. (1)–(3)]:(2), (3)
It is assumed that the dissolution is quick enough and does not limit the overall kinetics, which is supported by experiments on several salt and oxide precursors as well as by the chemical analyses discussed above, and that with an NH3 feed the acid only exists as an ammonium salt, which means that the acid–base reaction is fast under the conditions of NH3 bubbling. If we consider experiments with amide as the starting reactant and a continuous NH3 feed, we make the hypothesis that the reactions are elementary. This system can then be described with a simple linear differential system for the time evolutions of the salt, amide, and nitrile concentrations [Eqs. (4)–(6)]:(5), (6)
Fitting this model on the series of experiments with variable amounts of dissolved Zn yields values of kinetic constants that can be observed in Figure 14. Given the impossibility to observe the catalytically active form in situ and the complexity of the studied process, assumptions were made that the following statements depend on. As the analyses were performed ex situ and needed significant amounts of product (1–2 mL), sampling was performed hourly. Thus for each experiment, the database to be fitted is not extensive, but the parameters can still be fitted accurately, provided that the shapes of the curves are properly considered. Actually, by evaluating the time location of the maximum accumulation of ammonium carboxylate, β can then be estimated. This time (tm) is identified for [Zn]<0.46 mol %, between 0.7 and 4 h, and observed approximately afterwards.
As α has a very stable value close to −k2, solutions can be reduced to those that respect the estimation of β by Equation (14):
Although we stated that the catalyst has no significant impact on the acid conversion in experiments that started with the acid, the model leads to an evolution of k1 (ammonium carboxylate to amide) that features a linear trend until 0.4 mol % Zn followed by a plateau (1.2 h−1).
In the experiments that started from a batch of acid, it was observed that the kinetics was controlled by a zero law of NH3 supply, whereas starting from a batch of amide, this limiting step does not occur in the early stages of the reaction.
The evolution of k2 (amide to nitrile) can be assumed to correspond to the coexistence of mechanisms with and without catalyst. Then k2 appears as the sum of a constant term, determined by the blank experiment (0.075 h−1) and a variable term that follows a law of half of the Zn content [mol %] (0.28×[Zndissolved]0.5 h−1). The maximum k2 (0.40 h−1) obtained at 1.6 mol % Zn corresponds to 32 μmolnitrile s−1; however, the maximum value of the efficiency per mol of Zn is above 0.13 molnitrile molZn−1 s−1 for a low Zn content. No simple system corresponds to a kinetic law of order a half, relative to the amount of catalyst. However, one possibility is the existence of a dinuclear Zn complex in equilibrium with two identical and less stable mononuclear complexes of Zn that would react with amide to produce nitrile, and this latter step is the rate-determining step, which corresponds to k2 [Eqs. (15)–(17)]:(16), (17)
The dinuclear species needs to correspond to the main representative form of dissolved Zn and then the order of a half stems from the equilibrium constant with the mononuclear species. Dissolution experiments performed in this study suggest the coordination of two carboxylate groups per Zn atom and academic literature on dinuclear Zn species provides examples of 18-electron complexes, in which carboxylates appearing as LX-type bridging ligands.18, 19 However, amide does not help with the formation of the catalytically active form and modification of the NH3 flow rate does not change the kinetics. Thus we proposed a dinuclear species Zn2(RCO2)4, which has two bridging carboxylates, and a mononuclear species Zn(RCO2)2 (Figure 15). A catalytic mechanistic cycle is also proposed in Figure 15 with these hypotheses. Nothing in the experimental work suggests the absolute necessity of NH3 as a ligand; however, there probably exists some equilibrium between the bridged LX carboxylate and the X-coordinated carboxylate plus an L-coordinated NH3 complexes, and the simplest hypothesis was chosen.
The values of k2 obtained by linear fitting the logarithm of nitrile production as a function of the logarithm of the amide content are very similar to results of this model and follow the same trend.
The evolution of k′−1 (amide to ammonium carboxylate) could also be the sum of a constant term (without catalyst: 0.08 h−1) and a variable term (≈0.07 × [Zndissolved(mol %)]0.5 h−1), which could be explained by the presence at close range of amide and water in the coordination or solvation area of the catalytically active form or even by the possible transfer of a water molecule from a coordinated amide that is dehydrated towards the other coordinated amide to produce an acid and a nitrile from two amide molecules.
The experiments that start with acid can also be modeled, provided we make more hypotheses. First, let us assume that at 250 °C with the continuous bubbling of NH3 the acid completely reacts with NH3 to produce the carboxylate salt. The differential system is modified by an external parameter, which only appears in the expression of the ammonium salt content as a function of time, NH3 flow rate (F), NH3 content in the medium (n) and the kinetic constant of the proton exchange (k0). After one equivalent of NH3 has been provided, the acid content is given as zero, the usual differential system is valid, and continuity between both systems is ensured [Eq.(18)].(18)
If we consider the amide model, a set of parameters (k1=0.725 h−1, k′−1=0.090 h−1, k2=0.150 h−1) was obtained and reinserted in the acid model with success, which provided new fitting parameters (k0 n=0.072 h−1, F=86.7 mmol h−1).
Homogeneous In catalysis
A series of experiments that used In concentrations of 0.023–1.6 mol % was performed at 250 °C for 8 h that started from a batch of amide and a feed of the nitrate salt precursor, the dissolution of which was monitored by decantation and ICP-OES. Contrary to zinc oxide, indium oxide is poorly dissolved in the amide medium (≈3 mol %); therefore, In was introduced as a nitrate salt. However, for high loading, the nitrate is responsible for the appearance of a brown color, probably related to a side reaction. Although it was not observed, as for Zn, there is a maximum solubility for In, which should vary with the acid/amide/nitrile composition. With the same hypotheses as for Zn, a simple model was applied to this series of experiments.
As can be seen from Figure 13, similarly to the Zn model, k1 (ammonium carboxylate to amide) was found to evolve, which featured a linear trend until 0.4 mol % [In] followed by a plateau (at 1.5 h−1).
The evolution of k2 (amide to nitrile) also appears as the superposition of terms with and without catalyst. The constant part of k2 was determined by a blank experiment (0.075 h−1), and the variable term also follows a law of half of the In content [mol %] (0.35×[Indissolved]0.5 h−1). The maximum of k2 (0.50 h−1) obtained at 1.6 mol % In corresponds to 40 μmolnitrile s−1; however, the maximum value of the efficiency per mol of In is above 0.13 molnitrile molIn−1 s−1 for a low In content.
The evolution of k−1′ (amide to ammonium carboxylate) can also be considered as the sum of a constant term (without catalyst: 0.08 h−1) and a variable term (≈0.14×[Indissolved]0.5 h−1).
The fit of the kinetic parameters proposes higher values for In than Zn for both the forward (amide to nitrile 25 % faster, salt to amide 25 % faster) and backward reactions (amide to ammonium carboxylate 115 % faster). Notably, In is introduced as a trivalent cation and is most likely to stay in this oxidation state; therefore, it is expected that three carboxylates coordinate it, which form a species significantly different to that hypothesized for Zn.
Synergy between In and Zn
Zn and In were found to dissolve in the studied medium to probably form coordination complexes. Both metal complexes are thought to lead to catalytically active forms that enhance the nitrilation kinetics, and the ligands most probably still react in the late stages; thus it is expected that their coordination is reversible. There might even be exchange between the catalytic forms, thus synergetic effects were sought between these metals by providing both precursors in variable relative amounts to keep the total amount of dissolved metal at 0.23±0.01 mol %. Figure 16 displays the obtained nitrile yields as a function of time and In ratio. A small addition of In was found not to improve the Zn catalytic efficiency; however, the efficiency of this mixture outperforms the arithmetic addition of both separately, especially in the equimolar case, which is as efficient as the system with only In. Both catalytic forms are thought to undergo through similar mechanisms, which imply kinetic laws of order of a half; however, the experiments show that the marginal kinetics for low concentrations, which are more efficient, cannot be added, as if the catalysts do not interact.
Dehydration catalysts were tested for the batch nitrilation of unsaturated fatty acids by direct reaction with NH3 at 250 °C, which is lower than the state-of-the-art operating temperature. The use of lower operating temperatures is probably the most efficient way to decrease side reactions such as polymerization and isomerization, especially concerning unsaturated fatty compounds. The reaction without a catalyst was described and characterized for the first time; then catalysts with different acid–base surface features were screened, which included efficient catalysts reported in the patent literature. Zn and In species were found to display the best results at 250 °C, which correlated with the amount dissolved in the medium for both metals. The catalytically active forms for both metals could not be observed by using the usual techniques; however, several experiments suggested the existence of a coordination complex formed of carboxylate molecules and possibly NH3 around the cation. The nitrile kinetics was studied and a model was proposed, which displayed a half-order dependency on the catalyst content that could originate from an equilibrium between mono- and bimetallic coordination complexes. The kinetics of nitrile production for In was 25 % higher than that for Zn and reached 40 μmol s−1 for 1.6 mol % In, and half of the In can be replaced by Zn before the efficiency is reduced. However, both of the Zn and In catalysts did not reduce the activation energy enough to be good catalysts at temperatures lower than 250 °C. A rough estimation of nitrile production at 200 °C from the Arrhenius plot (250–300 °C range) yielded kinetics approximately eight times slower than that at 250 °C. Moreover, the amount of catalytically active form was limited by half of the acid content, above which the metals started to precipitate. A solution to overcome this problem would be to set up a continuous liquid-phase system with limitations of temperature and the nature and length of the fatty chains.
Commercial zinc oxide (Sigma–Aldrich), zinc oxide nanopowder (Nanostructured & Amorphous Materials Inc.), zinc nitrate (Strem), zinc methanesulfonate (Tib Chemicals AG), zinc stearate (Sigma– Aldrich), γ-alumina (Degussa), aluminum nitrate (Sigma–Aldrich), gallium oxide, and nitrate (Strem), indium oxide and nitrate (Strem), niobium oxide (Starck), tin dioxide (Strem), and magnesium oxide (Merck) were used. Tungstated zirconia was prepared by the calcination of commercial tungstated zirconium hydroxide (MEL Chemicals).12
The experiments were performed by using a lab-scale pilot apparatus for a batch process that consisted of a 500 mL Pyrex reactor equipped with five openings, which was heated to 250–300 °C. A lab-scale anchor impeller stirred the medium at 150 rpm while NH3 was fed at 3 g h−1. The reactor was connected to a demister thermostatted at 130 °C by an oil circulating system, and the demister was connected to a condenser (12 °C) and a cold trap (−78.5 °C) and repelled vapor towards a graduated funnel, which enabled the measurement of the collected ammoniated water while preventing NH3 from leaving the set-up (Figure 17). The reactor was filled with either 170±1 g of technical oleic acid (Oleon, Radiacid 0210, C18:1 at 71.7 wt %, C18:2 at 7.2 wt %, C16:1 at 7.1 wt %, C16 at 4.4 wt %, C14+C14:1 at 3.4 wt %, C18:3 and C18 at 1.1 wt %, C20+C22 at 0.2 wt %) or 85±1 g of technical oleic amide [TCI, C18:1 at >65.0 wt %, repartition similar to Radiacid observed by GC with flame ionization detection (FID)], which was heated to 150 °C under a flow of N2.
At 150 °C, NH3 was added and the temperature was increased until the desired value was reached (250–300 °C). The beginning of the introduction of NH3 was noted as t=0. Every hour, the amount of collected ammoniated water was noted and samples were withdrawn from the reactor for analysis. The experimental conditions in the lab-scale pilot apparatus were adapted to fit to the actual factory performance, such as the NH3 bubbling efficiency (shape and depth of the feeding tube), the stirring speed (150±10 rpm to homogenize without splashing), and the water deprivation circuit (demister with Rashig rings, water condenser, dry-ice trapping). This calibration step corresponds to 0.05 mol % of acid remaining, which is 99.95 % conversion, after 10 h of reaction at 300 °C with 0.23 mol % (0.065 wt %) crystalline zinc oxide feed at t0 and an NH3 flow rate of 3 g h−1 for 170 g of starting material. The system was not in equilibrium because of the continuous feeding of one of the reactants and the intended displacement of the dehydration equilibrium by evaporating the produced water and evacuating it through the demister. These parameters were of major influence in the apparent kinetics as they decided the displacement of the early and late steps of the reaction; thus, after calibration, their stability was controlled for the results to be comparable. The parameters examined in this study were the nature and content of the catalyst, the NH3 flow rate (considering its possible recycling), and the operating temperature.
Liquid acid–base titration analysis was performed on the withdrawn samples by using a Mettler-Toledo T70 titrator. For the titration of acids, the sample was diluted in a ternary solvent (10 % isopropanol, 15 % ethylene glycol, 75 % chloroform) and titrated by KOH 0.1 mol L−1 in methanol by using a DGi116-Solvent electrode. Both acids and ammonium carboxylates were indistinctly titrated with this method (i.e., only one signal was obtained that corresponded to the totality of acid plus ammonium carboxylate). As the ammonium salt was completely dissolved and reprotonation of the carboxylate was unlikely in this solvent, this indistinct titration could stem from the high difference in pKa with the titrating agent. For the titration of amides, the sample was diluted in acetic anhydride and titrated by 0.1 M HClO4 in acetic acid by using a HA405 DPA electrode. The standard deviation of this method reaches a relative 1 %, and the evolution of the withdrawn samples with time was observed as negligible.
In parallel to liquid titration, the withdrawn samples were analyzed by GC–FID (Perkin–Elmer Clarus 500). The sample was injected directly on-column in a 5 m precolumn connected to a 30 m (DB-WAX 30 m×0.53 mm×50 μm) capillary column. GC–MS analysis was also performed by using a 30 m (ELITE-WAX ETR 30 m×0.25 mm×0.5 μm) capillary column.
The analysis did not reveal any superposition of the tallest peaks; however, the MS are very similar for several mono-unsaturated C18 compounds, and 1H NMR spectroscopy is poorly effective in distinguishing a ω-9 from a ω-7 species or other internal unsaturation. 13C NMR spectroscopy could provide a good indication of the location of the unsaturation, which was presented by Knothe et al. for similar long-chain compounds.20 Until a distance of eleven carbon atoms with regard to the end of the chain, the carbon atoms of the unsaturation appear at two different shifts, and the difference between these shifts decreases obeying a law of order three for unsubstituted octadecenoic acids. This would be a very useful and adapted analytic tool if starting from a pure compound; however, in our study, the signal was more complicated to analyze as the reactant was a statistical mixture. Ozonolysis or softer oxidation conditions (e.g., dilute KMnO4, OsO4) could allow the study of such isomer mixtures by observing the fragmentation products; however, no such standard procedure was developed in the present study. HPLC or 2D GC could also be envisaged to separate such isomer mixtures.
1H NMR spectroscopic quantification of protons of the CH2 group in the α position of the functionality was also tried. However, the superposition of the amide signals and the protons in the α position of the unsaturation was observed. Treatment of the sample can cause the amide protons to be shielded; however, no standard experimental protocol could be established as the treatment can also shift the amide signal until superposition with that of the acid.
In this study, the conversion of every acid functionality without regard to chain length was sought, and thus liquid titration was more adapted as it provided information on the actual acid, ammonium salt, and amide contents. The nitrile content evaluated by GC–FID was in agreement with the value obtained by mass balance from liquid titration measurements. GC–FID analyses provided direct information on the absence of significant amounts of side products of similar carbon-chain-length products (below 20 carbon atoms). Regarding polymers, no direct measurements could be performed but, by deduction from both liquid titrations and GC, the concentration was considered to be below 0.5 mol %.
ICP-OES was performed on selected kinetic samples after decantation. This decantation proceeded as follows: the 1–2 mL capped samples were heated for 48 h in an oil bath at 90 °C, then the surface liquid material was removed by slow pipetting. The melting point for pure oleamide is 102–104 °C; however, technical amide with approximately 70 % oleic chain was used and its melting point measured by differential scanning calorimetry (DSC) was 70±1 °C.
Adsorption microcalorimetry coupled with volumetry was used to evaluate the acid–base features of the catalysts in matters of amount and strength. The sample was inserted inside a C80 microcalorimeter from Setaram set at 80 °C, and accurate dosing of NH3 or SO2 was performed. Adsorption–desorption–readsorption cycles provided measurements of the irreversible volume of adsorption by difference of the adsorbed and readsorbed volumes at 27 Pa equilibrium pressure, which is an estimation of the number of adsorption sites [μmol m−2], and the successive thermograms and corresponding adsorbed quantities provided measurements of the strength of the adsorption sites [J m−2]. The adsorption of NH3 probes the acidic sites, whereas the adsorption of SO2 probes the basic sites. Values measured for a selection of acid–base catalysts are displayed in Table 1.
Table 1. Acid–base properties and catalytic results for the investigated catalysts after 8 h reaction.
[a] BET specific surface area. [b] Irreversible volume of adsorption measured at 27 Pa. [c] Integrated enthalpy of adsorption over the irreversible volume.
The authors are thankful to the scientific services of IRCELYON and to the very helpful master work of Didier Grondin. The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 241718 EuroBioRef.