Environmentally benign and diastereoselective synthesis of 2,4,5-trisubstituted-2-imidazolines

School of Chemical Engineering & Analytic The Mill, Sackville Street, Manchester, manchester.ac.uk; Tel: +44 (0)161 306 4366 School of Chemistry, The University of Ma 9PL, UK † Electronic supplementary information ( NMR, C NMR, MS, HRMS, FTIR spec kinetic calculations; DFT computational m data, membrane separation. CCDC 157 crystallographic data in CIF or o 10.1039/c7ra11827a ‡ These authors equally contributed to th Cite this: RSC Adv., 2017, 7, 53278


Introduction
2-Imidazoline derivatives are an important class of heterocycles with a wide range of industrial applications (Fig. 1). They exhibit biological activity and can be found in natural products such as topsentines or spongotines, which have been attractive for biomedical purposes due to their antitumor, antiviral and anti-inammatory characteristics. 1 Several pharmaceutical compounds that contain a 2-imidazoline ring have been synthesised and studied for blood pressure control, hyperglycemia, depression and schizophrenia. 2 Imidazolines and imidazolinium ions have been used as surfactants, 3 ionic liquids, 4 ionic polymers, 5 and in the eld of coordination chemistry and catalysis. 6-10 Among the 2-imidazoline derivatives, 2,4,5-triaryl imidazolines (also called amarines) have been gaining increased interest because their hydrolysis provides 1,2-diaryl-1,2-diaminoethanes featuring a C 2 -symmetric structure, important in catalysis. [11][12][13] The reported syntheses for obtaining 2,4,5-trisubstituted-2imidazolines are summarised in Table 1. 2,4,5-Trisubstituted-2-imidazolines were successfully synthesised from benzylamine and carbon tetrachloride, 14 and 1,2-cyclic sulphates and amidine. 15 However, none of these syntheses were as prominent as the reaction from benzaldehyde and ammonia. 16 Benzaldehyde reacts with liquid ammonia to produce hydrobenzamide (also known as 1,3,5-triaryl-2,4-diazapentadiene), an intermediate which upon heating or employing strong base deprotonates and subsequently forms amarine, supposedly via disrotatory pericyclic closure. 11,17 The rate of pericyclisation usually depends on both the steric and electronic character of the substituents. 18 However, the effect of substituents on the synthesis of 2-imidazolines and the reaction mechanism have yet to be investigated.
Notwithstanding the moderate to good yields, most reactions require additives, extensive purication, or harsh conditions, i.e., high temperature, toxic or banned solvents. In addition, additives and catalysts were usually used for the preparation of the desired product, which results in additional waste that has to be managed. While attempts have been made towards a more sustainable synthesis of 2,4,5-trisubstituted-2imidazolines, there is a need for exploring green solvents and purication processes.
The 21st century's sustainability grand challenge requires the chemical industry to employ mild reaction conditions, and non-toxic, non-ammable chemicals and solvents, with the aim of minimising environmental impact. 21 Solvents for reaction and purication account for more than 80% of the total mass used for the production of pharmaceuticals, 22 and consume 60% of the total energy. 23 The solvent selection guides established by pharmaceutical companies provide practical advice to reduce the environmental burden associated with solvent usage. 24,25 Following the guidelines, sixteen different green solvents were screened for the syntheses presented in this work. As an effective process intensication tool due to the mild conditions and low energy consumption, 26 membrane technology is a sustainable alternative to conventional purication techniques such as crystallisation and chromatography, 27 which can be more easily implemented in continuous processing. 28 Consequently, the potential of membrane-based purication of the 2-imidazoline products was evaluated.
Herein, we report the development toward an environmentally benign, diastereoselective, one-pot synthesis of 2,4,5trisubstituted-2-imidazolines employing non-toxic ammonium salts, green solvents and purication. The kinetics and reaction mechanism were explored to reveal the governing factors of the pericyclisation. To the best of our knowledge the kinetic studies and quantum mechanical calculations for the reaction of aldehyde and ammonia to produce 2,4,5-trisubstituted-2imidazolines have not yet been reported.

Results and discussion
The reaction optimisation was initiated by investigating the effect of temperature and solvent on the reaction of 4-(tri-uoromethyl)benzaldehyde and (NH 4 ) 2 CO 3 ( Table 2). In line with the 12 principles of green chemistry, 21 all reactions were performed at mild temperatures (20-50 C) in solvents, which were chosen according to GlaxoSmithKline's solvent selection guide. 25 Although, the isolated yield markedly increased with the temperature up to 40 C (entries 1-3), a considerable decrease at 50 C was observed (entry 4), which indicates the  decomposition of (NH 4 ) 2 CO 3 . The inuence of solvent is presented in entries 3 and 5-16 in  30 Considering the tradeoff between selectivity and reaction rate, 2-MeTHF, ethyl lactate, propylene carbonate and sulfolane provide a suitable solvent for the reaction as they produced 1 with high selectivity in a manageable reaction time. However, since the product isolation from the last three is more solvent and energy intensive (extraction and evaporation), 2-MeTHF was used for further reaction optimisation.
To study the effect of ammonia sources (2 eq.) on the reaction, non-toxic ammonium salts of acids covering a wide range of pK a values (À6.30 to 10.32) were selected ( Table 3). The highest conversion was achieved using (NH 4 ) 2 CO 3 (entry 1), while moderate and poor yield was obtained with NH 4 HCO 3 and NH 4 OAc (entries 2 and 3), respectively. NH 4 HCOO (entry 4) did not provide the desired product at 40 C, but a moderate yield was obtained at the reux temperature of 81 C. Moderate yield was also achieved using NH 4 CF 3 COO at reux (entry 5). In contrast, no product was formed with NH 4 Cl (entry 6) under reux conditions. These results suggest that employing the ammonium salt of a weaker acid (therefore with higher pK a value) leads to increased reaction rate due to the more favoured dissociation of the salt and the more basic environment. In addition, the amount of ammonia source was also investigated (entry 1). The increase of ammonium carbonate from 1 to 2 eq. led to 23% yield enhancement. However, further increase to 4 eq. did not result in signicant improvement.
The effect of both electron withdrawing and electron donating p-substituents on the reaction were systematically investigated ( Table 4). Substrates with strongly or moderately electron-withdrawing groups (EWG) (entries 1-7), i.e., moderate to high Hammett sigma constants (s ¼ 0.45-0.82), provided the desired product in good to excellent yield. The substrate with weakly electron-withdrawing bromide group (entry 8; s ¼ 0.23) only afforded the 2-imidazoline derivative in poor yield. Reactions with non-EWGs (-H, -OH) were attempted but did not yield the imidazoline products.
In order to further explore the effect of EWGs on the reaction rate, the kinetics of the reaction was studied. The reactions of (NH 4 ) 2 CO 3 with 4-nitrobenzaldehyde, 4-(pentauorosulfanyl)benzaldehyde, 4-(triuoromethyl)benzaldehyde, and methyl 4formyl benzoate were chosen as model systems because they  Table 4 The effect of p-substituents on the reaction a provided the desired product under the same conditions with a considerable yield, and they cover a wide range of s (0.45-0.78). A consecutive reaction, 3A / B / C, was suggested for the reaction kinetics based on the formation of stable intermediate B identied as hydrobenzamide in the literature. 11 In order to obtain the reaction rate constants k 1 and k 2 , the rate eqn (1)-(3) have to be solved.
are the concentrations of component of the aldehyde, hydrobenzamide, and product, respectively; and n stands for the rate order of the rst reaction. The B / C reaction is considered as rst order because the transformation from the hydrobenzamide to product is an intramolecular ring closure. However, the order of the A / B reaction is not that evident because it is composed of several elementary steps including, the formation of aldehyde-ammonia adducts and water elimination. The rate order of the reaction was chosen based on the following method: the value of n was varied from 1 to 5 and then the obtained differential equations were tted to the experimental data with the least square method. The goodness of tting was evaluated with the sum of squared deviations (SSD) and a rate order was selected where the SSD (summarised in Table S1, ESI †) was at the minimum. The lowest SSD was obtained at n ¼ 3 in three out of four sets of experimental data. Consequently, a third order model was chosen as the overall rate order for the rst chemical transformation. As shown on the plots in Fig. 2, the conversion proles in most cases (-NO 2 , -CF 3 , -COOMe) are sigmoidal: the product formation is slow in the rst period, however, it increases in the second period, then it decreases again. Finally, a plateau is reached when no further product is formed. The sigmoidal plots indicate the formation and accumulation of hydrobenzamide intermediate, which reveals that the A / B reaction is faster than the B / C reaction. Due to the rapid reaction rate for the rst step of the -SF 5 derivative synthesis the sigmoidal shape is not prominent. In order to investigate the relationship between the reaction rates and the electron-withdrawing substituents, the calculated k 1 and k 2 values are displayed as a function of the s in Fig. 3a. Although increasing rate constants for both k 1 and k 2 were expected with increasing s no direct relationship between the three constants was found. The rate of hydrobenzamide formation increases from s ¼ 0.45 (-COOMe) until s ¼ 0.68 (-SF 5 ), however, k 1 signicantly drops at s ¼ 0.78. The rate constant of the intramolecular ring closure decreases between s ¼ 0.45 (-COOMe) and s ¼ 0.54 (-CF 3 ) and then grows with increasing electronwithdrawing property. These results could be the consequence of other factors, such as low solubility of the intermediate.
Density functional theory computational analysis (DFT B3LYP/6-31G(d,p)) revealed the cyclisation mechanism of 1 from the hydrobenzamide intermediate (1hb) (Fig. 4). The deprotonation of 1hb forms a high energy carbanion intermediate (1cU), which stabilises by the translation of the  ammonium counter ion. Subsequent protonation of 1cU on its nitrogen atom results in 1zw zwitterionic intermediate, which provides 1 via ring closure. The Gibbs free energy of both the cyclisation (DG r ) and the deprotonation (DG deprot ) was derived for substituents with different electron-withdrawing properties (different s values). The dependence of DG r on s is negligible; however, the DG deprot values signicantly decrease with higher s values (Fig. 3b). Consequently, the reaction favours EWGs, which is in line with the experimental observations (see Table 4).
The DFT results refuted two speculations about the mechanism of the cyclisation. First, the cyclisation from the carbanion species with electron-withdrawing substituents is not favourable thermodynamically (Table S2, ESI †), despite speculations that the reaction proceeds with the ring closure of the carbanion intermediate leading to the negatively charged imidazoline which is the deprotonated form of 2-imidazoline (Scheme S1a, ESI †). 11 Second, 1cU is the energetically most favoured carbanion conformer (Table 5). Despite speculations that 1cW has the lowest energy due to the lack of steric hindrance between H(4) and H(5) 16 1cU has a quasi-planar structure, which allows the conjugation of the p system to be extended to the whole molecule. However, the steric repulsion between the H(5) and H(34) atoms in 1cS and 1cW forces the ring to turn out of the plane, which decreases the conjugation. The cyclisation was further studied using nuclear independent chemical shi (NICS) calculations. 34 NICS (0) values of À1.94 and À3.55 ppm were obtained for the ve-membered rings in 1zw and 1, respectively. These relatively small negative NICS (0) values revealed the non-aromatic nature of the rings. In contrast, the large negative NICS (0) value of À11.34 ppm for 1 TS-IV proved the aromaticity of the transition state, and consequently, the pericyclic nature of the reaction. The lobes in the highest occupied molecular orbital (HOMO) of 1zw indicate that the cyclisation is allowed progress only in disrotatory mode leading to the cis product (1) according to the Woodward-Hoffman rule. 35 The experimental results are in good agreement with the calculations because for all substrates, the formation of solely the cis product was observed. The stereochemistry of 1 (Fig. 5a), 2, and 5 ( Fig. S115 and S116, ESI †) was conrmed by singlecrystal X-ray diffraction. The stereochemistry of 3, 4 and 6-8 was deduced to be cis as the synthetic methods were identical. This deduction was further supported by the high coupling constants (10.9-11.2 Hz) for the imidazoline hydrogens in the NMR spectra.
As a consequence of the electrocyclic nature of the reaction, the diastereoselectivity can be inuenced by the method of reaction initiation. Photoirradiation at 254 nm and 365 nm for 96 h resulted in 39% and 13% conversion with trans : cis product ratio of 1.42 and 1.61, respectively (Scheme 1).
Furthermore, the thermodynamically favoured trans 9 can be selectively obtained by alkali treatment of the cis isomer, similarly to a previously described procedure. 13 The trans stereochemistry of 9 (Fig. 5b) was indicated by the lower coupling constant of imidazoline hydrogens (J ¼ 8. 8 Hz) compared to the case of 1 (J ¼ 10.9 Hz).
Besides p-substituted aldehydes, the synthesis was also performed from o-, and m-substituted aldehydes, in order to investigate the effect of the substituent position on the reaction (Table 6). Moreover, attempts were made to expand the substrate scope to heterocyclic and non-aromatic aldehydes as well. Comparing the position of substituents (entries 1-2), the m-substitution provided the desired product in good yield, but the reaction was slower compared to the p-substituted analogue. The o-substitution did not afford the imidazoline product as a consequence of steric effects, i.e., the close proximity of the bulky -CF 3 substituent to the formyl group sterically hinders the reaction. When it was attempted to produce the product from 4-pyridinecarboxaldehyde and (NH 4 ) 2 CO 3 with the optimised conditions, a product mixture of imidazoline and imidazole (1 : 1) was obtained. However, when NH 4 OAc was used as ammonia source, the desired 2-imidazoline product was isolated in a good yield of 72%. This result suggests that, similarly to 3, product 11 is also sensitive to oxidation under basic conditions. The synthesis of 12 was performed using trans-cinnamaldehyde and (NH 4 ) 2 CO 3 as shown in entry 4 in Table 6. The reaction provided the product in good yield within 24 h, which indicates that highly conjugated aldehydes also afford the desired 2-imidazoline product. Therefore, the reaction was also attempted using ethyl 4-oxobut-2-enoate as a nonaromatic starting material. Similarly to trans-cinnamaldehyde, the consumption of starting material was virtually 100% within 24 h (entry 5). The reaction, however, did not stop at the 2imidazoline product, but through an intramolecular double bond migration a trisubstituted imidazole product was formed instead. Consequently, it could be concluded that a conjugated system which is sterically unhindered is required for the presented synthetic methodology to yield 2-imidazolines.
Most green chemistry articles solely focus on improving the synthetic step without considering the purication and isolation of the product. In order to address the inherent environmental burden of downstream processing, the isolation of the 2imidazoline products was aimed to be performed in a single crystallisation step employing a green solvent, which eliminated the use of tedious chromatography employed in previous  procedures (see Table 1). Furthermore, the considerable increase in molecular size during the reaction allowed the use of solvent-resistant nanoltration for product purication. Four membranes, namely GMT-oNF-1, GMT-oNF-2, NF030306, and Duramem 300, were screened at 10-40 bar pressure to obtain solvent ux and solute rejection for product 1, and the corresponding benzaldehyde substrate impurity (Fig. 6a). Flux is dened as volume of solvent that permeates the membrane per unit area in a given time, while rejection is the ratio of solute concentration in the permeate and the retentate. Permeance is dened as the ux normalised by the applied pressure. The permeances (expressed in L m À2 h À1 bar À1 ) obtained for GMT-ONF-1, GMT-ONF-2, NF030306, and Duramem 300 were 4.39 AE 0.6, 5.2 AE 0.6, 0.35 AE 0.04, and 0.26 AE 0.01, respectively. Despite the relatively low permeance obtained for Duramem 300, it showed excellent product rejection above 99%. The concentration and purity proles for single-stage and two-stage dialtration are shown in Fig. 6b. The detailed description of the mathematical framework can be found in the ESI. † The number of diavolumes is a time-like parameter indicating the progress of the ltration and it is dened as the ratio of permeate and retentate volume. The number of diavolumes required to obtain 99% product purity with a single-stage and two-stage dialtration is 5 and 6, respectively (see dotted lines). In line with expectations, 36 the two-stage cascade conguration markedly improved the product yield from 95% to 99.8%. The purication of a post-reaction mixture validated the calculated model. During the process, Duramem 300 at 20 bar demonstrated stable permeance and rejection performance over 8 days of continuous operation in 2-MeTHF (Fig. S118, ESI †). The solvent consumption of the ltration could be minimised by implementing an additional ltration stage for in situ solvent recovery. 28,37 Conclusions The diastereoselective synthesis of 2,4,5-trisubstituted-2imidazolines was successfully realised from aldehydes and ammonia sources in a one-pot procedure at mild temperature employing green solvents with moderate to high polarity. The effect of the chemical nature of the aldehyde, in particular the position and polarity of aromatic substituents, was systematically investigated. It was found that strongly EWGs with high Hammett sigma constant (e.g. -NO 2 , -CF 3 ) provide the 2- The reactions were performed with aldehyde (3 mmol), (NH 4 ) 2 CO 3 (6 mmol), and 2-MeTHF (6 mL) at 40 C. b Isolated yield. c The ammonia source was NH 4 OAc (6 mmol). d Formation of the desired 2-imidazoline was not observed, but 13 was obtained instead. imidazoline product in moderate to good yield. In contrast, weakly EWGs (-Br) and electron-donating groups (-OH) poorly or did not afford the desired product, respectively. Besides the aromatic aldehydes, the synthetic methodology is applicable to sterically unhindered, conjugated analogues. The DFT results refuted two speculations and conrmed that (i) the ionic intermediate gets protonated prior to cyclisation, (ii) the most thermodynamically stable carbanion conformer is U-shaped. DFT modelling, NMR and singlecrystal XRD conrmed that thermal initiation solely yields the cis isomer. It was demonstrated that the diastereoselectivity of the reaction can be changed by photoinitiation, resulting in excess trans isomer, which is a consequence of the electrocyclic nature of the reaction. The feasibility of membrane-based product purication was demonstrated, which provides a green alternative for the conventional chromatographic purication of 2,4,5-trisubstituted-2-imidazolines.

Analytical methods
All reactions were monitored with thin layer chromatography (TLC) using silica gel 60 F 254 (Merck) plates; and with high performance liquid chromatography (HPLC). The optimum reaction time, i.e., maximum achievable conversion was determined by HPLC. HPLC analyses were measured on a VWR Hitachi Chromaster instrument with 5160 pump, 5260 autosampler, 5310 column oven, and 5430 diode array detector (DAD). See ESI † for the HPLC methods. Melting points (values without correction) were measured on a Stuart melting point apparatus SMP3 using 5 C min À1 slope.
The products were characterised by IR, 1 H NMR, 13 C NMR, 19 F NMR, 1 H COSY, 1 H-13 C HMBC and 1 H-13 C HSQC, mass spectrometry (MS) and, where possible, single-crystal XRD measurements. IR spectra were recorded from dry samples using a Thermo Fisher Nicolet iS5 iD5 ATR-FTIR spectrometer. NMR analyses were performed on a B500 Bruker Avance II+ 500 MHz instrument. Spectra were processed using the MestReNova soware. Low resolution mass spectra were recorded on a Waters SQD2 Single l Quadrupole Mass spectrometer used with a Waters Acquity UHPLC using 100% methanol as mobile phase. Electrospray ionisation technique was used in negative or positive mode. High resolution mass measurements were performed on a Thermo Exactive plus EMR Orbitrap mass spectrometer, used with a Thermo Ultimate 3000 UHPLC using 100% methanol as mobile phase. Single-crystals were grown by slow evaporation at low-temperature (2 C) from acetonitrile for X-ray analysis. CCDC 1576088 (1), CCDC 1576089 (2), CCDC 1576090 (5) and CCDC 1582071 (9) contain the supplementary crystallographic data for this paper. †

General synthetic procedure
In a typical procedure, a benzaldehyde substrate (3 mmol) was dissolved in the solvent (6 mL), followed by the addition of the ammonia source (6 mmol). The reaction mixture was stirred vigorously at the desired temperature. The reaction was followed by TLC (MeOH-CHCl 3 1 : 10 v/v) and HPLC. The reaction was terminated when maximum conversion was reached by ltering the ammonia source. The solvent was evaporated and the crude product was recrystallised.
Cis-2,4,5-tris(4-(trimethylammoniumyl)phenyl)-4,5-dihydro-1H-imidazole iodide (2). Amendments to the general procedure: the aldehyde was dissolved in a mixture of 2-MeTHF (5 mL) and DMSO (1 mL). Upon termination of the reaction, the residue was poured onto 50 mL acetone, followed by the ltration of the precipitate. 1 , 0.48 mmol). The reaction mixture was heated to 155 C and was maintained at that temperature for 2 h, during which time the sodium salt of 9 precipitated. The reaction mixture was cooled down to room temperature, treated with acetic acid glacial (1 mL), diluted with ethanol (5 mL), followed by heating to 100 C until all remaining solid product dissolved. The mixture was cooled down and basied with 35% NH 4 OH (10 mL) to pH ¼ 9 upon which white solid precipitated. The precipitation was ltered, washed with ethanol (5 mL). 1   (108 mg) following recrystallisation from toluene (5 mL). Product 9 was obtained as white crystals. Cis-2,4,5-tris(3-(triuoromethyl)phenyl)-4,5-dihydro-1H-imidazole (10). Amendments to the general procedure: upon completion of the reaction, the residue was treated with 25 mL 1 M HCl solution in order to dissociate the hydrobenzamide intermediate. The product and starting material was extracted with ethyl acetate (20 mL), followed by 35% aq. NH 3 (20 mL) wash at pH ¼ 9. The organic phase was washed with water (20 mL), and brine (20 mL). Ethyl acetate and the starting material were evaporated at 83 C, 40 mbar. 1

Membrane separation
Membranes with an area of 52.8 cm 2 were screened using a mixture of 4-(triuoromethyl)benzaldehyde and product 1, at 0.5 g L À1 concentration each. The membrane screening was performed in a typical nanoltration rig operated in cross-ow conguration (Fig. 7). A recirculation pump set at 1 L min À1 ensured homogeneous solute concentration in the retentate loop. The membranes were conditioned at each pressure for 24 hours prior to measurements to ensure steady-state operation. The solvent ux was obtained by measuring the volume of permeate passing through the membrane (V p ) in a denite time (t) and membrane area (A) as shown in eqn (4), while the solute rejection of solute was calculated from the ratio of solute concentration in the permeate (C P ) and the solute concentration in the retentate (C R ) as dened in eqn (5). The ltration of postreaction mixture was performed at 20 bar using Duramem 300 membrane with an area of 52.8 cm 2 . Fresh solvent was continuously added to compensate the permeate volume, keeping the system volume constant at 60 mL. 100 mL samples of permeate and retentate were periodically taken for analysis.

Conflicts of interest
There are no conicts to declare.