Journal of Chromatography A, 1388 (2015) 141–150

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Interfacing supercritical fluid reaction apparatus with on-line liquid chromatography: Monitoring the progress of a synthetic organic reaction performed in supercritical fluid solution Edward D. Ramsey ∗ , Ben Li, Wei Guo, Jing Y. Liu Sustainable Technology Research Unit, University of Science and Technology Liaoning, Anshan 11401, Liaoning Province, China

a r t i c l e

i n f o

Article history: Received 3 November 2014 Received in revised form 12 January 2015 Accepted 11 February 2015 Available online 17 February 2015 Keywords: Supercritical Organic reaction Process analytical technology

a b s t r a c t An interface has been developed that connects a supercritical fluid reaction (SFR) vessel directly on-line to a liquid chromatograph. The combined SFR-LC system has enabled the progress of the esterification reaction between phenol and benzoyl chloride to synthesize phenyl benzoate in supercritical fluid carbon dioxide solution to be dynamically monitored. This was achieved by the periodic SFR-LC analysis of samples directly withdrawn from the esterification reaction mixture. Using the series of SFR-LC analysis results obtained for individual esterification reactions, the reaction progress profile for each esterification reaction was obtained by expressing the measured yield of phenyl benzoate as a function of reaction time. With reaction temperature fixed at 75 ◦ C, four sets (n = 3) of SFR-LC reaction progress profiles were obtained at four different SFR pressures ranging from 13.79 to 27.58 MPa. The maximum SFR yield obtained for phenyl benzoate using a standard set of reactant concentrations was 85.2% (R.S.D. 4.2%) when the reaction was performed at 13.79 MPa for 90 min. In comparison, a phenyl benzoate yield of less than 0.3% was obtained using the same standard reactant concentrations after 90 min reaction time at 75 ◦ C using either: heptane, ethyl acetate or acetonitrile as conventional organic reaction solvents. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Environmentally damaging, toxic and relatively expensive organic solvents are extensively used in synthetic manufacturing processes. These solvents are often destroyed by incineration once they can no longer be purified and recycled and this contributes to carbon dioxide emissions as well as hazardous transportation issues. It is well established that: non-toxic, non-flammable, noncorrosive, odorless, inexpensive, inexhaustible, ubiquitous, and low disposable cost carbon dioxide in its supercritical fluid state can be manipulated via density control to imitate a range of nonpolar organic solvents [1,2]. With careful consideration [3] this potentially makes supercritical fluid carbon dioxide (SF-CO2 ) an attractive option in appropriate circumstances to replace or partially replace the use of some organic solvents widely used for synthetic manufacturing processes. In addition, synthetic reactions that can be conducted in supercritical fluid solution may proceed at significantly faster reaction rates compared to what is achievable using conventional organic reaction solvents [4]. This is generally

∗ Corresponding author. Tel.: +86 412 5928239; fax: +86 412 5929627. E-mail address: [email protected] (E.D. Ramsey). http://dx.doi.org/10.1016/j.chroma.2015.02.037 0021-9673/© 2015 Elsevier B.V. All rights reserved.

attributed to more efficient mass transfer occurring in supercritical fluids whose physical state characteristically provides lower viscosities than liquids. At the end of SFR synthetic procedures involving the use of SF-CO2 , carbon dioxide can be conveniently expanded to waste compared to energy demanding organic solvent removal processes that are generally required to isolate products prepared using conventional synthetic procedures. Alternatively, depending upon the scale of the SF-CO2 process, it is possible to recycle decompressed carbon dioxide following its purification and recompression. Due to the potential benefits, a wide range of synthetic organic reaction procedures using SF-CO2 as reaction solvent have been developed [5–9]. Despite this, the use of SF-CO2 as an alternative solvent for synthetic reactions has not yet been widely adopted. In order to help further advance the development and optimization of synthetic procedures performed in supercritical fluid solution it is advantageous to monitor SFR processes on-line. Development of on-line process analytical technology (PAT) although being relatively new continues to evolve across various manufacturing sectors [10,11]. This involves a range of applications, particularly within the pharmaceutical industry [12–16] where rapid process development and optimization for the manufacture of high value products is of key importance. Real time or near time PAT is

142

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

differentiated from off-line laboratory techniques in that it typically provides process analysis results in seconds to minutes as opposed to hours or days. The majority of PAT procedures currently introduced or being developed for applications involving conventional solvent based synthetic organic chemistry utilize different types of on-line spectroscopic techniques [17,18]. However, on-line HPLC has also been used for the development of synthetic organic reactions performed in conventional solvents [19–21]. Only very few reports have described the use of PAT with synthetic organic chemistry SFR equipment. These PAT studies have involved the on-line coupling of spectroscopic techniques [22–24]. Although less rugged than PAT spectroscopic instrumentation, online chromatographic procedures are well suited for SFR feasibility studies and small-scale SFR process development studies that may result in a mixture of products being formed. Additionally, chromatography based PAT procedures provide potential for the rapid detection of unexpected by-products particularly at trace levels. Liquid chromatography is currently established as the dominant chromatographic technique used in the pharmaceutical industry. Accordingly, the principle objective of this study was to develop a near time PAT method for small scale SFR apparatus using on-line liquid chromatography. The interfacing of supercritical fluid extraction (SFE) systems, which utilize SFR type vessels, with liquid chromatography has been reviewed [25–27]. With coupled SFE–LC instrumentation, intermediate sorbent traps are often used to collect and preconcentrate extracts prior to LC analysis. Intermediate SFE–LC sorbent traps generally require the use of solvent flushing stages to transfer trapped extracts for analysis whilst preventing decompressed gas from being introduced into the HPLC system. Such sorbent trap arrangements have typically involved the use of either a ten port valve [28,29] or multi valve SFE–LC interface [30,31]. A simplified SFE–LC interface that utilizes a single six port valve with an analyte sorbent trap has also been developed [32]. This report describes the use of a two valve interface that enables the direct on-line coupling of SFR apparatus to a liquid chromatograph. The SFR-LC interface does not utilize an intermediate sorbent trap which simplifies its mode of operation and facilitates higher sampling rates making it more suitable for near time PAT applications. The SFR-LC interface design ensures that a representative sample of reaction mixture dissolved in supercritical fluid solution is rapidly withdrawn from within the SFR vessel for direct on-line liquid chromatographic analysis. Using SFR-LC, the progress of the esterification reaction between phenol and benzoyl chloride to synthesize phenyl benzoate in SF-CO2 was studied. In order to compare SFR performance with respect to reaction rate and product yield, the same esterification reaction was also conducted in three different aprotic conventional organic reaction solvents of different polarity.

2. Experimental 2.1. Materials and reagents Analar grade: phenol, benzoyl chloride, methyl benzoate, phenyl benzoate, and biphenyl were obtained from ACROS Organics (Shanghai, China). HPLC grade methanol and water were obtained from Sigma-Aldrich (Shanghai, China). Analar grade heptane, ethyl acetate and acetonitrile were obtained from Dikma Technologies (Beijing, China). SFR-LC interface chromatographic performance measurements were performed using: (i) the SFR vessel loaded with 20 ␮L methyl benzoate, 20 mg phenyl benzoate and 10 mg biphenyl dissolved in situ using SF-CO2 , and (ii) a standard solution composed of 20 ␮L methyl benzoate, 20 mg phenyl benzoate and 10 mg biphenyl dissolved in 25 mL ethanol. All SFR-LC

esterification reaction studies involved the use of a 2:1 mole excess of phenol to benzoyl chloride. This was achieved by initially loading 36 mg phenol and approximately 5 mg biphenyl into the SFR vessel after which 20 ␮L benzoyl chloride was injected into the SFR vessel at various pre-pressure values as described in Section 2.4. 2.2. SFR system A 25 mL SFE vessel obtained from Thar Designs (Pittsburgh, PA, USA) served as the SFR vessel. The original SFE vessel design was modified such that the sintered high pressure seal retainers were replaced with low dead volume direct feed through retainers. The SFR vessel was loaded with a close fitting ceramic marble. The SFR vessel was locked in an AT-330 column oven (Auto Science, Tianjin, China) whose temperature was maintained at 75 ◦ C. A Gilson 307 pump (Middleton, WI, USA) whose pump head was cooled to −15 ◦ C via a NESLAB RTE-110 recirculator (Newlington, NH, USA) was used to supply liquid carbon dioxide for all supercritical fluid solubility and SFR-LC studies. The cylinder containing 99.999% pure liquid carbon dioxide (AIRICHEM, Dalian, China) was equipped with a liquid draw-off tube. 2.3. SFR-LC interface The SFR-LC interface was constructed using two Rheodyne 7010 valves (Cotati, CA, USA). The valves were mounted on the door of a Gilson 831 oven whose temperature was maintained at 75 ◦ C. The sequence of key SFC–LC interface valve settings to make a sample injection are shown in Fig. 1. The sample injection valve was fitted with a 10 ␮L sample loop whereas the sample draw through valve was fitted with a 250 ␮L loop when standard solutions were formed in situ within the SFR vessel and a 650 ␮L loop for SFRLC studies. During sample loading, a pressurized stream of SF-CO2 solution from within the SFR vessel rapidly flows through the sample injection valve loop to fill the loop of the draw through valve. After sample injection, the aliquot of supercritical fluid solution held in the draw through valve loop was exhausted to atmosphere via a solvent wash system. At the end of completed SF-CO2 solubility or SFR-LC experiments, the SFR-LC interface was cleaned. The procedure involved disconnecting the SFR vessel and replacing the blank pin on port 3 of the draw through valve with a length of open tubing that discharged into a beaker. A syringe loaded with ethyl acetate was then connected to the line that had been used to connect the SFR vessel to the SFR-LC interface sample injection valve. Using the series of valve switch positions shown in Fig. 1, the SFRLC interface was cleaned by flushing with several 5 mL aliquots of ethyl acetate. The cleaned SFR-LC interface was then dried in situ using a flow of warm gaseous carbon dioxide. During disconnection from the SFR-LC interface, the SFR vessel and ceramic marble were cleaned by sonication in ethyl acetate and then thoroughly dried. After cleaning and drying, the SFR vessel and SFR-LC interface were reconnected and port 3 of the sample draw through valve was refitted with the blank pin. The complete cleaning and reconnection process could be accomplished within 25 min. 2.4. SFR system operation and SFR-LC sample injection protocol Phenol, phenyl benzoate and biphenyl samples were introduced into the SFR vessel on weighing boats made from thin aluminum foil whereas methyl benzoate was directly introduced into the open SFR vessel using a 25 ␮L syringe. Prior to the start of liquid carbon dioxide delivery, the loaded and sealed SFR vessel was allowed to thermally equilibrate for 30 min using an oven temperature of 75 ◦ C. The SFR vessel reached final selected operational target pressure via a Gilson 307 pump that utilized two flow rate programs. The first program provided a liquid carbon dioxide flow at 5 mL min−1

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

143

Fig. 1. Sequence of SFR-LC interface valve positions required to complete one sample analysis.

to the SFR vessel until a pre-pressure value of 2.76 MPa less than each selected final target pressure to conduct SF-CO2 solubility or SFR-LC studies was reached. The pre-pressure was maintained for 15 min to allow the SF-CO2 within the SFR vessel to thermally equilibrate at 75 ◦ C. During this period, the small oven containing the locked SFR vessel was manually rocked at 5 min intervals. During SFR vessel rocking, the mixing imparted via the rolling motion of the ceramic marble served to promote: (i) compound dissolution, (ii) even distribution of the concentrations of dissolved compounds, and (iii) even heat distribution throughout the SF-CO2 solution contained within the SFR vessel. At the end of the 15 min period, the pre-pressure value was increased to the final selected target SFR vessel pressure by the pump continually dispensing liquid carbon dioxide at a flow-rate of 2 mL min−1 . This second pump up stage was completed in 45–55 s during which time the SFR vessel was continually rocked. Rocking was maintained for a period of 1.5 min after final target pressure was reached. Immediately prior to making a SFR-LC sample injection, the SFR vessel was rocked for 30 s before loading the 10 ␮L sample valve loop. Loading the sample valve loop was accompanied by a small drop in the selected SFR vessel target pressure that was rapidly restored by the pump dispensing further liquid carbon dioxide at 2 mL min−1 to compensate for the total sample volume withdrawn to fill both the SFR-LC interface valve loops. A 10 s delay was introduced between loading the sample injection valve and making the SFR-LC sample injection to ensure the injection was made at the fully restored selected target pressure. For all supercritical fluid solubility and SFR-LC studies involving the use of benzoyl chloride, it was injected into the SFR vessel via a Rheodyne 7125 valve fitted with an 80 ␮L sample loop. This valve was plumbed into the supply line that connected the liquid carbon dioxide pump to the SFR vessel. Benzoyl chloride injections into the SFR vessel were made at the end of the 15 min pre-pressure thermal equilibration period, immediately when the pump started to continuously deliver the further liquid carbon dioxide required to reach the final selected SFR vessel target pressure.

DAD system, collecting data from 210 to 300 nm. For all analyses, the mobile phase was methanol:water (73:27 v/v) with a flow rate of 1 mL min−1 . The analysis column was a 150 mm × 4.6 mm i.d. DIAMONSIL® column obtained from Dikma Technologies packed with 5 ␮m C18 stationary phase. The analysis column was fitted with a Dikma EasyGuardTM pre-column containing a replaceable 10 mm × 4.4 mm i.d. cartridge packed with 5 ␮m C18 stationary phase. 2.6. Safety notes The SFR system and SFR-LC interface were housed within a fume hood. Special precautions were required to safely exhaust the SFR vessel at the end of the reaction period and also the exhaust from the SFR-LC interface after each sample injection. If the supercritical fluid reaction solution was decompressed directly to atmosphere, then micronization due to the rapid expansion of supercritical solution (RESS) could have resulted in the formation of organic compound nano-particles dispersed in a rapidly expanding aerosol type medium. Organic compound nano-particles can exhibit significantly enhanced biological activity [33] that may pose a severe health threat if contacted with the: skin, eyes, nasal tract, or digestive system. Consequently, both the SFR vessel contents and SFR-LC interface were exhausted through a three-stage gaseous exhaust washing system consisting of three 1 L Schott bottles (BDH, Poole, Dorset, UK) each containing 700 mL ethanol. The bottles were connected in series via tubes sealed within their screw caps. The bottles were housed within an enclosure fabricated from a metal case with a polycarbonate window. Once the flow of exhaust gas had passed through the sequential ethanol wash system, its pressure was significantly reduced enabling it to be safely discharged via an exhaust line contained within the fume hood. 3. Results and discussion 3.1. Stability of UV/vis detector

2.5. Liquid chromatography Liquid chromatography was performed using two Shimadzu LC-10AS pumps (Kyoto, Japan) interfaced to a Shimadzu SCL10A System Controller under Shimadzu Class VP software control. Detection was performed using a Shimadzu SPD-M10AVP UV/vis

Initial studies were performed to determine whether the SFR-LC system would provide stable UV/vis detector operation following direct injection of 10 ␮L samples of SF-CO2 . For these studies, the SFR vessel was operated at 75 ◦ C with pressures ranging from 13.79 to 27.58 MPa. After the short time taken for the initial 10 ␮L injection of supercritical fluid to decompress whilst passing through the

144

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

Fig. 2. Where: (a) is the model esterification reaction between phenol and benzoyl chloride in supercritical carbon dioxide solution and (b) is the reaction of benzoyl chloride with the HPLC mobile phase. Although not shown, hydrogen chloride is produced with each reaction.

columns, an intense gas peak was obtained at all DAD wavelengths. This is attributed to the large change in refractive index during the time required for the decompressed gaseous carbon dioxide to exit the detector flow cell. For the range of pressures tested, stable UV/vis detector performance was maintained with a methanol content of at least 50% (v/v) in an otherwise aqueous mobile phase using a flow rate of 1 mL min−1 . With lower methanol content, the UV/vis detector started to become highly unstable due to the retention of carbon dioxide gas bubbles within the UV/vis detection cell. These findings are in general agreement with a previous report [34] that has also studied this issue. During the current study it was established that for mobile phases with relatively high water content, post-column addition of methanol eliminated UV/vis detector instability problems caused by decompressed carbon dioxide bubbles. It has also been reported [35] that the use of a restrictor attached to the outlet of the UV/vis cell provides an effective means to maintain detector stability for SFE–LC. Throughout these current studies, the HPLC mobile phase consisted of methanol:water (73:27 v/v) such that no problems with UV/vis detector instability were encountered. 3.2. Supercritical fluid solubility studies In order to investigate the feasibility of developing a near time PAT procedure involving SFR-LC, the model reaction shown in Fig. 2a was selected. This involves the esterification reaction between phenol and benzoyl chloride to synthesize phenyl benzoate. As shown in Fig. 2b, any unreacted benzoyl chloride injected into the HPLC methanol:water mobile phase will react to form methyl benzoate and benzoic acid. Phenol is sufficiently soluble in SF-CO2 [36] for the experimental quantity used in the esterification reaction to be completely dissolved. However, a literature search failed to find any solubility data for benzoyl chloride in SF-CO2 . The SFR-LC system can also be used to obtain solubility data for a single compound or components within a mixture. In order to establish whether benzoyl chloride was sufficiently soluble in SF-CO2 for the scale of the model esterification reaction a solubility study was undertaken. This involved injecting 40 ␮L of benzoyl chloride, representing a two fold increase above that to be used in the model esterification reaction, into the SFR vessel. The procedures used to inject benzoyl chloride into the SFR vessel and make sample injections using the SFR-LC interface are described in Section 2.4. Since the first selected target pressure to investigate the

solubility of benzoyl chloride in SF-CO2 solution was 12.07 MPa, a SFR vessel pre-pressure of 9.31 MPa was used. Immediately after the first sample injection at 12.07 MPa, the pressure within the SFR vessel was rapidly increased to the second selected solvating target pressure of 13.79 MPa. During the 30–35 s required for the pump to dispense sufficient carbon dioxide for the SFR vessel to reach 13.79 MPa the SFR vessel was continuously rocked. The second target pressure of 13.79 MPa was maintained throughout the first 10 min taken to analyze the first sample during which the SFR vessel was rocked at 2 min intervals. A second injection of the SF-CO2 solution of benzoyl chloride obtained at 13.79 MPa was then made after 10 min. Thereafter, using the same procedure two further injections of the SF-CO2 solution of benzoyl chloride were made at 17.24 MPa and 20.68 MPa at 10 min intervals. Fig. 3 shows the series of chromatograms obtained for the four injections of benzoyl chloride dissolved in SF-CO2 solution made at 10 min intervals using the progressively higher SFR vessel target pressures. As anticipated, benzoyl chloride reacted with the HPLC mobile phase according to Fig. 2b to produce benzoic acid and methyl benzoate whose peaks were assigned using the retention times obtained for pure standards. The peaks obtained for methyl benzoate in the chromatograms shown in Fig. 3 exhibit severe tailing of a type not observed for methyl benzoate in the chromatograms shown in Fig. 4. This indicates that benzoyl chloride continually reacts with the HPLC mobile phase throughout the time required to elute benzoyl chloride from the columns. Peaks 3, 6, 9 and 12 in the chromatograms shown in Fig. 3 are tentatively assigned as unreacted benzoyl chloride. This assignment is supported by the fact that if 20 ␮L of benzoyl chloride is diluted and mixed with 25 mL methanol:water (73:27 v/v) for 40 min prior to conventional HPLC liquid sample injection, only two fully resolved symmetrical peaks with no significant tailing are detected at retention times corresponding to benzoic acid and methyl benzoate. Since the solubility of organic compounds generally increase with increasing supercritical fluid density [37], the intensities of the series of the chromatograms shown in Fig. 3 indicate that 40 ␮L benzoyl chloride is completely soluble in one SFR vessel volume of SF-CO2 at pressures at or above 12.07 MPa. This is surmised since the overall intensity of the chromatograms obtained for benzoyl chloride did not increase with increasing SF-CO2 pressure. Instead a small progressive decrease in the intensity of the series of benzoyl chloride chromatograms is shown in Fig. 3. This decrease is due to the relatively small volume of the SFR vessel compared to the

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

145

Fig. 3. Chromatograms obtained for benzoyl chloride dissolved in supercritical fluid carbon dioxide solution. First injection made at 0 min with pressure at 12.07 MPa. Second injection made at 10 min with pressure at 13.79 MPa. Third injection made at 20 min with pressure at 17.24 MPa. Fourth injection made at 30 min with pressure at 20.68 MPa. For benzoic acid peak numbers are: 1, 4, 7 and 10. For methyl benzoate peak numbers are: 2, 5, 8 and 11. For benzoyl chloride peak numbers are: 3, 6, 9 and 12. Initial gas peak for each injection is off-scale.

volume of SF-CO2 solution withdrawn to make each sample injection via the SFR-LC interface. Hence, each injection of the supercritical solution of completely solvated benzoyl chloride resulted in a small yet detectable dilution of the SFR vessel contents. Apart from the method of introducing the compound into the SFR vessel, the same procedure used to investigate the solubility of benzoyl chloride in SF-CO2 was applied to phenol, phenyl benzoate and biphenyl. For the phenol solubility study, the standard HPLC mobile phase composition of methanol:water (73:27 v/v) used for all other studies was changed to another methanol:water (50:50

v/v) composition to separate the phenol and carbon dioxide gas peaks. The results of the further solubility studies confirmed that each of the three compounds to be either used or formed during the esterification reaction would be totally soluble in one SFR vessel volume of SF-CO2 using the quantities of reagents specified in Section 2.1. Hence, it was established that SFR-LC could be used to monitor the changes in the concentrations of all the compounds in the reaction mixture throughout the entire time required to complete the esterification reaction. The solubility measurements therefore served to help validate the development of a SFR-LC

Fig. 4. Chromatograms obtained for four injections of a standard mixture of three compounds dissolved in supercritical fluid carbon dioxide solution at 20.68 MPa and 75 ◦ C. Injections using the SFR-LC interface made at: 0, 20, 40, 60 min. For methyl benzoate peak numbers are: 1, 4, 7 and 10. For phenyl benzoate peak numbers are: 2, 5, 8 and 11. For biphenyl peak numbers are: 3, 6, 9, and 12. Initial gas peak for each injection is off-scale.

procedure to quantitatively monitor the progress of the esterification reaction. Where: tr = retention time (min), N = theoretical plates, k = retention factor, Tf = asymmetry factorb (5% peak height), Rs = resolution. a n = 4. Values in brackets are R.S.D. (%). b Where: Tf = wb /wf (At a specified peak height: wb and wf are the respective partial peak width values measured from the back and front of a perpendicular line constructed between the peak apex and the baseline).

Tf k N tR

1.07 (0.96) 1.62 (2.7) 3.25 (0.32) 1.22 (1.7) 7.03 (0.19) 1.08 (0.4) 2022 (2.0) 3454 (2.2) 9403 (0.94) – 3.90 (0.5) 10.71 (1.24) 8.02 (0.24) 12.70 (0.67) 15.15 (0.17) 1.10 (0.94) 1.39 (2.7) 3.31 (0.53) 1.14 (1.9) 7.15 (1.29) 1.09 (2.8) Methyl benzoate 3.96 (0.49) 2114 (2.1) Phenyl benzoate 8.14 (0.41) 3777 (2.3) Biphenyl 15.37 (0.66) 9508 (1.6)

Pressure at 20.68 MPa

Rs Tf k N tR

Pressure at 13.79 MPa

Rs Tf k N tR

Direct injection at atmosphere Compound

Table 1 Mean calculateda chromatographic performance parameter values obtained for sample injections made at different pressures via the SFR-LC interface. Interface temperature at 75 ◦ C.

– 3.87 (1.0) 1992 (2.1) 10.28 (0.5) 7.94 (0.42) 3031 (2.0) 12.63 (0.83) 15.04 (0.21) 8876 (2.4)

Rs

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

1.05 (1.96) 1.65 (0.7) – 3.21 (0.55) 1.36 (3.6) 9.94 (1.56) 6.98 (0.23) 1.07 (0.38) 12.29 (0.4)

146

3.3. SFR-LC interface chromatographic performance In order to evaluate the impact on chromatographic performance using the SFR-LC interface, studies were undertaken that involved injecting standard SF-CO2 solutions containing a dissolved mixture of methyl benzoate, phenyl benzoate and biphenyl at two different SFR vessel pressures. Preparation of the two solutions involved loading the open SFR vessel with standard quantities of the three compounds. Once loaded, the SFR vessel was sealed and test solutions were prepared in situ at 13.79 MPa and 20.68 MPa using the procedure described in Section 2.4. For comparative purposes an ethanolic standard solution of the three compounds was also analyzed by injections made using the SFR-LC interface operated at atmospheric pressure. This was achieved by disconnecting the SFR-LC interface sample injection valve from the SFR vessel and draw through valve. Thereafter, the SFR-LC interface sample injection valve was manually loaded with the ethanolic standard solution using a liquid sample syringe whilst the SFR-LC interface oven temperature was maintained at 75 ◦ C. For each pressure used to evaluate SFR-LC interface performance, four replicate sample injections were made. The four chromatograms obtained for the analyses of the standard solution standard prepared in situ using SF-CO2 at 20.68 MPa and 75 ◦ C are shown in Fig. 4. Table 1 summarizes the chromatographic performance results derived from the chromatograms obtained at the three different SFR-LC interface injection pressures. Comparison of the results shown in Table 1 indicates that only a small progressive decline in chromatographic performance is obtained for the SF-CO2 standard solutions injected at increasing pressures compared to conventional liquid sample injection. 3.4. On-line SFR-LC esterification reaction monitoring studies A series of synthetic SFR studies were undertaken using a fixed SFR vessel temperature of 75 ◦ C. These SFR-LC studies involved monitoring the progress of the esterification reaction at four different SFR pressures ranging from 13.79 to 27.58 MPa. To initiate each esterification reaction, 20 ␮L of benzoyl chloride was injected into the SFR vessel using the method described in Section 2.4. This procedure ensured that both the solid phenol and biphenyl initially loaded into the SFR vessel had sufficient time to fully dissolve before introducing the benzoyl chloride into the SFR vessel. The first SFR-LC sample injection was made 10 min immediately after the benzoyl chloride was injected into the SFR vessel. Thereafter, a series of esterification reaction mixture samples were analyzed at 20 min intervals throughout a reaction time of 170 min. Following each SFR-LC sample injection, the SFR vessel was periodically rocked for 15 s at approximately 2 min intervals. The function of the non-reactive biphenyl was to act as a reporter species to confirm that no SFR-LC interface problems such as those caused by blockage or leaks occurred throughout the course of the SFR-LC studies. An example of the chromatograms obtained for the first three SFR-LC analyses for an esterification reaction performed at a SFR pressure of 13.79 MPa is shown in Fig. 5. The chromatographic peak for unreacted phenol is not observed since phenol eluted with the carbon dioxide gas peak. As shown in Fig. 5, the phenyl benzoate product peak becomes progressively more intense whereas the peaks obtained for and derived from the remaining non-reacted benzoyl chloride become progressively smaller as reaction proceeds. The progressive small decline in the intensity of the biphenyl peak observed in the SFC–LC chromatograms shown in Fig. 5 confirmed stable SFR-LC performance during the monitored reaction period using the small volume SFR vessel. Inspection of the peak

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

147

Fig. 5. First three SFR-LC analyses for the model esterification reaction performed in supercritical fluid carbon dioxide solution at 13.79 MPa and 75 ◦ C. Injections made at: 0, 20, 40 min equating to reaction times of: 10, 30, 50 min. For methyl benzoate peak numbers are: 1, 5 and 9. For benzoyl chloride peak numbers are: 2, 6 and 10. For phenyl benzoate peak numbers are: 3, 7 and 11. For biphenyl peak numbers are: 4, 8 and 12. Initial gas peak for each injection is off-scale. Phenol co-elutes with the gas peak.

amplitudes for the methyl benzoate and non-reacted benzoyl chloride shown in the chromatograms in Fig. 5, for unknown reason(s) do not exhibit the same relative intensity pattern compared to those shown in Fig. 3. The small variation in retention times for the peaks shown in Fig. 5 compared to the retention time data provided in Table 1, less than 3.5% for phenyl benzoate, is attributed to small variations in retentive capacity introduced via the periodic replacement of pre-column C18 cartridges. Replacement of these cartridges became necessary to maintain good chromatographic performance during the SFR-LC reaction monitoring studies. Pre-column cartridges were replaced if retention times or peak integral values for a test solution injected by conventional means varied by more than 4%. The chromatographic performance check was frequently performed using the SFC–MS interface operated at atmospheric pressure using the same three component ethanolic standard solution used to provide the data shown in Table 1. Three separate esterification reactions were performed at each of the four SFR pressures. A new method was used to determine the SFR yields of phenyl benzoate. Construction of the calibration graphs used for determining phenyl benzoate yields involved initially loading the SFR vessel with known quantities of phenyl benzoate acting as quantification standards. The SFR vessel was then sealed and a calibration standard solution was prepared in situ by dissolving the phenyl benzoate according to the procedure described in Section 2.4. For each of the four SFR pressures, a set of three calibration standard solutions were prepared in situ using exactly one SFR vessel volume of SF-CO2 . Once prepared, triplicate injections of each calibration standard supercritical fluid solution were made using the SFR-LC interface. With the origin considered as a data point, a linear calibration graph was obtained for each of the four SFR pressures with each calibration graph providing a correlation coefficient value of greater than 0.99. This approach was adopted since although the SFR vessel loaded with the marble was known to have a volume approximating to 22.5 mL its exact volume was not accurately known. Consequently, the use of conventional off-line phenyl benzoate calibration standards made in a suitable organic solvent was considered to be less accurate for determining SFR phenyl benzoate yields compared to using in situ prepared calibration standard solutions.

The series of SFR-LC analysis results obtained for each esterification reaction were used to construct the reaction progress profile for each individual reaction. This was achieved by expressing the measured yield of phenyl benzoate as a function of the reaction time when the SFR-LC sample injection was made. Fig. 6 shows the reaction progress profiles obtained for each separate esterification reaction. The results shown in Fig. 6 establish that an inverse relationship exists between reaction rate and supercritical fluid reaction pressure for the esterification reaction. For each set of three reaction progress profiles obtained at the four different SFR pressures, cross-over points are observed in Fig. 6. A possible cause for this observation is that the esterification reaction was only periodically mixed using a manual procedure. This may have resulted in slight variations in the SFR conditions at each reaction pressure. Ideally in order to help promote reproducible reaction progress and yields, all synthetic organic reactions performed in solvent

Fig. 6. SFR-LC reaction progress profiles obtained for each individual esterification reaction performed in triplicate at four different reaction pressures. Temperature for all reactions was 75 ◦ C.

148

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

Table 2 Meana phenyl benzoate yield (% theoretical) determined at different reaction times using on-line SFR-LC analysis. SFR vessel temperature at 75 ◦ C. Pressure (MPa)

10 min

30 min

50 min

70 min

90 min

110 min

130 min

150 min

170 min

13.79 17.24 20.68 27.58

28.5 (24.2) 11.7 (35.5) 11.3 (5.5) 4.2 (13.2)

63.3 (11.5) 36.5 (19.0) 18.7 (5.6) 8.0 (4.4)

71.0 (4.7) 49.6 (5.7) 22.9 (6.1) 10.4 (1.2)

72.6 (6.4) 55.0 (5.5) 24.9 (7.2) 11.6 (5.6)

72.7 (2.0) 57.3 (4.3) 27.8 (6.9) 11.3 (10.3)

70.4 (2.2) 57.5 (7.5) 28.2 (3.9) 12.1 (3.8)

68.8 (1.9) 56.6 (3.8) 29.2 (4.8) 12.2 (5.1)

66.7 (1.9) 55.7 (5.6) 29.9 (5.4) 12.3 (4.9)

64.2 (2.7) 54.5 (4.9) 30.3 (5.5) 12.3 (4.5)

a

n = 3. Values in brackets are R.S.D. (%).

systems should be continuously and efficiently stirred to ensure even distribution of reactant concentrations. Continuous stirring also helps provide even exposure of the reaction mixture to any reaction vessel hot spots and/or catalytic sites particularly in the walls of metallic reaction vessels. Table 2 summarizes the SFR-LC results obtained for the yields of phenyl benzoate at the four SFR pressures. Inspection of the data shown in Table 2 indicates that for the 13.79 MPa and 17.24 MPa SFR pressures, a maximum phenyl benzoate yield is obtained at 90 min and 110 min reaction time, respectively. Thereafter, for each of these two SFR pressures a reduction in phenyl benzoate yield was determined. This anomaly arises due to the small SFR vessel volume resulting in a small yet significant reaction mixture dilution following each sample injection at and after the maximum SFR-LC reaction yield was measured at these two reaction pressures. Despite this experimental apparatus shortcoming, provision of the type of data shown in Table 2 using off-line HPLC would require the analysis of nine times more individual esterification reactions compared to the SFR-LC method. To obtain each single off-line HPLC analysis result a number of steps would also be required, these include: (i) immediately stopping the esterification reaction at a specific reaction time, (ii) quantitative recovery of all the SFR vessel contents, (iii) sample preparation, and (iv) final HPLC analysis. These steps would involve considerable manual manipulation and would be difficult to reproducibly perform. Therefore, process optimization for the esterification reaction in SF-CO2 using off-line HPLC analysis apart from being far less convenient would require significantly greater time to provide equivalent results to those obtained using the on-line SFR-LC procedure. As shown in Table 2, the maximum SFR phenyl benzoate yield determined using SFR-LC monitoring was 72.7% (R.S.D. 2.0%). This yield was obtained using a reaction pressure of 13.79 MPa after 90 min reaction time. However, prior to this SFR-LC yield being determined four samples had been withdrawn from the SFR vessel for analysis. This would have served to progressively reduce the concentrations of the reagents, critically as indicated in Fig. 6 during the initial relatively fast reaction rate period. Therefore, in order to obtain a more accurate phenyl benzoate yield at 13.79 MPa after 90 min reaction time a further set of three esterification reactions were performed. The SFR-LC sampling protocol for each individual reaction involved making the first sample injection at 90 min reaction time after the reaction had been initiated followed by a second sample injection after 110 min reaction time. Throughout the 110 min reaction period, the SFR vessel was periodically rocked to emulate the mixing used in the multiple sample SFR-LC reaction progress monitoring studies. With this procedure the average yield obtained for phenyl benzoate after 90 min reaction time was 85.2% (R.S.D. 4.2%). The set of three chromatograms obtained, indicated that the esterification reaction was effectively complete after 90 min. This was surmised since the peak integration values for both phenyl benzoate and biphenyl declined in almost identical percentage value for the injection at 110 min relative to the corresponding peak integration values obtained at 90 min. The same SFR yield of phenyl benzoate was obtained at 13.79 MPa after 90 min reaction time when no biphenyl was added to the SFR vessel. This result confirmed that biphenyl played no part in the SFR process.

3.5. Synthesis of phenyl benzoate in conventional organic solvents In order to compare the rate of reaction and yield of phenyl benzoate obtained using non-polar SF-CO2 with that obtainable using conventional organic reaction solvents a series of experiments were conducted. Three conventional aprotic organic reaction solvents of different polarity were selected on the basis that this would provide a reasonable range for comparison with the SFR performance that had been obtained using non-polar SF-CO2 . The organic solvents selected were heptane, ethyl benzoate and acetonitrile. All organic solvent reactions were performed in a glass reaction flask equipped with a reflux condenser and a continuous mechanical stirrer. Organic solvent reaction mixtures containing dissolved phenol, benzoyl chloride and biphenyl were prepared on the basis that the SFR vessel had a volume of 22.5 mL. The initial concentration of the three compounds prepared in each of the three organic solvents was approximately the same as that used in the SFR-LC studies. In order to further directly compare the performance of the esterification reaction in the organic solvents with that obtained using SF-CO2 , the esterification reactions in the three organic solvents were each performed at 75 ◦ C for 90 min. At the end of 90 min reaction time an aliquot of organic solvent reaction mixture was withdrawn and analyzed using off-line HPLC. For each organic solvent a phenyl benzoate yield of less than 0.3% was obtained. This compared to a maximum SFR phenyl benzoate yield of 85.2% (R.S.D. 4.2%) using the equivalent reaction conditions. In order to improve the yield of phenyl benzoate using acetonitrile as the reaction solvent a more concentrated reaction solution was prepared. This solution provided a ten fold increase in the concentrations of each of the three compounds compared to that used in the SFR-LC studies. Reaction was performed at 75 ◦ C but the reaction time using the concentrated acetonitrile reaction solution was increased to 180 min to further help increase the yield of phenyl benzoate. After 180 min reaction time an aliquot of the concentrated reaction solution was withdrawn, immediately diluted ten fold with acetonitrile and analyzed by HPLC. Although the phenyl benzoate yield had increased, it remained low at 2.3%. This result confirmed that the synthesis of phenyl benzoate in SF-CO2 solution proceeds at a very fast reaction rate relative to that in any of the three organic solvents. The above results show similar characteristics to those reported [23] for the Diels–Alder reaction between N-ethymaleimide and anthracene. Using on-line UV/vis spectroscopy, the reaction rate for the Diels–Alder reaction was found to be twenty five times faster in SF-CO2 compared to the reaction being performed in acetonitrile. In addition, as with the esterification reaction, the Diels–Alder reaction exhibited an inverse relationship between reaction rate and the density of SF-CO2 reaction solvent. 4. Conclusions A SFR-LC interface constructed using only two valves housed within an oven has provided a successful means to rapidly monitor the progress of a synthetic organic esterification reaction performed in SF-CO2 solution. The SFR-LC near time PAT system provides the capability to rapidly assess the feasibility of performing and/or optimizing reactions in supercritical fluid solution.

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

On-line SFR-LC can provide substantial potential savings in terms of both time and cost compared to SFR process development using off-line analytical procedures. In addition, SFR-LC process monitoring is well suited for situations involving light and/or air sensitive compounds. The SFR-LC system also provides the capability to obtain solubility data either for a single compound or individual compounds in a mixture in supercritical fluid solution. Ideally, SFRLC studies should involve withdrawing samples from a reaction mixture whose constituent reactants and products remain totally soluble in the supercritical fluid throughout the time required to complete the reaction. In accordance with current PAT trends for manufacturing scale plant, it seems highly probable that more field rugged spectroscopic techniques such as those based on near infrared technology will continue to be far more frequently used than on-line chromatographic procedures for process monitoring. Therefore, SFR-LC near time PAT may be limited to initial small-scale research and development applications including SFR pilot plant studies. For such small scale SFR systems, techniques involving SFR-LC–MS or SFR-LC–MS–MS may be particularly well suited for the rapid detection and characterization of unexpected SFR process by-products especially any formed at trace levels. The results obtained in this study illustrate the potential for rapid and efficient synthetic organic reactions being conducted in SF-CO2 compared to the use of conventional organic reaction solvents. Since the model esterification reaction used in this study resulted in the synthesis of phenyl benzoate, an exhaustive series of PAT studies involving SFR-LC were not undertaken to fully optimize reaction pressure or evaluate reaction temperatures other than 75 ◦ C. This is because relatively low cost phenyl benzoate is commercially manufactured by reacting phenol dissolved in basic aqueous solution with benzoyl chloride, typically resulting in a 75% yield [38]. However, we now anticipate using SFR-LC apparatus to evaluate the feasibility of synthesizing relatively high value compounds. The successful outcome of SFR-LC feasibility studies to synthesize such compounds would serve to justify more extensive SFR-LC process optimization studies being undertaken. Automation of the SFR-LC interface would facilitate unattended SFR-LC near time PAT studies. An automated SFR-LC system would be especially suited for a PAT application that may involve obtaining many data sets for process optimization. Accordingly, we are currently evaluating the use of a programmable valve switching module with a view to constructing a SFR-LC interface with an auto-sampler type capability. The SFR-LC interface described in this report is ideally suited to interface SFR apparatus directly on-line to a supercritical fluid chromatograph. This conveniently enables changing the mode of chromatographic operation to perform alternative SFR-SFC reaction progress monitoring and/or orthogonal on-line product purity checks. In addition, SFR-SFC should be generally suitable for situations that require the analysis of moisture sensitive compounds that would react with water in reverse phase HPLC mobile phases. The interface used in the SFR-LC esterification study has now been successfully used, in conjunction with a larger volume custom manufactured SFR vessel equipped with a continuous mechanical stirrer, to perform a SFR-SFC study [39] of an organic reaction involving the use of an enzyme.

Acknowledgement Edward D. Ramsey wishes to express his sincere gratitude to the 1000 Experts Award Program sponsored by the Central Government of China, Project Number WQ20122100062, who provided funding to help facilitate these studies.

149

References [1] L.T. Taylor, Supercritical Fluid Extraction, John Wiley & Sons, New York, NY, 1996. [2] J.W. King, J.E. France, Basic principles of analytical supercritical fluid extraction, in: B. Wenclawiak (Ed.), Analysis with Supercritical Fluids: Extraction and Chromatography, Springer-Verlag, Berlin, 1992, pp. 32–60. [3] E.J. Beckman, Supercritical and near critical CO2 in green chemical synthesis and processing, J. Supercrit. Fluids 28 (2004) 121–191. [4] E.D. Ramsey, Q. Sun, Z. Zhang, C. Zhang, W. Guo, Mini-Review. Green sustainable processes using supercritical fluid carbon dioxide, J. Environ. Sci. 21 (2009) 720–726. [5] Y. Ikushima, M. Arai, Stoichiometric organic reactions, in: P.G. Jessop, W. Leitner (Eds.), Chemical Synthesis Using Supercritical Fluids, Wiley-VCH, Weinheim, 1999, pp. 259–279. [6] R. Scott Oakes, A.A. Clifford, C.M. Rayner, The use of supercritical fluids in synthetic organic chemistry, J. Chem. Soc., Perkin Trans. 1 (2001) 917–941. [7] D. Prajapati, M. Gohain, Recent advances in the application of supercritical fluids for carbon-carbon bond formation in organic synthesis, Tetrahedron 60 (2004) 815–833. [8] C.M. Rayner, P.M. Rose, D.C. Barnes, Synthetic organic reactions in supercritical fluids, in: P.T. Anastas (Ed.), Handbook of Green Chemistry Volume 4 Supercritical Fluids, Wiley-VCH, 2010, pp. 189–241. [9] S. Mayadevi, Reactions in supercritical carbon dioxide, Indian J. Chem. 51A (2012) 1298–1305. [10] J. Workman, B. Lavine, R. Chrisman, M. Koch, Review of key developments in process analytical chemistry from 2009 to 2011, Anal. Chem. 83 (2011) 4557–4578. [11] W. Chee, P. Sharratt, Critical review trends in process analytical technology, Anal. Methods 10 (2010) 1412–1438. [12] L.X. Yu, Pharmaceutical quality by design: product and process development, understanding, and control, Pharm. Res. 25 (2008) 781–791. [13] E.N.M. van Sprang, H.-J. Ramaker, M. Streefland, L.A. van der Pol, E.C. Benvery, A.G. Smilde, Manufacturing vaccines: an illustration of using PAT tools for controlling the cultivation of Bordetella pertussis, Qual. Eng. 19 (2007) 373–384. [14] Z. Chen, D. Lovett, J. Morris, Process analytical technologies and real time process control a review of some spectroscopic issues and challenges, J. Process Control 21 (2011) 1467–1482. [15] D.J. Ertl, S.E. Barnes, B.E. Cooley, C.A. Gross, Applications of PAT to API synthesis from R&D to manufacturing, Am. Pharm. Rev. May–June (2009) 66–72. [16] A. Chandra, A.M. Daly, D.A. Foley, M.A. LaPack, S. Mukherjee, T.D. Orr, G.L. Reid, D.R. Thomson, H.W. Ward, Industry perspectives on process analytical technology: tools and applications in API development, Org. Process Res. Dev. (2014), http://dx.doi.org/10.1021/op400358b. [17] A. Bakeev (Ed.), Process Analytical Technology—Spectroscopic Tools and Implementation Strategies for the Chemical and Pharmaceutical Industries, second ed., Wiley, Chichester, 2010. [18] R. Mendhe, A.S. Ruthore, I.S. Krull, Tools for enabling process analytical technology applications in biotechnology, BioPharm. Int. 25 (2012) 23–35. [19] O.W. Gooding, T. Lindberg, W. Miller, E. Munyak, L. Vo, Case studies in process screening and optimization utilizing a new automated reactor-analysis system, Org. Process. Res. Dev. 5 (2001) 283–293. [20] C.J. Welch, X. Gong, J. Cuff, S. Dolman, J. Nyrop, F. Lin, On-line analysis of flowing systems using microflow HPLC, Org. Process Res. Dev. 13 (2009) 1022–1025. [21] D.A. Foley, J. Wang, B. Maranzano, M.T. Zell, B.L. Marquez, Y. Xiang, G.L. Reid, Online NMR and HPLC as a reaction monitoring platform for pharmaceutical process development, Anal. Chem. 65 (2013) 8928–8932. [22] S.M. Howdle, M.W. George, M. Ploliakoff, Spectroscopy of SCF solutions, in: P.G. Jessop, W. Leitner (Eds.), Chemical Synthesis Using Supercritical Fluids, Wiley-VCH, Weinheim, 1999, pp. 147–164. [23] J. Qian, M.T. Timka, A.J. Allen, C.J. Russell, B. Winnik, B. Buckley, J.I. Stenfeld, J.W. Tester, Solvophobic acceleration of Diels–Alder reactions in SC-CO2 , J. Am. Chem. Soc. 126 (2004) 5465–5474. [24] H. Fisher, O. Gyllenhaal, J. Vessman, K. Albert, Reaction monitoring of aliphatic amines in SC-CO2 by proton NMR spectroscopy and implication for SFC, Anal. Chem. 75 (2003) 622–626. [25] A.T. Rees, Supercritical fluid extraction for off-line and online high-performance liquid chromatographic analysis, in: E.D. Ramsey (Ed.), Analytical Supercritical Fluid Extraction Techniques, Kluwer Academic, Dordrecht, 1998, pp. 287–352. [26] T. Hyotylainen, M.-J. Riekkola, Approaches for on-line coupling of extraction and chromatography, Anal. Bioanal. Chem. 378 (2004) 1962–1981. [27] J. Pan, C. Zhang, Z. Zhang, G. Li, Review of online coupling of sample preparation techniques with liquid chromatography, Anal. Chim. Acta 815 (2014) 1–15. [28] E.D. Ramsey, B. Minty, A.T. Rees, Drugs in water: analysis at the part-per-billion level using direct supercritical fluid extraction of aqueous samples coupled on-line with ultraviolet–visible diode-array liquid chromatography–mass spectrometry, Anal. Commun. 34 (1997) 261–264. [29] E.D. Ramsey, A.T. Rees, G. Wei, J.Y. Liu, X.H. Wu, Direct aqueous supercritical fluid extraction coupled on-line with liquid chromatography-tandem mass spectrometry for the analysis of polyether antibiotics in water, J. Chromatogr. A 1217 (2010) 3348–3356. [30] F. Ischi, W. Haerdi, Supercritical fluid extraction-high performance liquid chromatography on-line coupling: extraction of some model aromatic compounds, Chromatographia 41 (1995) 238–242. [31] R. Batlle, H. Carlsson, E. Holmgren, A. Colmsjo, C. Crescenzi, On-line coupling of supercritical fluid extraction with high-performance liquid chromatography

150

E.D. Ramsey et al. / J. Chromatogr. A 1388 (2015) 141–150

for the determination of explosives in vapour phases, J. Chromatogr. A 963 (2002) 73–82. [32] Z. Wang, M. Ashraf-Khorassani, L.T. Taylor, Design for on-line coupling of supercritical fluid extraction with liquid chromatography: Quantitative analysis of polynuclear aromatic hydrocarbons in a soil matrix, Anal. Chem. 75 (2003) 3979–3985. [33] K. Mishima, Biodegradable particle formation for drug and gene delivery using supercritical fluid and dense gas, Adv. Drug Deliv. Rev. 60 (2008) 411–432. [34] M. Ashraf-Khorassan, M. Barzegar, Y. Yamini, On-line coupling of supercritical fluid extraction with high performance liquid chromatography, J. High Res. Chromatogr. 18 (1995) 472–476.

[35] M.H. Liu, S. Kapila, K.S. Nam, A.A. Elseewi, Tandem supercritical fluid extraction and liquid chromatography system for determination of chlorinated phenols in solid matrices, J. Chromatogr. 639 (1993) 151–157. [36] G.-G. Julian, M.J. Molina, F. Rodriguez, F. Mirada, Solubilities of phenol and pyrocatechol in supercritical carbon dioxide, J. Chem. Eng. Data 46 (2001) 918–921. [37] R.B. Gupta, J-J. Shim, Solubility in Supercritical Carbon Dioxide, CRC Press, Boca Raton, FL, 2007. [38] Y.-S. Lee, M.-Y. Yen, Y.-P. Shih, Phase-transfer catalytic kinetics of the synthesis of phenyl benzoate, Ind. Eng. Chem. Res. 34 (1995) 1572–1580. [39] B. Li, G. Wei, E.D. Ramsey, unpublished results.