In terms of process development applications, FC provides additional advantages beyond the ones mentioned above for scale-up and for production-scale manufacturing. Foremost among them, the ability to the rapidly evaluate a wide range of reaction conditions over a short period of time is attractive for optimization of reaction conditions. For example, using a single flow-through reactor set-up, as will be described in this work, one can vary the individual pump flow rates to change the reaction stoichiometry. Alternatively, by increasing or decreasing the total flow rate, one can decrease or increase the reactor residence time and, consequently, optimize the reaction yield with respect to cycle-time productivity. Similarly, the reactor temperature and back-pressure can also be varied systematically to achieve rapid reaction optimization with respect to those two variables (typically in just a few hours). For completeness, it is pointed out here that reactions requiring microwave and/or light radiation are often more feasible in flow versus batch-mode processing [12], however, such reactions will not be discussed here further. Clearly, FC has emerged as a new tool for synthetic chemists and process engineers [13].
In order to realize the goal of real-time process optimization, there exist analytical challenges associated with on-line reaction monitoring, which is obviously the most desirable type of analysis to support continuous processing. To that end, techniques such as UV [14], NMR [15], MS [16], Raman [17], on-line HPLC [18], and FTIR [19] spectroscopy have been recently applied for the monitoring of flowing streams in "real time". In the present work, FTIR was selected because it is a non-destructive technique (unlike MS) that does not suffer from the clogging issues that can be problematic for on-line HPLC, particularly when working with heterogeneous systems. For solution-phase monitoring, FTIR is more universal than Raman spectroscopy (which is often better suited for analyzing solid samples than solutions) and, with respect to NMR, it is substantially cheaper/more readily available, and it does not require the use of deuterated solvents. It should also be noted that commercially available on-line FTIR systems are equipped with sophisticated chemometrics software, greatly simplifying the data interpretation (e.g. "auto-picking" components based on time- dependent spectral changes) and allowing the end user to monitor the relative compositions of multiple components without the analysis of raw spectral data (which is especially attractive in manufacturing settings). Moreover, it is possible to "train" the software to detect and monitor specifi c compounds, using spectra from "authentic" samples of reagents, solvents and products used in the process (as available). Quantification, as needed, can be achieved by calibrating the relative changes observed in the FTIR signals against the absolute changes in concentration as detected at specific times during the conversion by off -line HPLC analysis.
Strecker reactions [20] have previously been shown to be amenable to FC, and even improved by the use of fl ow [7]. The specific reaction discussed here uses commercially available starting materials, however it was modeled after a similar flow reaction used in the synthesis of a pharmaceutical compound currently under development. The goal of this paper is to demonstrate the utility of on-line FTIR in monitoring product formation, in real time, during the cyanation step of a Strecker reaction. Doing so ultimately allows for "on-the-fly" process optimization to quickly maximize the reaction yield of the product racemate. In this paper two new technologies are introduced. The first is a (custom-built) low-volume flow-cell that allows one to interface a standard in-situ fiber optic FTIR probe head (9.5 mm diameter) with the FC eluent stream from a reactor coil having a diameter of only 1.0 mm without loss of signal (note that a recent FTIR system with an integrated flow cell has been commercialized and marketed specifi cally for continuous process monitoring [21]; that system eliminates the fiber optic probe that is commonly available with more traditional FTIRs). The second FC technology demonstrated is a home-built, low internal volume in-line mixer that ensures that the eluent stream homogeneity is maintained as it leaves the reactor and enters the FTIR fl ow cell. Employment of such a mixer is found to be important in ensuring sample homogeneity when analyzing biphasic (liquid-liquid) reaction mixtures of the type analyzed in this work.
Experimental
Chemicals
Anisaldehyde (> 99% pure), α-methylbenzylamine (> 99% pure), dichloromethane, methanol, potassium cyanide (KCN), hydrochloric acid (HCl), trifl uoroacetic acid (TFA), glacial acetic acid, and de-ionized water were sourced commercially. In each case, the highest grade available was purchased and used without further purification.
FC Apparatus
A schematic of the experimental set-up is shown in Figure 1. Three independently controlled reagent streams were pumped by individual HPLC pumps into a four-point mixing union. Front-end pressure regulators (10 psi) ensured forward flow of the different reagent streams when operating at different flow rates in each line. The compositions of the three reagent streams were as follows, unless noted otherwise in the text:
Stream A: α-methylbenzylamine, anisaldehyde, dichloromethane and methanol, in the volume ratio: 21:20:160:37
Stream B: KCN in water, at a concentration of 0.625 g/ml
Stream C: glacial acetic acid and dichloromethane, in a 50:50 volume ratio
A 0.04-inch (1.0 mm) i.d. coil of stainless steel tubing, of 2.0 ml total capacity, comprised the reactor. A back-end pressure regulator (10 psi) was used to maintain a constant pressure drop across the reactor for all experiments. On-line FTIR (see next section) was used to monitor the eluent stream composition in order to gauge the yield impact of performing various changes to the reaction conditions. A custom-built, low internal volume (0.034 ml) fl ow cell was used in this work to interface with an existing fi ber optic probe (9.5 mm) connected to the spectrometer; an internal O-ring was used to ensure a good seal with the FTIR probe surface. The PEEK tubing connecting the fl ow cell to the reactor coil and to the product stream collection flask used finger- tight fittings: the process stream entered and exited the flow cell at 45° angles relative to the probe axis.
To ensure homogeneity of the biphasic reactor eluent stream specific to the reaction studied in this work, an in-line mixer was added prior to the analysis flow cell. The dynamic mixer was constructed using a bored-out HPLC column (5 cm length, 4.6 mm diameter) containing two small magnetic stir bars and enclosed by two column ends (10- 32). Mixing was accomplished by placing the home-built mixer on a magnetic stirring plate.
On-line FTIR Monitoring of Product Stream
The on-line FTIR used in this work was a commercially available system fitted with a standard 9.5 mm silver halide (AgX) fiber optic probe. The spectra were captured using the FTIR manufacturer’s software, which was also used to perform principle component analysis (PCA) in near real-time. Both the native IR spectra and the algorithm output (relative concentrations of each component of interest plotted as a function of the acquisition time) were exported into a spreadsheet for further data work-up/off -line analysis. Individual IR spectra, each consisting of 33 co-added scans, collected over the region 1600 – 1000 cm-1, were acquired in 15 s intervals.
O -line HPLC Monitoring of Product Stream
An HPLC system was used to obtain chromatograms separating the following components to allow for their off -line quantification (refer to Scheme 1): 1 (retention time, RT, ~ 2.10 min), 2 (RT ~ 0.34 min), 4 (RT ~ 3.60 and ~ 3.90 min; components in equilibrium). Detection was achieved using a diode array UV-visible absorbance detector monitoring the 210 nm wavelength. The column temperature was maintained at 40°C for all analyses. Separations were performed using a fused-core C-18 column (4.6 x 100 mm, 2.7μm particle size) and a mobile phase consisting of two solvents: Solvent A - aqueous phosphoric acid (0.1% v/v) and Solvent B - HPLC grade acetonitrile. The gradient program started at 90%A : 10%B, v/v, and achieved 5%A : 95%B after six minutes, with the final condition being held for two minutes. A two minute re-equilibration period (using 90%A : 10%B) was performed between sample injections. The flow rate was maintained at 1.8 ml/min for all runs. Samples from the FC reactor eluent stream were diluted ~ 100-fold in 70:30 v/v acetonitrile/water prior to injection of 5 μL aliquots onto the HPLC column.
Results and Discussion
For the Strecker synthesis of the α-amino acid, 5, the hazardous cyanation step (which evolves HCN gas) shown in Scheme 1 was performed using the FC set-up in Figure 1. The Imine intermediate (3) was formed in situ in Stream A (see Experimental section) via nucleophilic addition of the Amine (2) to the Aldehyde (1). Next, KCN (Stream B) and acetic acid (Stream C) were introduced at the mixing junction (see Figure 1). The resulting hydrocyanic acid reacts to form the Aminonitrile (4), which was actively monitored by FTIR and collected in the eluent from the flow reactor. Subsequently, through batch-mode aqueous workup (not discussed here), 4 can be hydrolyzed to generate the desired Amino Acid (5). The reagent stoichiometry and the various reagent stream compositions (see Experimental section) selected as a starting point for the process optimization discussed herein were based on previous experience with a FC Strecker synthesis of a proprietary compound.
In the experiments performed in this work, the low dead volume of the flow cell (0.034 ml, excluding the connecting tubing) allowed for rapid response times during IR monitoring of the product stream. In turn, the real-time monitoring of the reactor stream allowed rapid optimization of the Strecker reaction to be performed, as will be discussed below. Samples were pulled from the stream, during the equilibrated portions of the experiment, for parallel analysis using off - line HPLC in order to confirm the spectral assignments and to ascertain the relative concentrations of each species of interest.
FTIR spectra of all solvents and starting materials were obtained prior to performing the reaction monitoring (spectra not shown). Since both the Amine starting material (2) and the cyanide solution, Stream B, had relatively weak IR signals in the region monitored (1200 to 1450 cm-1) and because of the equilibrium that is established in solution between compounds >2 and 3, the reagent/intermediate mixture is simply referred to as the "reagent fraction" in this work. In comparison, 4 (the "product" of interest here generated in flow) was, in most cases, very easy to identify spectroscopically.
During initial FC development of the process shown in Scheme 1, the IR spectral signature of 4 was not known. To obtain the reference spectrum, 4 was generated using the flow-through reactor maintained at room temperature (~ 22°C), with Streams A-C pumped at the following flow rates: 6.5 ml/min, 0.8 ml/min, and 0. 8 ml/min, respectively. The reference spectrum was then acquired using the on-line FTIR. From the reaction profile data shown in Figure 2, 4 was generated when, at t = 13 min, the water flowing in place of Stream B was replaced by the KCN-containing solution (see Experimental section). That "new component" (i.e., non-reagent chemical), whose concentration can be seen to begin to increase at ~ 16 min in Figure 2 and then plateau at ~ 18 min, was verified to be 4 by HPLC analysis using a previously determined retention time marker. Note that the level of 4 can be seen to decrease beyond t = 24 min as the Aminonitrile was flushed out of the system during a solvent flush of the reactor coil. Via off -line HPLC analysis, the degree of conversion at steady-state production of 4 was estimated at 73% (corresponding to ~ 19 min in Figure 2).
Once the initial experiment was completed and the identity of the components of interest were confirmed using off -line HPLC, optimization of the FC conditions was initiated. First, the overall flow rate through the reactor was varied by adjusting the speeds of each pump, and fixing the relative flow ratio. This allowed the impact of the reactor residence time on the degree of conversion (product yield) to be assessed. Figure 3 demonstrates that at a constant pump flow ratio of 8.13 : 1: 1 and a fixed reactor temperature of 25°C, the slowest overall (total) flow rate tested, 4.05 ml/min, provided the best yield (77% conversion; see Table 1). A subsequent experiment utilizing a total flow rate of 2.03 ml/min (data not shown) also indicated ~77% conversion, suggesting that the reactor residence time could thus be considered optimized.
As the product stream collected after passing through the FTIR flow cell was clearly biphasic (the emulsion phase separates in minutes upon standing), there was concern about the impact of stream homogeneity aff ecting the FTIR response, particularly when changing the relative ratios of the various reagent streams. For example, different ratios of water and dichloromethane could produce emulsions of differing stability. Since the product, 4, is soluble predominantly in the organic layer, a less stable emulsion could present less of the product to the FTIR probe surface for detection than is representative of the product stream as a whole. Indeed, our experiments revealed that both higher Stream B (aqueous) relative flow rates and higher temperatures produced more "noise" (randomness/time-dependent instability) in the IR signal. That behavior was attributed to "breaking" of the emulsion as it passed from the reactor coil to the flow cell for detection. To remedy the problem, a home-built, dynamic in-line mixer (see Experimental section) was constructed and incorporated into the experimental set-up approximately 15 cm upstream from the FTIR flow cell. The efficient "regeneration" of the product stream emulsion provided cleaner signals, as evidenced by comparison of the transients in Figure 3 with those in Figure 4.
The reagent ratio optimization of the reaction was performed in a separate experiment using the previously optimized conditions as the starting point (condition "C" in Table 1); the corresponding IR data is presented in Figure 4. In the time period between 0 and ~ 32 min, the fl ow rate of either Stream B or C was varied while maintaining that of Stream A at 3.25 ml/min. It was found that increasing the flow rate of Stream B from 0.4 ml/min to 1.0 ml/min produced a notable improvement in yield of 4 (note the corresponding increase in the consumption of the reagent fraction). Someof the signal decreases observed in the product transient in Figure 4, which do not exhibit corresponding "mirror image" changes in the reagent fraction transient, are likely artifacts as evidenced by the HPLC data presented in Table 2. Lastly, from Figure 4/Table 2 it can be seen that increasing the reactor temperature from 25°C and 50°C had a minimal effect on the productivity of the process as determined by HPLC (the corresponding FTIR product signal change was minimal); going from 50°C to 75°C had no effect as determined by either analytical approach.
Ultimately, using the optimized conditions in Table 2 (i.e., a total flow rate of 5.25 ml/min, a reactor temperature of 50°C, and the following reagent stream flow rates: Stream A = 3.25ml/min, Stream B = 1.0 ml/ min and Stream C = 1.0 ml/min), the productivity of the 2.0 ml reactor coil was estimated at approximately 60 g of 4 per hour.
In the final part of the process optimization of the reaction in Scheme 1, the acid type in Stream C was varied systematically. The results of this experiment are summarized in Table 3. When using hydrochloric acid in place of acetic acid, although a generally lower product yield was obtained (confirmed by HPLC), the analytical challenge encountered in using that acid turned out to be the noisy IR signal that was observed (transients not shown). The root cause of the noise was attributed to the emulsion phase separation, whose stability was further diminished by use of the aqueous mineral acid (even with the in-line mixer in place), causing signal loss. As the use of trifluoroacetic acid (TFA), a stronger acid but one that is closely related to acetic acid, also showed limited promise in improving the reaction yield, this probe experiment was discontinued. 
Conclusion
On-line FTIR was shown to be an effective tool for performing near real-time optimization of the cyanation step of a Strecker synthesis conducted inside a flow-through reactor. The process investigated was optimized with respect to the overall flow rate/reactor residence time, the stoichiometry of the three reagent streams used in the reaction, the reactor temperature and the acid type, all within a period of only a few hours. The optimized conditions are expected to produce almost 60 g/h of the product, 4, per hour of operation using the 2.0 ml internal volume stainless steel reactor coil.
Performing on-line analyses of flowing streams is not without challenges. In the case of the FTIR used in this work, interfacing the 9.5 mm diameter probe head with the 1.0 mm diameter reactor coil was accomplished using a custom-designed/built flow cell. Given that the sampling depth of the evanescent light emanating from the probe surface is on the order of microns, the flow cell volume (< 0.1 ml) was able to be kept at a minimum so as not to compromise the response time of the FTIR analyses (which would be the result of excessive "dead volume" in the flow cell). Another technology introduced in this paper, for laboratory-scale FC applications, is the home-built dynamic in-line mixer. That mixer, while simple and inexpensive to construct, proved to be key in maintaining the homogeneity of the product stream emulsion entering the flow cell for analysis, ultimately improving the analytical precision of the FTIR measurements. Together, those two technologies, coupled to the automated data collection/PCA treatment offered by the FTIR software, allowed for successful monitoring and optimization of the Strecker reaction discussed herein.
In addition to facilitating certain types of chemistry, the use of the small-volume flow-through reactor in this work was able to minimize potential worker exposure to hazardous HCN gas relative to a batch reactor, whereby, a much larger amount of gas can be produced in a shorter period of time. While alternative analytical techniques (see Introduction) are currently under evaluation for various FC applications, on-line FTIR clearly possesses key advantages over many other techniques in its ability to monitor multiple components simultaneously, with data acquisition rates on the order of seconds. Additionally, the FTIR flow cell does not clog as easily as some on-line HPLC sampling devices. While FTIR does not have the sensitivity or selectivity of HPLC or NMR, in many FC applications (such as yield optimization) this is not necessary; i.e., only in cases where the production of critical low-level impurities (< 3%, relative to the main component) need monitoring are such techniques warranted.
Reference
- Roberge, D. M.; Zimmermann, B.; Rainone, F.; Gottsponer, M.; Eyholzer, M.; Kockmann, N. Organic Process Research & Development 2008, 12, 905.
- Kovacs, I.; Jones, R. V.; Niesz, K.; Csajagi, C.; Borcsek, B.; Darvas, F.; Urge, L. Journal of the Association for Laboratory Automation. 2007, 12, 284.
- Jas, G.; Kirschning, A. Chemistry - A European Journal 2003, 9, 5708.
- Fletcher, P. D. I.; Haswell, S. J.; Pombo-Villar, E.; Warrington, B. H.; Watts, P.; Wong, S. Y. F.; Zhang, X. Tetrahedron. 2002, 58, 4735.
- Wild, G.; Wiles, C.; Watts, P. Letters In Organic Chemistry. 2006, 3, 419.
- Baumann, M.; Baxendale, I. R.; Martin, L. J.; Ley, S. V. Tetrahedron. 2009, 65, 6611.
- Wiles, C.; Watts, P. Organic Process Research & Development 2008, 12, 1001.
- McMullen, J.P., Ducry, L., Bieler, N., Cretton, P., Zimmerman, B., Chem. Eng. Technol., 2005, 28 No.3.
- Bogdan, A.R. Poe, S.L. Kubis, D.C., Broadwater, S.J., McQuade D.T., Angew. Chem. Int. Ed. 2009, 48, 1-5.
- Baxendale, I.R., Ley, S.V., Mansfield, A.C., Smith, C.D., Angew. Chem. Int. Ed. 2009, 48 4017.
- Valera, F.E., Quaranta, M., Moran, A., Blacker, J., Armstrong, A., Cabral, J.T., Blackmond, D.G., Angew. Chem. Ind. Ed. 2010, 49, 2478.
- Criado, A.; Cárdenas, S.; Gallego, M.; Valcárcel, M. Talanta. 2000, 53, 417.
- Wiles, C.; Watts, P. European Journal Of Organic Chemistry. 2008, 1655.
- Jackman, R. J.; Floyd, T. M.; Schmidt, M. A.; Jenson, K. F.; Micro Total Analysis Systems 2000, Proceedings of the μTAS Symposium, 4th, Enschede, Netherlands, May 14-18, 2000. 4, 155.
- Bart, J.; Kolkman, A. J.; Oosthoek-de Vries, A. J.; Koch, K.; Nieuwland, P. J.; Janssen, H. J. W. G.; van Bentum, J. P. J. M.; Ampt, K. A. M.; Rutjes, F. P. J. T.; Wijmenga, S. S.; Gardeniers, H. J. G. E.; Kentgens, A. P. M. Journal of the American Chemical Society 2009, 131, 5014.
- Martha, C. T.; Elders, N.; Krabbe, J. G.; Kool, J.; Niessen, W. M. A.; Orru, R. V. A.; Irth, H. Analytical Chemistry 2008, 80, 7121.
- Leung, S.; Winkle, R.; Wootton, R.; deMello, A. Analyst. 2005, 130, 46.
- Welch, C.; Gong, X.; Cuff, J.; Dolman, S.; Nyrop, J.; Lin, F.; Rogers, H. Organic Process Research & Development 2009, 13, 1022.
- Hubner, S.; Bentrup, U.; Budde, U.; Lovis, K.; Dietrich, T.; Freitag, A.; Kupper, L.; Jahnisch, K. Organic Process Research & Development 2009, 13, 952.
- Atherton, J.; Blacker, J.; Crampton, M. R.; Grosjean, C. Org Biomol Chem. 2004, 2, 2567.
- Carter, C. F.; Lange, H.; Ley, S. V.; Baxendale, I. R.; Wittkamp, B.; Goode, J. G.; Gaunt, N. L. Organic Process Research Development. 2010, 14, 393.
Author Biographies
Frederick T. Mattrey received his B.S. degree in chemistry from DeSales University of 2001. He has been with Merck for eight years and currently holds the title of Project Analytical Chemist. He has worked as an analytical chemist in drug product and drug substance pharmaceutical development. His scientific interests are in the areas of chromatography and process analytical technologies.
Raised in Edmonton, Alberta,Sarah Dolman received an Hon. B.Sc. from the University of Toronto in 1999, where she studied Mathematics & Chemistry. She then pursued graduate studies at the Massachusetts Institute of Technology under the guidance of Professor Richard Schrock, obtaining her Ph.D. in Organic Chemistry in 2004. Sarah joined the Process Research Department at Merck & Co., Inc. as a Senior Research Chemist in Montreal, Quebec. In 2008, Sarah moved to the Rahway, NJ site of Merck & Co., Inc. where she is currently a Research Fellow.
Jason Nyrop received his B.S. degree in chemical engineering from Bucknell University in 2007, and worked in the chemical process development and scale up of active pharmaceutical ingredients using both in-silico and laboratory techniques until 2010. He is currently in the MBA program at the Johnson Graduate School of Management at Cornell University.+
Peter J. Skrdlareceived his Ph.D. degree in chemistry from the University of Arizona in 1999. He has been with Merck for over twelve years and currently holds the title of Senior Investigator. He has conducted scientific research in various fields dealing with the application of analytical chemistry to the development of active pharmaceutical compounds. He has authored/co-authored over 40 papers in peer-reviewed journals, to-date.