NMR Reaction-Monitoring as a Process Analytical Technique

The data that can be mined from the nuclear magnetic resonance (NMR) analysis of compounds in solution provides chemists with access to structural information as well as to the quantitative information that is inherent to this spectroscopic technique. These characteristics are of significant utility when applied to process analytical technology (PAT), particularly in the development of robust synthetic processes (1).

Traditionally, one associates solution NMR with structural identification of isolated compounds, dissolved in the appropriate deuterated solvent in an NMR tube, which is then placed in the magnet, and the relevant processes (i.e., locking, shimming, data acquisition, and processing) produce the NMR spectrum for the compound of interest. Although this method provides a wealth of information, many obstacles hinder its application to a PAT platform. The most significant obstacle is the prohibitive cost of conducting reactions, even on small scale, due to the requirement of deuterated solvents for signal lock. Even the use of partially deuterated solvent mixtures (~10% v/v deuterated in protonated solvent) can be costly. Manual sampling can prove time-consuming and alter the reaction mixture due to sample isolation. The advent of solvent suppression techniques, coupled with the advances in NMR flow-cell technology originally applied in the area of liquid chromatography (LC)–NMR and high throughput NMR spectroscopy, circumvented these problems and has led to the adoption of NMR as an online reaction-monitoring technique.

NMR analysis of reaction mixtures has advantages over traditional reaction-monitoring techniques. Apart from the aforementioned quantitation and structural elucidation, NMR is also highly discriminatory in reaction media containing multiple species in solution, including structural isomers and diastereomers. This selectivity can be enhanced using additional nuclei, such as ¹⁹F or ³¹P as reaction-monitoring handles, if possessed by the analyte species under investigation. Because this technique is also noninvasive, reactive intermediates with sufficient lifetimes can be tracked and identified, leading to a much greater understanding of the mechanistic details of a particular reaction under investigation without the need for expensive and time-consuming isotope labeling experiments. This means that the process can be investigated as is, generating a true picture of what is occurring during the course of the reaction. In addition to reaction-mixture structure elucidation, kinetic information can easily be extracted from the data furnished by NMR reaction monitoring.

Several financial and physical barriers may prevent broad implementation of this technique as a method of analysis in manufacturing processes. It is unlikely that each reactor on a manufacturing site could be coupled to a high field NMR spectrometer for reaction monitoring. There are much more cost-effective and mobile PAT alternatives available. The development of low-field portable magnets has helped to overcome the logistical issues associated with fixed high field superconducting magnets (2). Although these instruments are unlikely to provide detailed structural information on species in the reaction medium being analyzed, the data generated is nonetheless quantitative and can easily be used to monitor the progress of reactions on large scale. Alternatively, the knowledge gained via NMR reaction monitoring during the development phase of a process can readily be transferred to a more portable PAT tool (e.g., infrared [IR], Raman, or LC), which can be implemented at the manufacturing stage, or as part of a QbD strategy.

NMR reaction-monitoring systems

The use of NMR as a reaction-monitoring tool has increased in recent years (3, 4). The basic process involves removing a flowing stream of the reaction mixture from the reactor vessel, which is then pumped to the magnet using tubing. There, it enters a flow cell in the rf coil region of the probe before being returned to the reaction vessel (see Figure 1). The volume removed from the reactor at any one time is typically less than 5 mL; h flow rates depend on the experimental setup, but should be slow enough to give sufficient residence time in the high magnetic field region to assure adequate polarization and minimize the T₁ bias to peak areas (5).

Figure 1: Basic NMR reaction monitoring setup.

Many of the reports on NMR reaction monitoring have employed the use of flow cells that are fixed inside the probe, but modified NMR tubes have been used as a low cost alternative (6, 7). The replacement of these flow tubes is considered easier than replacing flow cell probes, which have to be returned to the vendor for repairs and can result in instrument downtime if additional flow probes are not available. Flow tubes offer the added advantage that any spectrometer with a conventional probe may be readily converted to be used for NMR reaction monitoring, thereby allowing for the analysis of a wide range of NMR active nuclei and temperatures.

Flow NMR can be used to investigate batch reactions. The technique also lends itself to monitoring flow and continuous reactions. The output from a packed bed or thermal tube reactor can be simply diverted to the flow NMR for in-line analysis of the product stream.

Data processing

Advances in instrumentation facilitate the generation of large amounts of data and the potential to uncover vast amounts of information. Processing of multiple individual spectra (typically > 100) proved time-consuming and labor-intensive; therefore an efficient method of processing and mining the data was needed if this method of reaction monitoring was to be viable. The development of automated processing software has expedited this procedure, allowing hundreds of 1-D NMR spectra to be processed and useful information extracted in a matter of minutes (8). Peak and spectral alignment algorithms provide the ability to accurately integrate a particular resonance of interest whose chemical shift may migrate across the spectrum during the course of a reaction. Exploiting the fact that under the correct experimental conditions the integral for a particular signal can be directly correlated to the concentration of that species, a graphical representation of concentration versus time is generated, thereby providing an accurate picture of the reaction's progress (see Figure 2). The data generated may then be transferred to various kinetic and modeling software packages to produce kinetic results for chemists and engineers. Offline, stopped flow, or on-flow 2-D NMR experiments can be conducted to determine the nature of reactive intermediates formed during the course of the reaction.

Figure 2: Data processing of multiple NMR spectra, showing the hydrolysis of acetic anhydride as a simple model reaction.

Application to the pharmaceutical industry

The value of NMR reaction monitoring has been increasingly documented in the literature and is being implemented in the development of multiple synthetic steps by groups at Pfizer, Lilly, Roche, and AstraZeneca, to name but a few (4, 9–11). A recent report on the optimization of the synthesis of an imidazole derivative demonstrated the value of flow NMR reaction monitoring for understanding the formation of an undesired byproduct (11). The process proceeded through an intermediate that could not be detected by other analytical techniques because of the instability of the intermediate. However, examination of the process by flow NMR allowed the detection and characterization of this intermediate as well as elucidation of the mechanism of the reaction. As a result, a method of producing the required imidazole as the sole product was found.

Conclusion

In summary, NMR technology is a valuable addition to the arsenal of PAT tools currently available in the pharmaceutical industry. The detailed picture of a process that is gained from a single experiment provides valuable information and a deeper understanding for scientists across a broad range of disciplines. This knowledge enables a robust process to emerge from the process research and development stage of drug development.

David A. Foley is a postdoctoral fellow, Mark T. Zell is a senior principal scientist, and Brian L. Marquez* is an associate research fellow, all at Pfizer Global Research and Development, Groton/New London Laboratories, Brian.Marquez@pfizer.com. Andreas Kaerner is a research advisor in analytical sciences R&D at Eli Lilly and Company.

*To whom all correspondence should be addressed.

References

1. M.A. Bernstein, Spectrosc. Eur. 19, 11–13 (2007).

2. S.K. Kuster, E. Danieli, B. Blumich, and F. Casanova, Phys. Chem. Chem. Phys. 13, 13172–13176 (2011).

3. M. Maiwald, H. H. Fischer, Y.K. Kim, K. Albert, and H. Hasse, Jrnl. Magn. Reson. 166 (2) 135–146 (2004).

4. M.A. Bernstein, M. Stefinovic, and C.J. Sleigh, Magn. Reson. Chem. 45 (7) 564–571 (2007).

5. J. Sudmeier, U.L. Gunther, K. Albert, and W. Bachovchin, Jrnl. Mag. Res. A, 118 (2) 145–156 (1996).

6. D.B. Green, J. Lane, and R.M. Wing, Appl. Spectrosc. 41 (5) 847–851 (1987).

7. M. Khajeh, M. A. Bernstein, and G. A. Morris, Magn. Reson. Chem. 48 (7) 516–522 (2010).

8. Mestrelab Research, http://mestrelab.com.

9. M. Zell et al., presentation at SMASH—Small Molecule NMR Conference (Portland, 2010).

10. D.A. Foley et al., J. Org. Chem. (2011) in press.

11. A.C. Barrios Sosa et al., Org. Process Res. Dev. 15 (4) 444–454 (2011).