The Ins and Outs of On-line Process Control with Raman and FT-IR Spectroscopy

Using Raman and FT-IR spectroscopy for on-line monitoring of manufacturing processes offers advantages such as improved quality control, nondestructive analysis, and reduced costs. Jim Rydzak has more than 20

years of experience leading teams in applying on-line process control, in both the pharmaceutical and consumer goods industries. He recently talked to Spectroscopy about that work, including what they achieved and how they overcame challenges.

You have applied both FT-IR and Raman spectroscopy for pharmaceutical process monitoring. How do you decide which technique to apply to a given application?

I subscribe to the philosophy of using the right tool for the job. In pharmaceutical and consumer product development, I have worked on processes in which the starting materials and the desired products were known. This knowledge enabled me to use the strengths of both FT-IR and Raman spectroscopy to their full advantage. For Raman, its strength is looking at double and triple bonds such as a carbon–carbon triple bond, which was key in one of the steps in work I have done on an oligonucleotide process. Other opportunities for applying Raman have included looking at the structure change in the backbone carbon chain of a compound and monitoring crystallization for conversion to optimize a process parameter (to ensure the proper polymorph change), as well as some cases that used water as the sole solvent.

Another reason to choose Raman is often overlooked. There are cases where you are looking predominately at the suspension in a slurry such as crystallization. Since Raman is a scattering technique, it is more sensitive to the suspended particles. Conversely, FT-IR is better at looking at reactions that produce colored solutions or products that can cause fluorescent interference in Raman where large, broad fluorescence bands can dwarf the bands of interest. Even clear, particle-free solutions many times can be more simply handled with FT-IR, where Raman often requires a longer focus probe to get good quality solution spectra. FT-IR is also somewhat simpler to set up and is a less expensive tool, so a company could deploy more FT-IR instruments for cases where there is not a clear Raman advantage.

Are the two techniques equally amenable to real-time monitoring?

Yes, the two techniques are both equally applicable to real-time monitoring with the caveats that I mentioned, though Raman has the advantage when the application is in an industrial setting. In the plant, the instrument can be situated remotely and thus a less industrially hardened and less expensive Raman instrument can be used. The fibers used to carry the signal from the probe to the instrument do not attenuate the signal as much as the infrared fibers, are less expensive, and can be tens of meters in length, enabling remote location of the Raman instrument.

Do you generally need to apply multiple techniques to a single manufacturing process?

For every process, there is a risk assessment performed and critical process parameters that could cause process failure are determined. For example, a critical process parameter could be the production of an impurity that results from running the reaction too long or other process faults in temperature, pressure, mixing, or human error. Another common critical process parameter is the crystallization stage, where it is essential to produce the correct polymorphic crystal, as often there are bioavailability differences in the different forms of the crystal. So, to answer the question, it is very feasible that more than one technique could be deployed to address the critical process parameters, especially as most pharmaceutical products involve multistep syntheses and isolation of the intermediate products.

You have applied both process FT-IR and Raman spectroscopy to the synthesis of oligonucleotides (1). What particular challenges did this present?

A combination of FT-IR and Raman were used in that feasibility study. There were some colored reactions that caused fluorescence, where FT-IR was favored. Some of the steps were also moisture sensitive, so the ability to detect water was a benefit for the FT-IR system. There was one key step in the preparation of a starting material that necessitated the breakdown of a carbon–carbon triple bond to form the catalytical form of a reactant that was key to the process, and there, the Raman technique had the clear advantage.

In that work, what types of potential errors did you want to be able to detect?

Most of the starting materials for this process were extremely expensive. This synthesis is done using an automated, continuous, flow reactor that uses a cyclic four-step process to add each oligomer to the chain. For a 10-unit oligonucleotide chain, the cyclic process was repeated 10 times. So, if there was an upset in any of the process steps it would be better to divert the compromised product than contaminate the good product already produced. The monitoring could also stop the process as soon as possible, during the repeating steps, rather than use up the expensive reactants.

There were five potential risks of error that we were looking to address. The flow reactor has over a dozen solvent and reactant inputs. The first risk was whether all the reactants were connected to the correct port, as introducing the reactants in proper sequence was essential to the synthesis. The next risk was whether the concentrations were correct for correct stoichiometry. A third risk was that the carbon–carbon triple bond was cleaved, activating one of the reactants. There was also a risk of moisture contamination, which could deactivate some of the reagents, as a result of seasonal fluctuations in humidity and potential raw material contamination. The final risk was being able to detect process upset conditions such as a pump or seal starting to fail and causing less-than-stoichiometric amounts of reactants being delivered.

Was the system able to detect them?

The first risk was easily managed by developing a discriminate chemometric model to accurately identify the correct reactant. A quantitative chemometric model was used to ensure that the proper concentration of each reagent was being delivered. Raman spectroscopy was able to monitor the absorbance of the carbon–carbon triple bond and trend its disappearance, indicating that the cleavage of the triple bond had happened and providing the assurance that the reactant was activated. FT-IR spectroscopy was surprisingly sensitive detecting moisture in the various solvents (as low as 50 ppm water in acetonitrile) and flow stream. Further work was needed to determine if the detection limit of water was adequate to prevent impurities. The last risk, pump or valve failure, was mitigated using multivariate statistical process control to monitor various inputs, including spectral data, that produce a composite model of reactants and processing conditions to provide a steady state trend. When the trend dipped below the 99% confidence level, an upset situation was detected, and by clicking on the active trend, a chart of contributing variables and their acceptable range was displayed. The use of spectral data in this contribution plot enabled interpretation of absorbance bands from the contribution plot to provide indications of the effect of the physical upset condition on the chemistry to be deduced. This information could help determine which seal or pump in the process was failing.

For what other types of pharmaceutical manufacturing process steps have you applied Raman spectroscopy, or seen examples of others applying it?

Raman spectroscopy is a key tool in monitoring the crystallization step in a pharmaceutical manufacturing process, especially if the low-frequency Raman region of the spectrum below 300 cm^-1 down to 5 cm^-1 near the laser line is used to expose the crystal lattice vibrations of the material.

Raman is also being used successfully in fermentation monitoring and processing in the biopharmaceutical area. FT-IR has also had success in this area.

Raman is also used for reaction monitoring, especially when monitoring reactions that have double and triple bonds forming or being broken and structure change in the backbone carbon chain of a compound.

In your work applying on-line Raman and FT-IR spectroscopy to manufacturing process control, what have typically been the biggest challenges, and how did you overcome them?

Often the biggest challenge is gaining acceptance of the new on-line technique over traditional lab analyses. In such cases, rigorous justifications need to be prepared to show the return on investment in addition to the enhanced quality assurance that the on-line monitoring provides.

Other challenges involve placement of the probe into a vessel that may not have a port in the best position to monitor a critical process parameter. These challenges have often been resolved through the use of dip tubes that can protect the probe and allow it to be placed where it needs to be in the vessel.

Cleaning of the probe in process and providing a clean background scan have also caused problems. These challenges have been overcome by employing an interface that can retract the probe, sealing it off from the vessel, to a position where it can be manually or automatically cleaned and dried. A fresh background or dark scan can then be made to minimize instrument drift and ensure the integrity of the method.

Have you been involved in cases where you tried to apply process Raman spectroscopy and determined that it was not suitable?

In my experience, where the complete process development and optimization and pilot plant scale-up is done before plant deployment, a person has multiple opportunities to determine whether on-line Raman spectroscopy will work. I have had one installation where Raman was used in development for monitoring a monohydrate to anhydrate crystal formation step where we chose to use a different technique that was easier to maintain and implement in the manufacturing plant than the Raman approach.

Reference

• J.W. Rydzak. D.E. White, C.Y. Airiau, J.T. Sterbenz, B.D. York, D.J. Clancy, and Q. Dai, Org. Process Res. Dev.19(1), 203–214 9 (Process Analytical Technologies 14 special issue), dx.doi.org/10.1021/op500035j

James (Jim) Rydzak is currently consulting under the company name Specere Consulting, providing expertise in process analytical technology, vibrational spectroscopy, reviewing instrument and software development, writing application notes, and providing market research information. His experience includes over 16 years of spectroscopy experience in the pharmaceutical industry with GlaxoSmithKline and 16 years in the consumer product sector with Colgate-Palmolive, including starting on-line process analytical groups in both organizations. He also has experience in in the polymer area with El Paso Chemical and Polyolefins. He can be contacted through LinkedIn.