I have been suggesting for some time that the generic and contract pharmaceutical industries would benefit from the application of PAT, QbD, and, eventually, continuous manufacturing. These techniques have a proven track record and the leading technology for process monitoring/control has been Near Infrared spectroscopy. While I have been using NIRS for four decades (using it to put a roof over my head and my kids through college), the reason for its commanding lead was that it is, by its nature, difficult to interpret.

“Why,” you might ask, “is that a good thing?” Well, in short, the good news/bad news of NIRS is simply stated: the good news is that NIR “sees” everything (chemically and physically) in the sample; the bad news is that NIR sees everything in the sample. Since a diffuse refection NIR spectrum is affected by the chemistry, particle size, morphology of the ingredients, and even the location of solvents/moisture (part of the crystal of surface), complex math algorithms are needed for many applications.

The use of derivatives and numerous other math and statistical manipulations are called Chemometrics. And, for anything other than a single component analysis, which probably would be done easier with UV or fluorescence, using Beer’s law, a more complex analysis is required. Since NIRS isn’t a primary method [spoiler alert: only weight and electrodeposition are “primary” methods] and needs a previous validated compendial method to calibrate the standard samples to generate a working calibration model. The upside to NIRS is that, despite the arduous calibration steps, the application in production settings is very rapid and can be performed with no sample prep or solvent use.

Now, on to Raman. Since Raman spectra are as crisp and specific as mid-range infrared spectra it and does not “see” water, making aqueous measurements simpler, why hasn’t it been used in PAT/QbD sooner? Well, as I alluded last column, and if you are familiar with Raman, the physics of Raman makes it a difficult tool to use in situ in production. When the laser beam (monochromatic light is a necessity) strikes the sample molecule(s), the photon(s) can do three things: 1) they can scatter inelastically (the vast majority doe this) and bounce back at the incident wavelength, 2) a photon may strike a molecule as add energy, or 3) strike a molecule and take away some energy (see Figure 1).

The elastic scattering (a.k.a., Rayleigh scattering) accounts for almost all the light striking the sample, but a simple notch filter—an interference filter that blocks only a selected wavelength, mainly, the incident laser light—is used to keep the detector from being overwhelmed by the incident light. The photons that gain energy are called Stokes radiation, while the photons that lose energy are Anti-Stokes radiation. Since each molecular bond has a different energy, the resulting scattered light gives a distinct spectrum. Both sets of wavelengths may be used, but the Stokes set is normally stronger.

However, like the Hope Diamond, Raman spectroscopy comes with a curse: the source laser generates massive amounts of fluorescent radiation from the sample! Elimination of the Rayleigh light is relatively simple since it is monochromatic. Fluorescence, however, is emitted as a full spectrum, potentially at many orders of magnitude greater than the Raman spectra. The resultant mix of spectra (fluorescence plus Raman) often resembles a pile of sand more than a spectrum.

Early attempts at obviating the fluorescence included using longer and longer wavelengths of lasers. As the wavelength approaches the near-infrared (~900-1000nm), the fluorescence quickly diminishes and disappears. The bad news is the strength of the Raman radiation diminishes almost as rapidly and, since it is weak at the best of times, trace analyses are nearly impossible.

Several manufacturers suggest spatially offsetting the return light collector (fiber optic or otherwise), since fluorescent radiation tends to reabsorb in a condensed medium. This approach is not well adapted to any application other than a contact type—touching or located very, very close to, say, a tablet or capsule. This is an impractical process tool, but useable for lab work.

On large manufacturer uses a patented Sequentially Shifted Excitation, or SSE, technology, employing distributed Bragg reflector LASER-diode that shifts excitation wavelength as a function of temperature. Raman spectra shift with incident wavelength, but fluorescence spectra do not. An on-board data processing algorithm distinguishes the spectrum of elastically scattered light. While the system can produce high-quality spectra, it large and expensive to accommodate a DBR laser. It mandates frequent laser replacement due to the constant temperature cycling of the laser.

Another method is a “gated” device, where a pulsed laser beam strikes the sample, the Raman radiation enters the device, but the window opacifies before most of the fluorescence reaches the detector. Since fluorescence is generated by an electron promoted to a higher energy state and as a finite relaxion time and RAMAN is a scattering effect, Raman radiation returns to the instrument in a shorter time period—long enough to “slam the door” to the fluorescence photons. While this device works well, it is large and expensive. It can do some interesting work but needs to be stationary and does not lend itself to process analyses—it can be placed at a fermenter, of course, but is not easily moved from place to place.

One unit I like is produced by Metrohm. Their method is named XTR (for Raman eXTRaction) and is their patent-pending proprietary method for differentiating the signals from fluorescence and Raman scattering and build them into two distinct spectra. The software (MIRA XTR DS) produces a fluorescence spectrum with its corresponding intensity and shape and generates a pure spectrum of just the Raman scattered light. Figure 2 shows Raman spectra at 785nm, 1064nm, and XTR-augmented spectrum.


Most of the modified Raman instruments mentioned above may be used to generate research-grade spectra, but this level of precision is unseen in other portable (hand-held) equipment. Figure 3 shows the spectra for two subtly different fentanyl molecules—acetyl and butyryl fentanyl. This level of differentiation cannot be achieved with “regular” Raman instruments but is common for this instrument.


How does this play into a PAT narrative? PAT begins with raw material verification. Until now, NIRS has been the go-to technology for raw materials qualification. With a high-quality hand-held unit, incoming materials can be inspected at the loading dock and accepted or rejected outright, obviating sampling, labeling, transporting to the QC lab, wet-analyses, and approval or rejection, days later, while the materials sit in the quarantine area.

Another application is in the tablet coating process (see Figure 4). The Raman unit or a benchtop model thereof can be positioned to monitor the coating process. The general approach, like a NIR probe, would be to watch the spectrum of the core diminish while the spectrum of the coating material increases. Many coating solutions now contain metals: Fe, Ti, Mg, etc. So, it may be necessary to simply follow the core spectrum disappear to not the endpoint of coating.


While Raman does not “see” water, it isn’t the first choice for drying, but the specificity of the spectra versus NIR spectra, it makes a very good tool in the PAT/QbD toolbox. When properly adapted to the drying unit, the specificity of the Raman spectra may be used to “see” the hydrates—a back-door method for water determination—but it may also follow the morphology of the API. Newer, low-dose, highly reactive, poorly soluble APIs would especially benefit from Raman since it can “see” lower levels of API than NIRS. So, one may hope that Raman moves into the fray, not replacing current techniques, but complementing them.

Also, did I mention the hand-held unit is far less expensive than a benchtop, research-grade unit?

---

Emil W. Ciurczak is President of Doramaxx Consulting. He has worked in the pharmaceutical industry since 1970 for companies that include Ciba-Geigy, Sandoz, Berlex, Merck, and Purdue Pharma, where he specialized in performing method development on most types of analytical equipment.