Raman spectroscopy is a technique used for observing vibrational modes in a molecule. Raman spectra reflect scattering abilities of chemical bonds in a molecule and can be used for chemical identification and quantitation of substances. Raman spectra are obtained by exciting the sample with a laser. The inelastic interaction between the laser beam and the sample leads to a small change in the energy of the excitation beam which can be captured as Raman signal. Position and strength of Raman bands are dependent on the chemical structure of the excited molecules. Raman spectrometers and microscopes are increasingly used in various industries and academia for fundamental and applied research. The use of Raman spectroscopy for pharmaceutical applications is steadily increasing.

Raman spectroscopy in pharmaceutical industry is used for:

  • Analysis of active ingredient and excipients in variety of solid dosage formulations, qualitative and quantitative, via bench-top or handheld Raman spectrometers or Raman microscopes.
  • Online, real-time monitoring of chemical reactions, or tablet coating.
  • Identification and quantification of polymorphs, in pure drug substances or in formulations, due to high sensitivity of Raman spectra to polymorphic variations.
  • Chemical imaging of API or excipients in tablets, blends, beads, granules.
  • Analysis and imaging of biomedical samples (skin, cell, tissues).

Of particular interest is use of Raman spectroscopy during development of the formulations as APIs are normally much stronger Raman scatterers than the excipients, which leads to relatively easy detection of API Raman signal.

In principle, there are three groups of Raman spectrometers: Dispersive, Fourier transform (FT) and Transmission. The first two are based on reflection / backscattering collection of Raman signal and thus primarily provide information about the surface of the sample while in the transmission Raman configuration the signal is generated across the sample.

Transmission Raman spectroscopy has only recently been commercialized to a significant extent in particular owing to TRS100 instruments manufactured by Cobalt (Oxford, UK). Strong Raman spectra can be relatively easily and quickly obtained on a TRS100 despite the laser beam passing through the sample.

Transmission Raman spectroscopy can be used for:

  • Content uniformity analysis for all solid formulations.
  • Polymorph analysis/detection/quantification.
  • Rapid sample screening, e.g., polymorphs, counterfeits, process variation monitoring.

Benefits:

  • Speed of acquisition and quality of spectra.
  • Various samples easily accommodated – powders, tablets, capsules.
  • Minimal sample preparation.
  • Very high throughput for a Raman spectroscopy technique.
  • Ideally suited for method development, in particular for process analytical technology (PAT) applications.

Case Example: Transmission Raman Spectroscopy Quantitative Analysis

  • A simple binary mixture of mannitol in glucose is analyzed.
  • Mannitol concentration varying from 0.5 to 5% wt/wt, 6 mixtures.
  • Focus on an isolated mannitol band.
  • Quantitative analysis of mannitol.
  • Compare the results with those obtained by Fourier Transform (FT) Raman spectrometer.
  • The model is intentionally simple to allow for very straightforward assessment of the technique.
  • Also, the model realistically mimics cases often seen in pharmaceutical industry of determining low concentration of an undesired polymorph in an API in drug substance purity analyses.

Goals:

  • Demonstrate the quality of spectra, efficiency and easiness of acquisition.
  • Demonstrate accuracy of multivariate calibration based on transmission Raman spectra.
  • Demonstrate the superiority of transmission Raman versus FT Raman measurements.

Sample preparation:

  • Sample preparation amounts to merely filling zip-lock (plastic) bags with mixtures.
  • Multiple scans of the sample can be obtained easily and in an automated manner. Sampling of the mixtures in transmission and with multiple spectra per sample ensures an excellent interrogation of the sample, much better than for FT Raman or IR technologies , and with minimal laboratory effort.

 

 

 

 

Figure 1. (top): Tray for holding powder bags. A Raman spectrum is obtainable from each shown hole; (middle): Ziploc bags filled with powder sample ( ~1g of powder). Nearly the entire amount of the shown powder is sampled; (bottom): Tray with powder bags, ready for acquisition.

Transmission Raman spectra (baseline corrected and normalized):

 

Figure 2. Baseline corrected transmission Raman spectra.

Experimental conditions – 0.1W laser power at 830nm; 2s acquisition; 4 repeats; spectra obtained from all four holes in the tray (four transmission spectra per mixture); 4mm laser beam; 1 cm-1 resolution.  The entire set of 24 spectra is obtained for about 5min in a single, unattended run! Very high signal-to-noise ratio.

Figure 3.  Normalized transmission Raman spectra.

Significant variation in intensities (common in Raman spectroscopy) is removed through unit area normalization. Standard deviation (StDev) of noise 6.7*10-5 in no-signal regions. The normalized spectra are imported into stand-alone chemometrics package, Solo, which is linked with the TRS100 software for PLS regression.

PLS Regression Results:

 

Figure 4. Top row: Hotelling T2 vs. Q residuals, and Leverage vs. Studentized Residuals; Bottom Row: True Mannitol vs. Predicted Mannitol, and PC1 vs. PC2 score bi-plot. True vs. Predicted plot matters most here.

An excellent model − only 2 factors in the model with leave-one-out cross validation error of 0.25% wt/wt for the range of concentrations of 0.5 to 5% wt/wt. Outlier criteria (Hotelling T2, Q residuals, score bi-plot) pass all samples.

FT Raman spectra:

Figure 5. Baseline corrected FT Raman spectra obtained from the same samples as above.

Two spectra per sample obtained from the powder in NMR tube – much smaller amount of sample (10-15mg), two spectra acquired in order to reduce inhomogeneity risk. Acquisition details: 64 scans (2 min per spectrum), 0.5W laser power at 1064nm, 2 cm-1 resolution, 50µm beam size. StDev of noise 7.5*10-4, x10 higher than StDev of noise in transmission.

FT and transmission Raman normalized spectra:

Figure 6: Comparison of FT (left) and transmission (right) Raman baseline corrected and normalized spectra.

  • Both signal to noise ratio and intensity of mannitol peak of interest at 874 cm-1 look much better on the transmission instrument.
  • Substantially better quality of spectra (signal-to-noise (s/n) ratio and reproducibility) coupled with much more thorough sampling, automated and rapid acquisitions, clearly favors transmission Raman over FT Raman.

FT and transmission Raman PLS models based on normalized spectra:

Figure 7: Comparison of FT (left) and transmission (right) Raman based PLS regression models.

  • Apparently poorer s/n ratio and reproducibility of FT Raman spectra vs. transmission Raman spectra leads to no meaningful PLS model being produced from the FT Raman spectra (left, R2=0.4) as opposed to a very good model based on transmission Raman spectra (right, R2 = 0.97) with a cross-validation error of only 0.26% wt/wt.

Conclusions:

  • A large number of high-quality transmission Raman spectra was acquired from a series of powder mixtures in a highly automated manner, with minimal labor, and quite rapidly.
  • High signal to noise ratio of the spectra allows for developing an accurate multivariate model for the component of interest with the leave-one-out cross validation error of 0.26% wt/wt for the calibration range of 0.5 to 5% wt/wt .
  • These experiments conclusively prove transmission Raman to be superior to FT Raman in terms of quickness, easiness and efficiency of acquisition, as well as the overall quality of spectra.
  • Owing to the better s/n ratio and reproducibility, the accuracy of the transmission Raman multivariate model is much better than that based on FT Raman spectra, which in fact fails to produce a meaningful model for the given range of concentrations in the analysed binary mixtures.