Manufacturers of pharmaceuticals and other products serving consumer health are faced with an increasingly fragmented, global supply chain of raw materials. At the same time, in the aftermath of the highly publicized tainted pet food and heparin scares of the last years, consumers are growing more concerned about the safety of the products they purchase and consume. As a result, the demand for cost-effective, accurate ways for manufacturers to verify the identity of incoming raw materials is growing.
Along with the resurgent Raman technology, mid-infrared (M-IR) and near-infrared (NIR) spectroscopic techniques form a trident of vibrational spectroscopy tools that are well suited for raw material identification and verification. In vibrational spectroscopy, vibrations at the molecular level produce a unique spectral fingerprint for each compound and enable accurate and conclusive differentiation.
M-IR spectroscopy, configured as Fourier transform infrared (FTIR) spectroscopy, has for the last 60 years been used extensively for material identification and authentication.1 When infrared light is passed through a compound, some wavelengths of the light may be absorbed, while others merely pass through the sample unaffected. The frequencies of the light that are absorbed correspond to the vibrational frequencies of the chemical bonds within the sample’s molecules. The absorption frequencies in FTIR (expressed as wave numbers [cm-1]) range from 400 cm-1 to 4,000 cm-1.
Because absorptions correspond to the vibrational frequencies of different bonds within the molecule, they can be used to identify a particular functional group such as C-O, C=O, O-H, or N-H, because the vibrational frequencies of these bonds in the mid-infrared range are well known. Because each bond may have several vibrational frequencies and a molecule may incorporate many different bonds, the infrared spectrum of a material may be complex. Fortunately, it is not necessary to identify every absorption frequency, because it is possible to match an infrared spectrum against others, providing they were all obtained under the same conditions. The fingerprint region (900 cm-1 to 2,000 cm-1) is especially interesting because of its greatest number of absorptions.
FTIR absorptions are strong, a characteristic that offers benefits but also poses challenges. Because glass absorbs FTIR wavelengths strongly, it cannot be used to contain materials during FTIR measurements. Materials must be measured directly, and the use of transmission sampling makes it necessary to dilute the sample by preparing a KBr pellet or Nujol (oil) mull. More recently, attenuated total reflectance (ATR) sampling techniques have been improved. They are now popular for FTIR because they remove the need for tedious sample preparation. Despite improvements, current sampling techniques introduce uncertainties into the analysis.
For ATR (the most common sampling method today), the sample of interest must be in direct contact with the ATR crystal, and only a few microns of sample are actually interrogated. This procedure raises significant issues for analyzing heterogeneous materials. ATR accessories can be difficult to use and are breakable and somewhat operator dependent. They also introduce characteristic distortions to the FTIR spectrum, making it necessary to use a software algorithm before searching against libraries generally created from transmission measurements. This correction procedure introduces additional uncertainties into the interpretation.
Traditional FTIR instruments are large, have moving parts, and generally require operation via a computer, making them impractical for use outside the laboratory. While fiber optics for the mid-infrared spectral region exist, they are delicate, have extremely poor transmission (one meter maximum), and are very expensive and are therefore not routinely used. Despite the practical difficulties, FTIR is still extremely popular for all chemical identification needs.
Because FTIR has exquisite molecular selectivity, it is simple to interpret, making it useful in sample identification. Compared to NIR and Raman, there are many more reference texts, correlation charts, and electronic libraries containing spectra of compounds. Automatic search methodologies can be employed to compare the spectrum of an unknown sample against different spectra in a library, including simple peak matching and discrimination analysis (Euclidean distance). Analysts must use caution when interpreting the results, however. Searching techniques can easily generate spurious results, because the closest known match may actually be significantly different from the unknown material; the search simply reports the best-known match from the spectral library, even though the unknown may be from a very different class of materials.
NIR characterizes the material based on its absorption in the approximately 4,000 to 12,500 cm-1 wavelength region, corresponding to vibrational overtone (harmonic) and combination modes that are much weaker than the fundamental modes measured in the mid-infrared. Each fundamental absorption in the mid-infrared region has several corresponding overtone and combination bands, many of which overlap and are broadened. Because of this, NIR spectra are characterized by much broader features than FTIR spectra, and it is often impossible to make direct bond assignments to particular frequencies.
The bands observed in NIR arise predominantly from stretching of O-H, C-H, and N-H bonds. Because these substructures are common across organic molecules, the differences between NIR spectra of different compounds are often subtle, resulting in a much lower inherent molecular selectivity than FTIR. The sampling method is generally diffuse reflectance, making the physical nature of the sample extremely important in NIR analyses; NIR spectra contain both chemical information like peak shape and position and physical state information such as baseline slope.
The weakness of NIR bands can often be used as an advantage because there is no requirement to specially prepare a sample as there is in FTIR. Additionally, samples can be analyzed through some translucent packaging. This characteristic has led to interest in the use of NIR to identify raw materials in pharmaceutical quality control. The most commonly used sampling mode for solids with NIR spectroscopy is diffuse reflectance. Trigger-operated fiber probes—usually a multi-fiber bundle for solids—and integrating spheres made of polytetrafluoroethylene or coated with gold, are both popular diffuse-reflectance sampling configurations.
Probes and spheres pose challenges for solids interrogation, however. Probes can produce operator-dependent results because the pressure on a powder sample causes baseline movement in the NIR spectrum. The baseline of the spectrum also rises and falls if there is any movement in the fibers during sample collection. Integrating spheres necessitate collecting and placing a sample directly onto the window of the accessory—or putting it into a suitable container first, albeit an inexpensive one, negating one of NIR’s benefits and exposing the operator to the material.
NIR bands are generally broad and ill-resolved, lacking the specificity of FTIR, which is so prized for identification of chemical spectra. As such, peak picking algorithms are not used to identify a material. Chemometric techniques such as multivariate discriminant analysis are required for sample qualification with NIR data. There are many different methodologies available, but they can be thought of as methods to reduce the dimensionality of the data and model the inter- and intra-class dispersion of the data, thereby classifying new samples that are in accordance with historical training/calibration data.
Extensive and comprehensive library construction is critical to accurate interpretation of NIR spectra. Generally, an NIR library can be produced by training the system to recognize what is representative of a sample. Because NIR spectra are affected by moisture content, as well as particle size and density, however, compiling a library for any method requires great care to ensure that representative materials are used and a full validation procedure takes place. Otherwise, the system may fail quite soon after commissioning, as an acceptable sample may display some subtle physical change from the samples used to prepare the library. The new sample may then have to be included as a representative sample and the library updated. The cost of preparing and maintaining an NIR library is often recognized as one of the most significant costs of operation.
Unlike FTIR and NIR, which are absorption techniques, Raman is a scattering technique. The sample is illuminated with an intense single wavelength light source. Most of the light scatters from the sample without any change in wavelength; this is the elastic scatter or Rayleigh-scattered light. A very small proportion of the light is inelastically (Raman) scattered. The frequency of the Raman-scattered light has shifted from the original wavelength, with the difference in frequency corresponding to the vibration frequency of bonds within the molecules of the sample. Typically, one photon in 106 to 108 is Raman scattered; the rest are Rayleigh scattered or absorbed.
The mechanism causing Raman scatter is different from that found in FTIR or NIR absorption. FTIR and NIR require a change in dipole movement in the vibrating bond for absorption to occur; a change in polarizability in the vibrating bond is needed for Raman scatter to occur. A molecule showing the change in polarizability required to make it Raman active may or may not show the FTIR/NIR activity (i.e., a change in the dipole).
Many absorptions that are weak in FTIR are strong in Raman; mid-IR and Raman are therefore said to be complementary. Symmetric vibrations give rise to intense Raman lines; nonsymmetric ones are often weak and sometimes unobservable. In Raman, as in FTIR, fundamental vibrational modes are interrogated, resulting in outstanding molecular selectivity with little dependence on physical properties such as particle size.
One of the greatest strengths of Raman over other technologies is the ease of sampling it affords. Glass, plastic film, and water are very weak Raman scatterers, enabling sampling through containers and packaging that would not be possible in FTIR; for example, Raman technology can sample through an ultraviolet cuvette, NMR tube, capillary tube, vial, plastic bag, or bottle. Raman sampling is non-contact, non-destructive, and can be made through the double-bagged internal containment in drums. Raman technology can be used to measure aqueous solutions, interrogating the dissolved analytes while analytically ignoring the water. The sensitivity of contemporary Raman instrumentation is such that acquisition times of seconds, comparable to FTIR and NIR, are now the norm.
A major benefit of Raman over NIR is its insensitivity to the physical form of the sample. This advantage results in a much simpler approach to data interpretation. Peak picking routines can be used with Raman libraries because the peaks are distinct and sharp and do not move unless the chemical moiety is affected. Raman measurements may be limited by the phenomenon of fluorescence; fortunately, this effect is not frequently observed with today’s 785 nm or longer Raman excitation lasers.
All three vibrational spectroscopy techniques have their place in the pharmaceutical industry. FTIR is still the most widely implemented for identity testing. Its sharp peaks make it ideal for qualitative analysis, and its output is most often used with a peak picking routine or full-spectrum search to confirm the identity of incoming raw materials.
The sampling and current size requirements of FTIR limit it primarily to lab use, however. In addition, water’s extremely strong and broad absorbance in FTIR can pose a problem. Significant amounts of water in a solid sample will likely prevent the measurement of some useful information. The costs associated with lab-based FTIR analysis, including sample collection, health and safety issues, quarantine requirements, and training of highly skilled lab personnel, have encouraged the investigation of both NIR and Raman as alternative and complementary techniques.
Having gained popularity over the past 20 years mostly due to its sampling convenience through some container materials, NIR is present in a variety of applications such as blend uniformity, raw material verification, moisture determination, and particle sizing.2 For raw material verification, the sensitivity of NIR to physical and chemical attributes makes it superior to FTIR because FTIR sample preparation destroys the original particle size distribution, and sensitivity to water prevents measurement of high moisture content samples.
This combination of physical and chemical information also renders simple and direct interpretation of the spectrum an arduous task at best, however, leaving NIR at a disadvantage when compared to Raman.3 In a raw material identification application, a quality control group faces many suppliers and products with vastly different physical characteristics, a situation that leads to lengthy and continual calibration set maintenance, making validation difficult and expensive.4 Subtle differences in particle size and other sample characteristics have to be captured, leading to the requirement of multiple calibration reference samples.
Raman instrumentation, on the other hand, is sample-container independent with easily interpreted results and small calibration sets. Raman combines the selectivity of FTIR and ease of sampling of NIR to make an ideal raw material inspection tool that will likely see widespread implementation in the pharmaceutical industry. Because a Raman spectrum is largely unaffected by particle size and moisture, continual library maintenance is not required.
In the past, Raman was not as popular as FTIR and was relegated to the research and development lab. Laser sources were large, unreliable, expensive, and difficult to maintain, and sensitivity was poor, requiring many hours for each measurement. The development of FT-Raman in the 1980s and 1990s improved matters significantly, and the development of high-performance optical blocking filters and charge-coupled device detectors in the 1990s was another significant improvement.
But the instruments that emerged from these advances were still large, costly, and suited only to controlled laboratory environments. Recently, technological advances have enabled the development of compact, rugged, self-contained Raman instruments that can be reliably used in harsh environments. Miniaturization and improved speed of analysis, for instance, have allowed use in warehouse and loading dock locations.
One instrument is the TruScan from Ahura Scientific. Handheld Raman instruments such as the TruScan feature rugged hardware designed to withstand field use, an easy-to-use interface, and fast, cost-effective validation and analysis. n
Bradley is the director of business development and Prulliere is TruScan product manager at Ahura Scientific Inc. Reach Bradley at (978) 642-2563 or email@example.com; reach Prulliere at (978) 342-2536 or fprulliere @ahurascientific.com.
1. Ryan JA, Compton SV, Brooks MA, et al. Rapid verification of identity and content of drug
formulations using midinfrared spectroscopy.
J Pharm Biomed Anal. 1991;9(4):303-310.
2. Blanco M, Coello J, Iturriaga H, et al. Near-infrared spectroscopy in the pharmaceutical industry. Analyst. 1998;123(8):135R-150R.
3. Ulmschneider M, Wunenburger A, Pénigault E. Using near-infrared spectroscopy for the noninvasive identification of five pharmaceutical active substances in sealed vials. Analusis. 1999;27(10):854-856.
4. Gemperline PJ, Webber LD, Cox FO. Raw materials testing using soft independent modeling of class analogy analysis of near-infrared reflectance spectra. Anal Chem. 1989;61 (2):138-144.