Ian Lewis

Fats and oils are ubiquitous in natural and processed foods, providing necessary energy storage. Fat and oil content in foods also have important contributions to the shelf life, texture, compatibility with processing operations, and sensory profiles of food products. Understanding the molecular properties of fats and oils separately within a heterogeneous food matrix requires a multidisciplinary approach. Vibrational spectroscopy techniques are used throughout the food industry to gain product understanding, identify adulterated products, ensure quality, and control processes. In analyzing fats and oils in food, near-infrared (NIR) spectroscopy is an established analytical technique, and there are other growing applications of infrared (IR) and Raman spectroscopies. In particular, Raman spectroscopy is well suited to measure fats and oils because their C-H and C-C bonds are polarizable. In this article, we review the historical use of Raman spectroscopy in studying fats and oils in foods from Fourier transform (FT)–Raman spectroscopy to dispersive Raman spectroscopy. We also provide an overview of various Raman approaches to understand fat compositional heterogeneity in solid foods, identify polymorph or crystallinity, and measure fatty acid saturation. Examples in a variety of fat-containing foods demonstrate feasibility for Raman applications in the laboratory and process environments.

Lipids are a class of biological molecules that provide important energy storage, metabolism, and cell membrane structure functions. They encompass sterols, fatty acids, glycerides, phospholipids, and waxes. Saturated fatty acids are also called fats. Fats have higher melting points, are solid at room temperature, and are more stable. Unsaturated fatty acids are also called oils. Oils have lower melting points, are liquid at room temperature, and are less stable. Understanding lipids for biomedical, food and beverage, and now lipidomics applications, requires a variety of analytical tools and techniques, including chromatography, wet chemistry, molecular probes, mass spectrometry (MS), and vibrational spectroscopy (1–3). Raman spectroscopy and nonlinear Raman spectroscopy techniques are particularly well suited to measure lipids because of a high Raman cross section and high optical scattering. A book, titled 

 by Rich McCreery, starts with the theory and goes into calculating the cross sections of some simple molecules. C-H and C-C bonds are polarizable, and they comprise the majority of a lipid. As a result, it stands to reason that lipids would have a higher Raman cross section than other biological molecules, such as proteins or nucleic acids (4). This assumption was verified by calculating the cross-sections of cholesterol and proteins, where the cross section of the 1440 cm

 band is three times higher than that of proteins (5). Another reason that lipids are easily measured by Raman spectroscopy is because of optical scattering, particularly in biological tissues. The density of the scattering is considered in calculating the Raman intensity, but that calculation only considers Raman scatters. Optical scattering also occurs elastically, so that would affect light fluence and sampling volume. Fatty tissues have a higher optical scattering coefficient than muscle tissues (6,7).

Because of their importance and strong Raman scattering, lipids have been studied by Raman spectroscopy since 1936. A 1936 paper by Jogarao is the first acknowledged paper on Raman spectroscopy analysis of food oils (8). Olive, gingelli, castor, coconut, groundnut oils, cocogem, and buffalo ghee were examined by Raman spectroscopy using an exposure of nearly 48 h to collect a spectrum. Bands at 1302, 1469, 1661, and 2873 cm

 were identified and assigned to C-H, C-O-C, and C=C groups. Other early work by van Den Hende examined various saturated fatty acids and identified cis- and trans-isomers in oleic acid (articles in French) (9,10). Additional early work studying oleic acid was described in papers written in 1945 and 1951, which focused on the chemistry of lipids and contained references to research papers written in 1943 and 1950 that were on Raman spectroscopy. However, those research papers could not be obtained by the authors. An experimental challenge that may have limited the number of Raman-based studies, in addition to the long exposure times, was the ability to purify samples (9). Thus, we can also draw on conclusions from relevant work that was performed in long chain hydrocarbons (11,12), small-chain carboxylic acids (13), and other model compounds (14).

The perception that oil or fat purification was a limitation of the Raman technique resulted in few papers being published in this application again until the 1970s. In the 1970s, studies using Raman spectroscopy to examine lipids received little attention relative to other biopolymers such as proteins, amino acids, and nucleic acids (15). However, the papers that were published in this decade demonstrated an increased sophistication of the Raman technique, including the use of Fourier-transform (FT)–Raman spectroscopy, and the scientific questions that Raman spectroscopy was trying to address. Some example applications include biophysics on mixed lipid-protein systems, measurement of chain length, cis- and trans-isomer composition, lipids in natural products, and classification of lipids (16–20).

Today, dispersive Raman, FT-Raman, enhanced Raman, and nonlinear Raman techniques are used to measure lipids. In particular, dispersive and non-linear techniques are growing in basic science and industrial applications. Readers are referred to excellent reviews by Li and others on coherent anti-Stokes Raman spectroscopy and by Prince and others on stimulated Raman scattering (21,22). This paper focuses on dispersive Raman spectroscopy of oils and fats in food and beverage applications.

A review of the literature was performed in Google Scholar using the terms 

 to describe the technology, and the search terms lipids, oils, fats, and fatty acids were used to describe the target molecule family. The references in several review papers were used to retrace the history of Raman spectroscopy to the earliest papers.

For experiments performed by the authors, such as the fats and oils identification and the salmon and chocolate studies, a Raman Rxn2 (Endress+Hauser) Raman analyzer operating at 785 nm or 1000 nm was used. Raman measurements were performed with a Rxn-10 collimated probe equipped with a non-contact optic, immersion optic, or a Rxn-20 large volumetric probe (Endress+Hauser). Measurements were performed at ambient laboratory conditions of commercially available samples with no additional sample preparation. Data were preprocessed using Grams/AI spectroscopy software (Thermo Scientific) and using routines written in-house.

Industrial Applications of Dispersive Raman

More recent applications have expanded on the historical uses to now include structure elucidation, the role of lipids in health and disease, and lipidomics (23). Within food and beverage applications, there has been an increased use of portable, benchtop, and process Raman spectroscopy in edible oil measurements. We highlight a few applications in oil and fat quality, and oils and fats in foods demonstrate the utility of Raman spectroscopy in the field. They establish scientific feasibility for additional tests on polymorphism, accelerated stability, or coproducts.

Figure 1 shows the Raman spectra of canola, peanut, and avocado oils between 600 and 1775 cm

 as representative oils. Visual inspection of the band position, width, and intensity provides an initial understanding of multiple functional groups. The main bands of fats and oils are noted in Figure 1, with the band assignments referenced in Table I. A parameter of interest is saturation. The position and width of the band at ~1444 cm

 shifts to a lower Raman shift and becomes more narrow with increasing saturation. Total unsaturation can be determined quantitatively by measuring the ratio of the C=C stretch (υC=C) centered at 1661 cm

. Calculating band intensity ratios is a straightforward approach to quantifying total unsaturation. Cis-isomer content can also be determined from the Raman spectrum. The band at 1272 cm

 is attributed to in-plane =C–H deformation in an unconjugated cis double bond. The intensity of this band ratioed to the in-phase methylene twisting vibration at 1306 cm

 provided a direct cis-isomer measurement, which is illustrated by comparing the Raman spectrum of vegetable oil with that of partially hydrogenated vegetable shortening. Hydrogenation results in a preferential decrease in the cis- or trans-isomer ratio.

FIGURE 1: Raman spectra of canola, peanut, and avocado oils were collected from commercially available samples. Major bands in the oil spectra are noted and assigned.

Authentication, Adulteration, and Blending

An early application was in oil discrimination and identification of adulterated oils. Adulteration is the deliberate and undeclared substitution of low-quality, but indistinguishable using senses, ingredients to hide poor quality or increase the volume of high-value foods. Adulteration represents a significant food safety, labeling, and product quality concern. Olive oils are frequently adulterated because they are high-quality foods, and adulterated oils often are indistinguishable through physical or chemical constants such as iodine value, saponification, refractive index, or density. A 1994 review by Li-Chan summarized advances in chromatographic and spectroscopic techniques to measure chemical composition markers of adulteration including Raman spectroscopy (24). Since the time the review was published, studies have reported application success using FT-Raman spectroscopy, Raman microscopy, handheld Raman, and fiberoptic Raman for a variety of adulterants and data analysis approaches (25–29). A comprehensive three-year longitudinal study by Kwofie and others enabled an understanding of supplier variations and yearly variations for 15 varieties of oils (30). As hardware and chemometric techniques improved along with a better understanding about the field, we saw many sophisticated questions answered. Two recent papers that used Raman spectroscopy to authenticate geographical location, identify adulterants, and discriminate cultivars demonstrate how the hardware and chemometric techniques have significantly improved (31,32).

Univariate or multivariate analyses are useful for Raman-based oil adulteration measurements. Univariate analyses are straightforward and allow for chemical changes to be easily interpreted. Univariate markers of total unsaturation and cis isomer content have been proposed in the literature. Partial least squares (PLS) analysis was successful in identifying canola oil, rapeseed oil, and spent oil adulterants. The principles of measurement for adulteration can be applied toward blending because both involve the identification and quantification of a mixed product. Blending is the deliberate and declared substitution of different ingredients to improve labeling claims or sensory attributes. Some common examples of a blended product are butters blended with margarine or olive oil. Butter blended with margarine, starch, or plant-based oils is also common practice. An important aspect to blending is accurate and rapid quantification because errors can lead to product giveaway or reduce process efficiency. Similar to adulterated products, blended products generally do not change the taste of butter and detection relies upon analysis. Laboratory-based measurements can be expensive and require specially trained staff and consumables. A paper by Nedeljković and others demonstrates the Raman-based identification of margarine in butter–margarine mixtures (33). The motive behind the study was for adulteration purposes, but the study can also be viewed in the context of a product blending application.

In the study by Nedeljković and others, margarine was distinguished from butter using bands at 1268 and 1656 cm

 associated with the C=C double bond and at 973 cm

 associated with the choline group in the phospholipid. Raman-based calculations were similar to known values measured by weight. PLS regression further confirmed the very good overall performance of the model, with the coefficient plot showing large positive values at the locations of the peaks discussed above. For the calibration data, 

= 0.991 and RMSECV = 3.199. For the test data, 

 = 0.994 and RMSEP = 2.757. Principal component analysis (PCA) data showed a relationship between the second principal component and the margarine content. The loading plot confirms that the known chemical moieties of margarine (the C=C double bond and the choline group) vary together with margarine concentration because of the consistently negative peaks at 973, 1268, and 1656 cm

, the intensity of which closely follow the margarine content in the calibration samples. The PCA of model mixtures demonstrate that the three Raman peaks at 973, 1268, and 1656 cm

 can be attributed entirely to vegetable fat in margarine and those peaks can be used to quantify margarine content in butter. This important feature means that the fat does not have to be isolated from the rest of the sample as was necessary in a chromatographic method. The removal of the fat-extraction step is an improvement toward streamlining the analysis for the laboratory or process environment.

Different types of macronutrients are commonly found in foods, with fats, proteins, and carbohydrates being the main macronutrients. Measurement of these macronutrients in a heterogeneous matrix can be achieved using Raman spectroscopy. Milk powders is an example of a common food that contains fats, protein, and carbohydrates. Cheese is another dairy product that contains fats and proteins. Solid milk powders with various concentrations of fats, carbohydrates, and proteins as shown in Table II were obtained and analyzed as-received. All milk powder samples were analyzed in a laboratory using a Raman Rxn2 Hybrid analyzer (λ = 785 nm). The analyzer was equipped with a 6-mm Rxn-20 probe, which provided a 6-mm spot size. The Rxn-20 probe was positioned approximately 25 cm from the solid powder samples. Raman measurements of the milk powders were collected for 2 min directly from open containers or as loosely packed in a plastic container.

Figure 2 shows the representative Raman spectra from four different milk powders. Bands on the spectra arising from lipids (or fats), proteins, and carbohydrates were identified using known literature values (15,34). Importantly, bands from the individual components were well-resolved and enable univariate models for quantification. Although there is spectral overlap of the constituents, the noted areas in Figure 2 can be used to monitor changes in macronutrients. In the example of the 1620–1800 cm

 region, a more narrow shape of the carbonyl stretch envelope at ~1650 cm

 and the emergence of relatively narrow bands at ~1740 and 1301 cm

 indicates a preponderance of lipids in powders 3 and 4. More intense and broader bands in the 1040–1120 cm

 region is indicative of a higher carbohydrate amount in powders 3 and 4. By contrast, a broader carbonyl stretch envelope at ~1650 cm

 and a narrow ring breathing band at ~1005 cm

 indicates an observable presence of proteins in powders 1 and 2.

FIGURE 2: Milk powders of varying protein, carbohydrate, and fat concentrations were measured by Raman spectroscopy. The specificity of the Raman spectrum enabled simultaneous measurement of the three macronutrients and contributions from fats (*), protein amino acids (◊), and carbohydrates (‡) were observed.

Fish and meats are other common foods where there are different types of macronutrients, which are mostly proteins and fats. Raman spectra were collected from store-bought samples of Atlantic salmon, wild coho, and smoked salmon. Two experiments were performed. In the first experiment, laboratory tests were performed using a benchtop Raman analyzer (Endress+Hauser) operating at 1000 nm equipped with a small-area (measured area ~100 μm

) contact probe. The probe was manually placed on the surface of the fish and signal collected for 10–60 s. In the second experiment, a large volumetric noncontact probe (measured volume ~3–6 mm

) was used to assess compatibility with processing plant conditions. Spectra were presented without any preprocessing. The initial tests performed on the store-bought salmon indicated a feasibility of 1000 nm excitation for laboratory measurements. The small-area probe was able to collect the signal from each of these zones without significant interference from adjacent zones. Figure 3 shows the representative Raman spectra from Atlantic salmon, where there were visually apparent fat-rich and muscle-rich zones. The Raman spectra were also visually different. The top spectrum contained features from mostly lipids in fat, and the bottom spectrum contained the protein features. The approach of using a small-area contact probe was applicable when more precise zonal measurements were required for quality control purposes. Notably, there were no bands observed from carotenoid pigments in spectra from Atlantic salmon. By contrast, the spectra of wild coho salmon contained a signal from carotenoid pigments in addition to lipid and protein bands (data not shown).

FIGURE 3: Raman spectra showing main protein (black) and fat (red) related bands for fresh salmon. Contributions from fats (‡) and protein amino acids (*) are shown.

Mock process measurements were performed on Atlantic salmon. The large volumetric probe provides a noncontact measurement over a large volume, and the resulting spectra had contributions from the fatty and muscle portions of the tissue. Measurement times were optimized to be compatible with conveyor belt speeds. The Raman spectra collected at 1, 3, and 5 s were suitable for input into a univariate or multivariate model with good signal-to-noise (S/N) and spectral resolution. Similar to the measurements collected in dairy solids, the width of the amide I envelope at ~1660 cm

 and the presence of the ring breathing band at ~1000 cm

The primary constituents of chocolate are cocoa butter and sugar, and measurement of dark or milk chocolate is a challenge for Raman spectroscopy performed in the visible or shortwave NIR because cocoa solids strongly fluoresce (35). To mitigate fluorescence, FT-Raman and surface-enhanced Raman spectroscopy (SERS) was used to study chocolate, cocoa butter, and fermented cocoa beans (36–39). Recent studies reporting dispersive unenhanced Raman with 1064 nm excitation piqued our curiosity to use a benchtop 1000 nm Raman system for chocolate measurements (40). A 1000 nm Raman Rxn2 analyzer (laser intensity = 100 mW), equipped with an Rxn-10 probe and non-contact optic that resulted in a spot size of 150 μm, was used to collect Raman spectra from commercially available white-, milk-, and dark-chocolate samples (41). Raman spectra were compared with literature studies in cocoa butter polymorphs, lipid extractions of chocolate, and chocolate samples (35,36,42). Figure 4 shows an example dark-chocolate spectrum collected with 1000 nm excitation. Qualitative and quantitative analysis of several functional groups of cocoa butter and sucrose was performed directly from the chocolate Raman spectrum without further sample preparation. Our initial results showed that Raman spectra collected at 1000 nm demonstrated sufficient background reduction to enable observation of bands throughout the fingerprint region for cacao nibs and white-, milk-, and dark-chocolate samples.

FIGURE 4: Raman spectrum of chocolate sample showing the major band positions.

Raman spectroscopy is well suited for measuring oils and fats because it provides a highly specific, multicomponent measurement in a process or laboratory environment. The specificity of Raman spectroscopy enables a simple visual inspection or spectral library comparison for qualitative insight or quantification using univariate or multivariate models. Examples in the historical literature and recent laboratory-to-process studies show that there is still much potential in the technique for basic science to process control applications. We are enthusiastic about emerging nonlinear techniques, new applications in darkly colored foods, and their increased use in manufacturing.

(6), 888–895 (2010). DOI: https://doi.org/10.1016/j.cell.2010.11.033.

(634), 12–14 (1929). DOI: https://doi.org/10.1039/AN9295400012.

Official Methods and Recommended Practices of the AOCS

 (American Oil Chemists Society, Champaign, IL, 7th Ed., 2017).

 (John Wiley & Sons, Inc., Hoboken, NJ, 2000) Vol. 157.

(5) R. Manoharan, J.J. Baraga, M.S. Feld, and R.P. Rava, 

(2), 211–33 (1992). DOI: https://doi.org/10.1016/10111344(92)80009-K

(6) A. Kienle, L. Lilge, M.S. Patterson, R. Hibst, R. Steiner, and B.C. Wilson, 

(13), 2304–2314 (1996). DOI: https://doi.org/10.1364/AO.35.002304

(11), R37–R61 (2013). DOI: https://doi.org/10.1088/00319155/58/11/R37

(1), 88–95 (1945). DOI: https://doi.org/10.1002/bscb.19450540108

(9–12), 328–338 (1947). DOI: https://doi.org/10.1002/bscb.19470560907

(9), 711–720 (1940). DOI: https://doi.org/10.1063/1.1750742

(12) J.W. McCutcheon, M.F. Crawford, and H.L. Welsh, 

(1), 9–11 (1941). DOI: https://doi.org/10.1007/BF02544221

(1), 88 (1939). DOI: https://doi.org/10.1063/1.1750332

(2), 191 (1947). DOI: https://doi.org/10.6028/jres.038.009

Biochim. Biophys. Acta BBA - Lipids Lipid Metab.

(2), 245–255 (1973). DOI: https://doi.org/10.1016/0005-2760(73)90250-6

(8), 494–498 (1972). DOI: https://doi.org/10.1007/BF02582487

(23), 3901–3904 (1973). DOI: https://doi.org/10.1139/v73-581

(4), published online only (1973). Available from: https://pubs.acs.org/doi/pdf/10.1021/ jf60188a001. DOI: https://doi.org/10.1021/ jf60188a001

(20) V. Baeten, P. Hourant, M.T. Morales, and R. Aparicio, 

(7), 2638–2646 (1998). DOI: https://doi.org/10.1021/jf9707851

(21) S. Li, Y. Li, R. Yi, L. Liu, and J. Qu, 

, available online only (2020). Available from: https://www.frontiersin.org/article/10.3389/ fphy.2020.598420. DOI: https://doi.org/10.3389/ fphy.2020.598420

(22) R.C. Prince, R.R. Frontiera, and E.O. Potma, 

(7), 5070–5094 (2017). DOI: https://doi.org/10.1021/acs.chemrev.6b00545

(23) K. Czamara, K. Majzner, M.Z. Pacia, K. Kochan, A. Kaczor, and M. Baranska, 

(1), 4–20 (2015). DOI: https://doi.org/10.1002/jrs.4607

(1), 3–11 (1994). DOI: https://doi.org/10.1016/0924-2244(94)90042-6

(25) V. Baeten, M. Meurens, M.T. Morales, and R. Aparicio, 

(8), 2225–2230 (1996). DOI: https://doi.org/10.1021/jf9600115

(26) N. McReynolds, J.M. Auñón Garcia, Z. Guengerich, T.K. Smith, and K. Dholakia, 

(11), 1872–82 (2016). DOI: https://doi.org/10.1177/0003702816645931

(27) T.K. de Lima, M. Musso, and D. Bertoldo Menezes, 

, 127454 (2022). DOI: https://doi.org/10.1016/j.foodchem.2020.127454

(28) H. Jin, H. Li, Z. Yin, Y. Zhu, A. Lu, D. Zhao, et al, 

, 130191 (2021). DOI: https://doi.org/10.1016/j.foodchem.2021.130191

(29) Y. Lee, J. Kim, H. Jeong, and H. Chung, 

(2022). DOI: https://doi.org/10.1080/05704928.2022.2051535

(30) F. Kwofie, B.K. Lavine, J. Ottaway, and K. Booksh, 

(31) E. Kakouri, P.K. Revelou, C. Kanakis, D. Daferera, C.S. Pappas, and P.A. Tarantilis, 

(7), 1565 (2021). DOI: https://doi.org/10.3390/foods10071565

(32) S. Portarena, C. Anselmi, C. Zadra, D. Farinelli, F. Famiani, C. Baldacchini, et al, 

, 137–145 (2019). DOI: https://doi.org/10.1016/j.foodcont.2018.09.011

(33) A. Nedeljković, P. Rösch, J. Popp, J. Miočinović, M. Radovanović, and P. Pudja, 

(5), 1315–1320 (2016). DOI: https://doi.org/10.1007/s12161-015-0317-1

(34) J. De Gelder, K. De Gussem, P. Vandenabeele, and L. Moens, 

(9), 1133–1147 (2007). DOI: https://doi.org/10.1002/jrs.1734

(35) I.A. Larmour, K. Faulds, and D. Graham, 

(9), 1230–1232 (2010). DOI: https://doi.org/10.1039/C0AY00320D

(36) L.N. de Oliveira, J.C. de R. Castro, M.A.L. de Oliveira, and L.F.C. de Oliveira, 

(6), 1598–607 (2015). DOI: https://doi.org/10.5740/jaoacint.15-083

(37) P. Vargas Jentzsch, V. Ciobotă, W. Salinas, B. Kampe, P.M. Aponte, P. Rösch, et al, 

, 274–280 (2016). DOI: https://doi.org/10.1016/j.foodchem.2016.05.017

(38) H.G.M. Edwards, S.E.J. Villar, L.F.C. de Oliveira, and M.L. Hyaric, 

(1), 175–180 (2005). DOI: https://doi.org/10.1016/j.aca.2005.02.039

(39) N. Devos, D. Reyman, and S. Sanchez-Cortés, 

, 128301 (2021). DOI: https://doi.org/10.1016/j.foodchem.2020.128301

(40) J. Heuler, S. He, S. Ambardar, and D.V. Voronine, 

(1), 9833 (2020). DOI: https://doi.org/10.1038/s41598-020-66820-1

(41) K. Esmonde-White, M. Lewis, and I. Lewis, “Raman Spectroscopy of Chocolate using 1000 nm Excitation,” presented at SciX Conference, Providence, RI (2021).

(42) S. Bresson, D. Rousseau, S. Ghosh, M.E. Marssi, and V. Faivre, 

(8), 992–1004 (2011). DOI: https://doi.org/10.1002/ejlt.201100088

 are with Endress+Hauser. Direct correspondence to: 

Articles

Several types of Raman spectroscopy, including Fourier transform (FT)–Raman and dispersive Raman, are well suited to examine and understand the fat compositional heterogeneity in solid foods, identify polymorph or crystallinity, and measure fatty acid saturation.

Raman Spectroscopy in Analyzing Fats and Oils in Foods

Raman spectroscopy is now a first-choice process analytical technology for several reasons. Within a short time frame, Raman spectroscopy has transitioned from an esoteric laboratory technique to a practical industrial tool in laboratory and manufacturing applications. We highlight important technology milestones in Raman spectroscopy’s history, modern approaches to sampling probes, and the role of instrument accuracy in analytical model development. Applications of Raman spectroscopy in process control and understanding its fundamentals form the basis for expanding Raman spectroscopy in new applications and integrating Raman spectroscopy into new process automation strategies.

Raman spectroscopy is an optical spectroscopy technique based on inelastically scattered light. The resulting Raman spectrum provides a “molecular fingerprint” of the sample, delivering highly specific information about chemical composition and molecular structure without sample preparation. We provide an overview of Raman spectroscopy for industrial applications.


The specificity of a Raman spectrum is powerful because it provides a chemical fingerprint of a sample, thus allowing a qualitative assessment of a material’s chemical composition and molecular structure. From the first reports of the Raman effect in 1928 until the early 1970s, technology for certain components, such as lasers and scientific cameras, enabled Raman spectroscopy to be used for academic laboratory studies in geology, biology, and chemistry fields. Although these developments in Raman spectroscopy represented a significant advance over original instrumentation, Raman spectroscopy instrumentation was still complex, inefficient, and difficult to operate. One reason for this difficulty was the inefficiency and instability of optical components such as filters and wavelength separators. Until the early 1990s, it was a challenge for Raman spectroscopists to simply measure a usable Raman spectrum. However, the academic community consensus was that the challenges of using Raman spectroscopy were worthwhile to get the composition and molecular structure information unobtainable via other spectroscopy techniques. In the 1990s, however, process and quantitative applications were limited because of instrument instability.

The standardization of optical components represented the next generation of enabling technologies for Raman spectroscopy outside the academic laboratory. Standardizing optical components and simplifying instrument design in the 1980s and 1990s resulted in the first generation of commercially available Raman instruments for microscopy and benchtop applications. Quantitative Raman spectroscopy for industrial applications became more practical during the 1980s and 1990s. Since then, additional improvements in optical components, laser stability, integrated spectrograph design, and fiber-optic sampling probes have helped make Raman spectroscopy a practical technology outside the laboratory. Today, Raman spectroscopy can be applied in various research settings. These include portable, industrial, environmental, and clinical research settings. In industrial process applications, Raman spectroscopy is a proven process analytical technology (PAT) in pharmaceutical, bioprocessing, chemical, and food applications.

As a PAT, Raman spectroscopy can be integrated into an in-process or manufacturing site in several ways.


Figure 1 shows the differences between inline, online, at-line, and offline installations. There are many factors that affect choices about the type of installation, such as process chemistry, cycle time, presence of hazardous conditions at the site, skill level of the operator, and application needs.

Several excellent in-depth discussions of Raman spectroscopy instrumentation fundamentals have been published (1–3). Briefly, instrumentation used to collect a Raman spectrum include a light source, a sampling optic, a signal collection optic (interchangeably called a 

), and a low-noise detector. The light source, typically a laser, is delivered to the sample either by free-standing optics, a microscope, or a fiber-optic sampling probe. A spectrograph based on holographic volume technology enables simultaneous collection of all wavelengths without moving parts, resulting in a fast, robust, and reliable instrument. The use of non-moving parts in the spectrograph imparts the advantage of consistently high spectral performance. This consistency allows the user to perform calibration transfer from instrument to instrument seamlessly and prevented calibration rework if there is a laser or
detector change.

Sampling from Nanoscale to Megaton Production

One benefit of Raman spectroscopy is sampling flexibility. For industrial uses, sampling is available in a microscope, sample chambers, or fiber-optic probes for immersion sampling or through sight glass. At the nano- and micrometer scale, Raman microscopy is a valuable laboratory instrument because it provides detailed chemical information at high spatial resolution (4). Commercially available Raman microscopes, first introduced in the early 1970s, enabled point spectroscopy and imaging studies in academic research laboratories in biomedical, chemical, materials, semiconductor, and pharmaceutical fields (5–8). Raman microscopy applications have continued to grow and have expanded into the nanoscale. Tip-enhanced Raman spectroscopy (TERS) incorporates a metallic atomic force microscopy (AFM) needle into a Raman microscope. With TERS, Raman spectroscopy on the nanometer scale can be achieved (9,10). Surface-enhanced Raman spectroscopy (SERS) is another enhancement approach that can increase Raman signal 102 to 106 times and provide detailed chemical information (11,12). New substrates for SERS have expanded its application from the analysis of single molecules to biomarkers. In addition to enhancement mediated through metal-substrate plasmonics, as in TERS and SERS, coherent Raman spectroscopy approaches such as coherent anti-Stokes Raman spectroscopy and stimulated Raman spectroscopy can provide rapid and high-quality signals for imaging or spectroscopy applications (13,14). These nano- and micro- scale approaches are undergoing a resurgence of interest because of the development of new plasmonic materials, the extension of the techniques to biological and clinical applications, and commercial interest. These approaches are used primarily in the laboratory across many industries. Industry applications of Raman nanoscopy and microscopy can include testing of multilayer films, understanding degradation mechanisms, quality control, and high-resolution imaging in melts, solid dispersions, or polymers.

In environments outside the laboratory at larger scales, fiber-optic probes provide a convenient and reliable sampling format. Sampling probes are useful because they can be directly immersed into a process, into a slip stream, or outside a sight glass. The use of visible or near-infrared fiber optics means that the probe can be placed in a process that is meters away from the Raman spectrograph. Therefore, the probe can stay inside the process media, and Raman spectra can be recorded automatically without needing to physically collect a sample for laboratory measurement.

Fiber-optic sampling probes are a key technology enabling process Raman spectroscopy for many reasons. One benefit of using sampling probes is that many industrial processes pose contamination and safety risks when manually collecting a sample for laboratory analysis. For example, in the chemical industry, processes may involve toxic chemicals or hazardous conditions, increasing the safety risks of manual sampling (15). In upstream bioprocessing, manual sampling increases the risk of contaminating the sterile cell culture.

Another reason that sampling probes are beneficial is because they can measure in solids, liquids, gases, or turbid media without requiring specialized adaptors or sample preconditioners. Its ability to maximize signal collection in each phase of matter requires a fundamental understanding of photon migration and the physical conditions at the probe–process interface before building specialized probes based on that understanding.

Chemical analysis of solid or turbid media requires special consideration to capture the spatial heterogeneity of the layers or bulk product. Raman spectroscopy is widely used to analyze solid heterogeneous samples because it provides detailed chemical information and requires no sample homogenization, allowing for bulk or cross-sectional analysis in a spatially heterogeneous sample. Some examples of solid or turbid materials monitored by process Raman spectroscopy include pharmaceutical tablets, slurries, graft copolymers, and processed foods. For these materials, their optical scattering properties are important to understand because they have an impact on the measured sampling volume (16).

Understanding the effects of optical scattering on Raman signal recovery guides the design and selection of a fiber-optic probe. There are applications that should harness the effects of optical scattering to obtain a representative sample or minimize the effects of optical scattering from bubbles or particulates. Figure 2 shows the principle of photon scattering through a layered solid system and how fiber geometries can be tailored to optimize collection of a surface or subsurface Raman signal.

Figure 3 shows commercially available variants of sampling probes for Raman spectroscopy using optical fibers. A commonly used approach is backscattered Raman spectroscopy, which is highlighted in Figure 2. In backscattered Raman spectroscopy, incident and Raman-scattered light share the same optical path. This geometry allows the collection of a small volume. In the layered system shown in Figure 2, a backscattered probe will collect a signal from the surface layer. This type of probe geometry can be applied toward measurement in liquids, especially to reduce potentially deleterious effects of bubbles, turbidity in the liquid, or immiscible liquids. This type of probe can also be used to measure surfaces of solids.

To obtain signals from a larger volume, collection fibers can be offset from the excitation fiber. Large volumetric Raman spectroscopy uses a wide laser beam and multiple collection fibers to obtain data from surface and subsurface layers. Large volumetric sampling is a useful approach for in-process measurements of content uniformity or processing solids and turbid media. Spatially offset Raman spectroscopy (SORS) and transmission Raman spectroscopy utilize a single-point excitation with spatially offset collection fibers. In SORS, the fibers are typically in the same plane, and in transmission Raman spectroscopy, the collection fibers are typically 180° from the excitation fiber. These probes are used in solids applications ranging from through-container raw material identification to in-process quality assurance of processes involving solids or turbid media such as blending. Together, these four variants provide sampling flexibility from early development to manufacturing, in applications from laboratory quality assurance (QA) to in-process monitoring, testing, or control.

There are many approaches to modeling spectroscopic data with either a univariate or multivariate model. Multivariate models are more commonly used, and these models can incorporate first principle bounds in the model (known as 

) or be based on correlation or covariance without any a priori knowledge (known as 

) (17). One important consideration for industrial applications is model transferability. 

 can refer to a) transfer across different instruments that measure the same process, or b) expanding the scope of an existing model to account for changes in process conditions, such as pressure, temperature, media composition, and reactor size. An excellent 2018 review on calibration transfer discusses how hardware design affects model transfer (18). Hardware built for high-accuracy with a uniform internal design, but with housing configurable to the installation environment, has a few advantages, including reproducible cross-instrument performance, compatibility with “turn-key” sampling probes, and ease of servicing. This flexible design allows the user to perform calibration transfer from instrument to instrument and prevents calibration rework in the event of subsampling, or a laser or detector change. Another important consideration is the choice of probe. Probe selection impacts representative sampling and model development, and certain probe optics may adversely affect model transferability or robustness in turbid media and solids (19,20).

Raman spectroscopy has been successfully applied to upstream biopharmaceutical development in industrial settings since 2010 (17). Whether for cell culture or fermentation bioprocesses, in batch or continuous mode, Raman spectroscopy has many proven benefits, including the ability to simultaneously measure nutrients, metabolites, and cell viability. Raman spectroscopy also allows for the cross-scale method transfer from the laboratory to a current good manufacturing practice (cGMP) environment without significant method adjustments (21–23). Recent studies have demonstrated that Raman spectroscopy-based feedback control of glucose feeding or lactate accumulation in cell cultures has benefits of improved monoclonal antibody product quality and increased titer (24–26). Feedback-based control was also achieved in highly fluorescing cell culture and fermentation applications using wavelengths closer to shortwave infrared (1–1.4 μm) (27,28). More recent work has further expanded upstream monitoring to amino acids and titer predictions (29,30). In downstream applications, Raman spectroscopy can be equally powerful and we have seen recent successful cases of Raman spectroscopy used to quantify protein concentration and to monitor chromatography-based purification and crystallization (31,32).

As performance demands on polymeric materials have increased, polymerization processes have become more complex and often require tighter control of process parameters. Enabling mega-volume processes that can consistently make a quality product safely have increased the demand for PAT with high uptime that enables real-time, closed-loop process control. An example of successful polymer analysis by Raman spectroscopy is polystyrene, an important polymeric material. Figure 4 shows a Raman spectrum of polystyrene. What should be noted about the spectrum is that the peaks correspond to known chemical moieties, and they are sharp, identifiable, and quantifiable, which makes robust process monitoring and control possible. Styrene polymerization can be easily quantified using the area or intensity ratio of the styrene vinyl bond at 1630 cm

Raman spectroscopy-derived concentration predictions closely match measurements carried out by gravimetry, but without the need for sample extraction. The work reported by Brun’s team indicates that Raman spectroscopy could be part of a control strategy to avoid process upsets and ensure quality of a polymer product (33). Specifically, these results demonstrate the utility of Raman spectroscopy for monitoring the polymerization of styrene into polystyrene. The data were able to be generated quickly (in a matter of seconds), and they corresponded well with data obtained by classical gravimetric methods that are known to give good results but are also cumbersome and time-consuming.

Csontos’s group developed an effective feedback control loop based on Raman spectroscopy for a hazardous exothermic oxidation reaction (34). They were able to determine the end point of the process reaction, following the quantification of reaction components and an unstable intermediate. This was another successful example of Raman spectroscopy giving visibility into an industrial chemical reaction allowing for real-time control of the overall process. Similarly, Hart’s group was able to develop an accurate Raman spectroscopy calibration model for end-point determination of an etherification, where the residual level of chloropyrazine starting material needed to be minimized (35). At the outset, they were aware the reaction may be scale dependent because a heterogenous base (K

) was used. They successfully demonstrated that the calibration work for reaction end-point determination done at the laboratory scale was easily transferred to the scaled-up pilot plant, with predicted results closely matching manual off-line high-performance liquid chromatography results.

Raman spectroscopy has also been used to improve our understanding of the reaction kinetics of many processes, such as mechanochemical Knoevenagel condensations (36). In one study, crystallization properties were examined for the condensation of three fluorinated benzaldehyde derivatives and malononitrile by Hart’s group using in situ
Raman spectroscopy.

As we continue to find ways of dealing with climate change, Raman spectroscopy will play an increasingly significant role in controlling and optimizing industrial processes that involve greenhouse gases like CO

. Recently, Jinadasa’s group demonstrated the ability of Raman spectroscopy to effectively monitor speciation of a CO

capture process for both the lean and rich amine streams (37). Other chemical processes involving carbon chemistry, such as the common Suzuki-coupling reactions of the C–C bond, can also benefit from Raman spectroscopy through yield optimization. In this scenario, parameters like temperature and concentration can be varied independently within reaction limits to find optimal operating conditions. In doing so, Heteni and Janagap were able to develop a robust partial least squares (PLS) model using their experimental Raman spectra, and online predictions from the model agreed well with offline gas chromatography–mass spectrometry (GC–MS) measurements of reaction yield (38).

Raman has been used to examine crystallization and the solid state form of small-molecule active pharmaceutical ingredients (API) since the 1980s (39). As in many industries, applications of Raman spectroscopy in small-molecule pharmaceuticals have surged in the past three decades as technology has improved. Because of these technological advances, Raman spectroscopy is now a practical analysis technique in the pharmaceutical laboratory and process environments. As a PAT in API manufacturing, Raman spectroscopy has demonstrated value from scientific understanding to process control. A recent paper by Nagy’s group highlights the industrial trend for using Raman spectroscopy for real-time monitoring and control for secondary processing batch steps (40). We are enthusiastic about recent reports in continuous solids monitoring and solid phase unit operations, including in situ control of crystallization and polymorphism, process-induced transformations, low-dose formulations, and tablet coating. Automated Raman spectroscopy measurements are compatible with the fast cycle times and can provide process understanding for continuous solids production approaches, including twin-screw granulation and hot melt extrusion (41,42).

The food and beverage industry works in a highly regulated environment with strict controls on quality and safety. In this respect, there are similarities between food and beverage production and biopharmaceutics and small molecule API manufacturing. Although the PAT framework has been successfully used in pharmaceutical and biopharmaceutical manufacturing since 2004, its application has lagged in the food and beverage industry until recently. A more current interpretation of PAT in food and beverage is that it represents “a silent revolution in industrial quality control in food processing” (43). As a result, there are more reports of Raman spectroscopy in food processing (44,45). One application area that is particularly benefitting from Raman spectroscopy is meat and fish processing. Meat and fish quality is based on taste, texture, and appearance. All of these quality attributes have a basis in the specimen’s chemical properties, making Raman spectroscopy well-suited to providing a fast, non-destructive, and multiattribute quality measurement (46). Table I shows an overview of Raman spectroscopy-provided laboratory measurements for meat or fish quality attributes.

Laboratory studies demonstrate that Raman spectroscopy can provide both a fast and effective analytical method for determining iodine value, polyunsaturated fatty acids, monounsaturated fatty acids, and saturated fatty acids in samples of pork back fat (47–49). Even in a highly sophisticated meat processing plant, a fair amount of byproducts are generated. One strategy to minimize by-product waste is to use it as input for creating other valuable materials such as hydrolyzed proteins, and new research is showing the promise of Raman spectroscopy-based feedback control in this application (50,51).

Raman spectroscopy is a valuable PAT for many applications across multiple industries. The molecular specificity, ability to be directly coupled to a reaction vessel, and compatibility with solids, liquids, gases, and turbid media are advantages of Raman spectroscopy. Application success and rapid return-on-investment are realized in Raman spectroscopy-based feedback control because it enables in-process corrections and ensures process and product quality. The established history of process Raman spectroscopy in molecular identification, quantification, and process monitoring form the basis for continuing to expand Raman spectroscopy to new applications. We are enthusiastic about bringing Raman spectroscopy even further into industry through the adoption of industry 4.0 principles in PAT, increased automation, and even more integration of Raman spectroscopy into process control applications.


We thank Dr. Linda Kidder and Dr. Casey Kneale for helpful discussions during manuscript preparation.

Handbook of Raman Spectroscopy: From the Research Laboratory to the Process Line (

CRC Press, London, United Kingdom, 2001).

, K.A. Bakeev, Ed. (Blackwell Publishing, Oxford, United Kingdom, 2005).

Practical Raman Spectroscopy: An Introduction

Raman Microscopy Developments and Applications

Raman, Infrared, and Near-Infrared Chemical Imaging

, Y.O. Slobodan Šašić, Ed. (John Wiley & Sons, Inc., New York, New York, 2010), pp. 109–131.

Raman Imaging: Techniques and Applications

 (Springer Berlin Heidelberg, Berlin, Heidelberg, Germany, 2012), vol. 168. 

https://doi.org/10.1007/978-3-642-28252-2

R.M. Stöckle, Y.D. Suh, V. Deckert, R. Zenobi, 

https://doi.org/10.1016/S0009-2614(99)01451-7

P.L. Stiles, J.A. Dieringer, N.C. Shah, and R.P. Van Duyne, 

https://doi.org/10.1146/annurev.anchem.1.031207.112814

C.L. Haynes, A.D. McFarland, and R.P. Van Duyne, 

E. Berrocal, D.L. Sedarsky, M.E. Paciaroni, I.V. Meglinski, and M.A. Linne, 

K.A. Esmonde-White, M. Cuellar, C. Uerpmann, B. Lenain, and I.R. Lewis, 

https://doi.org/10.1007/s00216-016-9824-1

E.F. Olsen, C. Baustad, B. Egelandsdal, E.O. Rukke, and T. Isaksson, 

https://doi.org/10.1016/j.meatsci.2009.12.008

H.J. van Manen, R. Bloemenkamp, and O.F. van den Brink, 

B. Berry, J. Moretto, T. Matthews, J. Smelko, K. Wiltberger, 

H. Mehdizadeh, D. Lauri, M.K. Karry, M. Moshgbar, R. Procopio-Melino, and D. Drapeau, 

N.R. Abu-Absi, B.M. Kenty, M.E. Cuellar, M.C. Borys, S. Sakhamuri, D.J. Strachan, M.C. Hausladen, and Z.J. Li, 

https://doi.org/10.1016/j.jprocont.2014.02.007

B.N. Berry, T.M. Dobrowsky, R.C. Timson, R. Kshirsagar, T. Ryll, and K. Wiltberger, 

T.E. Matthews, B.N. Berry, J. Smelko, J. Moretto, B. Moore, and K. Wiltberger, 

(11), 2416–2424 (2016). https://doi.org/10.1002/bit.26018

T.E. Matthews, J.P. Smelko, B. Berry, S. Romero-Torres, D. Hill, R. Kshirsagar, and K. Wiltberger, 

S. Goldrick, D. Lovett, G. Montague, and B. Lennox, 

https://doi.org/10.3390/bioengineering5040079

H. Bhatia, H. Mehdizadeh, D. Drapeau, S. Yoon, 

S. André, L.S. Cristau, S. Gaillard, O. Devos, E. Calvosa, and L. Duponchel, 

https://doi.org/10.1016/j.aca.2015.08.050

D. Yilmaz, H. Mehdizadeh, D. Navarro, A. Shehzad, M. O’Connor, and P. McCormick, 

F. Feidl, S. Garbellini, F.M. Luna,S. Vogg, J. Souquet, H. Broly, M. Morbidelli, and A. Butté, 

N. Brun, I. Youssef, M.C. Chevrel, D. Chapron, C. Schrauwen, S. Hoppe, P. Bourson, and A. Durand, 

I. Csontos, H. Pataki, A. Farkas, H. Bata, B. Vajna, Z.K. Nagy, G. Keglevich, G.J. Marosi, 

R.J. Hart, N.I. Pedge, A.R. Steven, and K. Sutcliffe, 

S. Haferkamp, W. Kraus, and F. Emmerling, 

https://doi.org/10.1007/s10853-018-2492-0

M.H.W.N. Jinadasa, K.J. Jens, L.E. Øi, and M. Halstensen, 

https://doi.org/10.1016/j.egypro.2017.03.1282

https://doi.org/10.1016/j.vibspec.2018.11.008

B. Nagy, A. Farkas, E. Borbás, P. Vass, Z.K. Nagy, and G. Marosi, 

https://doi.org/10.1208/s12249-018-1201-2

W. Meng, A. Román, S. Panikar, C. O’Callaghan, S. Gilliam, R. Ramachandran, and F. Muzzio, 

https://doi.org/10.1016/j.apt.2019.01.017

Raman spectroscopy is a valuable process analytical technology (PAT) for many applications across multiple industries, as a result of its many advantages, such as molecular specificity, ability to be directly coupled to a reaction vessel, and compatibility with solids, liquids, gases, and turbid media.

Ian Lewis

Articles

Raman Spectroscopy in Analyzing Fats and Oils in Foods

Making Industrial Raman Spectroscopy Practical