Lipids are perhaps
the most diverse of the main food groups, with the accepted definition grouping
together a huge range of types of molecule on the basis of their shared
insolubility in water 1. The most common class of lipids that are found
in foods are based on the fatty acid, which usually occurs in nature as
straight, even-numbered chains, with or without unsaturated bonds. Three of the
most common fatty acids; stearic acid, elaidic acid and oleic acids have 18-carbon
chains, these are shown in Figure 1. It is the composition of the fatty acids
that determine the properties and uses of many edible fats since the chain
length and unsaturation level determines melting points, energy available for
metabolism, polarity and crystal packing1-3.
The standard shorthand
nomenclature for fatty acids is sufficient to inform the reader of the full
chemical structure of the acid. The notation is 2:
Nc: Nu i
Nc is the number
of carbons in the acid/acyl chain, including the carbonyl and methyl terminals.
Nu is the number of unsaturated bonds in the chain. If 0
then formula is complete.
i is the isomerisation of the
P is the position of the first carbon of the unsaturated bond,
counting from the carbonyl end. For centrally positioned monounsaturated bonds
the positional number for these isomers is simply the chain length divided by
In my studies two main
classes of fatty acid based lipids were used; Fatty Acid Methyl
Esters (FAME) were used as model systems4-6 while triglycerides were studied in edible fats7-9. Shorthand codes for triglycerides consist of one or two letter
codes identifying each fatty acid in the order they are bound to the glyceride
with position 1 first. For example, POS consists of a palmitoyl bound to
position 1 on the glycerine, oleoyl bound to the middle position and stearoyl
bound to the third position. A short list of the common abbreviations is given
in Table 1
Code in Triglyceride
Code in Triglyceride
Table 1 List of the shorthand codes for selected fatty acids in
triglycerides. The triglycerides’ codes are as found in Stauffer 1996.
Early applications of
Raman spectroscopy to fat and oil analysis
Raman spectroscopy has been
used in a number of investigations of lipid composition, crystal structure and
purity assays. Early work was primarily focused on phospholipids, with
extensive research into the effect of internal rotation of the fatty acid
chains. In 1972 Bailey published a paper10 in which Raman spectroscopy was used to predict the cis/trans
isomer ratio from model mixtures of pure triglycerides and FAMEs. During the
next eighteen years one other paper was published on the application of Raman
spectroscopy to the chemical analysis of edible fat, a short paper on the
prediction of saturation levels 11. Sadeghi-Jorabchi re-ignited interest with a pair of papers in
which he determined the unsaturation levels of oils and fats 12,
13. Since then, papers have been published on oil
unsaturation parameters 14-17, oil adulteration 18-21, biopsy22,
23, frying oil
deterioration 24and lypolysis 25.
My Raman spectroscopic investigation of fatty acids
I started the study of
fatty acid based lipids with a critical evaluation of a number of homologous
families of fatty acid methyl esters. A homologous (literally ‘same word’)
family is a series of chemical compounds that differs in discrete chemical
units. The homologous series chosen where saturated fatty acids (varying in
chain length from 4 to 22 carbons long, in one carbon increments),
monounsaturated fatty acids (each had a central unsaturated bond and chain
length varied from 14 to 22) and 18 carbon (all fatty acids had 18 carbons in
the chain, but the number of unsaturated bonds ranged from 0 to 3).
Figure 2 shows the Raman
spectra of some fatty acids, illustrating a saturated and an unsaturated fatty
acid in the liquid state, along with a fatty acid in the molten state. The relationship
between the position of the Raman bands and the chemical unit giving rise to it
are well understood, and the assignments for each of the bands are given in
table 2. The band labelled 1, which ranged from 1730 to 1760 cm-1 in
the samples investigated arises from the carbonyl mode at the ester linkage. This
proved to be a very useful band for normalising the Raman signal as this
chemical unit is present in every fatty acid chain, thus standardises against molar
(per molecule) quantities. Band 2, found between 1610 and 1670 cm-1,
is a particular importance for nutritional information as it arises from the unsaturated
bonds in a fatty acid. Band3 and 4 arise from the methylene units that make up the
length of the fatty acid chain, although they do differ significantly in their
behaviour (band 3 contains contributions from all C-H bonds in the acid, band 4
only from CH2 groups not adjacent to either end of the molecule).
The relationship between
chemical structure and the Raman spectrum can readily be seen in Figure 3, which
compares the chemical structure of four saturated fatty acids with their
respective Raman spectra. A number of bands can be observed to increase with
the chain length and the inset plots the intensity of one of these (normalised
to the carbonyl mode) against chain length. There is a clear correlation
between the chain length and the intensity of this scissor mode at 1440 cm-1.
However, given that these were pure reference compounds the correlation was disappointing,
with R2 = 0.95. above 14 carbons in length the scatter increases,
and appears to have a see-saw effect about the central trend lines. Splitting
the trend into odd and even length fatty acids improves the trends to R2
= 0.99. The mystery behind this see-saw was unveiled by density functional
theory simulation of the Raman spectrum based on the chemical structure. The
modes affected by the see-saw in fact involved all the methylene groups moving
in synch with each other, i.e. adding each methylene unit affected the vibration
along the entire chain, rather than each methylene unit acting in isolation.
Odd and even chains have different centres of symmetry as odd chains are symmetric
about the central methylene unit, whereas even chains are symmetric about the
central C-C bond. Symmetry of the molecule determines vibrational modes, which
explains why the odd and even molecules do not fit the same trend. From this result we
can see that Raman should work well for analysing edible fats (the vast
majority are even chains) but care must be taken in more unusual fats. It would
be expected the difference between odd and even chains would actually be of
benefit in multivariate analysis as it creates additional variance.
In comparison the trends
observed in the 18 carbon homologous series were much more straightforward.
Figure 4 shows the Raman spectra (normalised about the carbonyl mode) and
chemical structure of the 18 carbon fatty acids studied. It is readily apparent
that several bands increase with unsaturation, while several decrease. The
reason why the CH2 modes decrease is that each unsaturated bond
eliminates two CH2 groups, replacing them with two CH groups. The
inset shows that the trends between the Raman bands for unsaturation and the degree
of unsaturation are very smooth, in contrast to the chain length trend. This is
because the unsaturated bonds are isolated from each other and their vibrations
are uncoupled. However, it is important to use th correct band for normalising;
the carbonyl mode for per molecule (molar) determination and a CH2
mode for mass (molal/degree of) unsaturation. As aluded to above the CH2
modes are not equal and the scissor mode is preferable as it also contains
contributions from CH2 modes adjacent to the ends as well as CH, CH3.
My Raman spectroscopic investigation of edible fats
We have seen above the close
relationship between the chemical composition of a lipid sample and its Raman
spectrum. These relationships strongly suggest the possibility of employing
Raman spectroscopy to analyse the fatty acid profile of more complex fat
samples. In order to assess this capability I measured the fatty acid profile
of butters and adipose tissue using gas chromatography and used these results
to regress against the recorded Raman spectra from the same samples.
Figure 5 shows the mean
Raman spectrum recorded from each of the four species studied in the adipose investigation,
pork, chicken, lamb and beef. Pork had the highest unsaturation per fatty acid
chain, while chicken had the shortest chain length. The beef and lamb spectra
were very similar, with both being much more saturated than either the pork or
chicken, while having an average chain length comparable to the pork. These
differences are borne out across the range of samples studied, as is evidenced
by the principal component analysis (PCA) score plot shown in Figure 6. PCA summarises
the differences in the data, with the consequence that samples that show little
difference lie close together while samples that are very different are widely
separated. The chicken samples are very isolated in the scatter plot, forming a
very tight group remote from any other samples. This indicates that this
species is radically different from the others, and indeed this is unsurprising
as the chicken is, of course, avian while the remaining species are mammalian. Within
the mammalian species there is another distinct separation, with the pork
samples barely touching the lamb samples, while the separation of lamb and beef
is limited. Unlike pigs, cows and sheep are both ruminant animals which proves
to be important for fatty acid accumulation. The ruminant digestion depends on
bacteria fermenting difficult to digest molecules such as cellulose, but these bacteria
also hydrogenate the unsaturated fatty acids present in the food. This action
was critical in the context of the butter study, in which the level of polyunsaturated
fatty acids was enhanced by dietary intervention by coating the linseeds so
that they were only digested in the true stomach after rumination.
PCA is an unsupervised
method, meaning that it is given no prior information on the samples before the
analysis. This means that its information is directed to the largest effects
that occur naturally in the data, but these may not be the same effects that best
describe the differences between the samples. The differences between poultry
and mammal (or between ruminant/non-ruminant) are sufficiently large that PCA
captures this separation, but the differences between beef and lamb are too
slight for it to pick up. A number of approaches were used that included
training a statistical method by defining the groups each sample belonged to.
More information can be found in reference 7, but Figure 7 shows one of these ‘supervised’
techniques known as the Kohonen neural network. In this plot the data space is
divided into a grid, with each intersection representing a spectrum that varies
along each axis. The data is plotted closest to the spectra it is most similar
to. Thus samples within each box are grouped together, and this is found to be
largely successful. Only four samples, indicated by arrows, were incorrectly
classified and all these were ruminant samples.
The Raman signals were also
correlated with the concentrations of individual fatty acids, and Figure 8 shows
the regression plots between the Raman spectrum and the GC measured fatty acid
content for two major fatty acids, oleic (18:1c) and linoleic (18:2c) acid. The
samples have been colour coded to demonstrate where each species lies within
the range. The Raman spectrum could be used to predict the content of 15 fatty
acids, giving a comprehensive summary of the fatty acid composition of these
edible fats in a single rapid measurement from an unextracted sample.
1. Hui, Y. H., Bailey's Industrial Oil and Fat Products. 5 ed.; John Wiley &
Sons Inc.: New York, 1996; Vol. 1.
2. Stauffer, C. E., Fats and Oils. 1 ed.; Eagan Press: St Paul, Minnisota, 1996.
3. Groff, J. L., et al., Advanced Nutrition and Human Metabolism.
2 ed.; West publishing Company: St Paul, 1995; p 575.
4. Oakes, R. E., et al., DFT Studies Of
Long-Chain Fames: Theoretical Justification For Determining Chain Length And
Unsaturation From Experimental Raman Spectra. Journal of Molecular Structure-Theochem 2003, 626, 27-45.
5. Oakes, R. E., et al., Conformations,
Vibrational Frequencies And Raman Intensities Of Short Chain Fatty Acid Methyl
Esters Using DFT With 6-31 G(d) And Sadlej pVTZ Basis Sets. Journal of Molecular Structure 2002, 586, (1-3), 91-110.
6. Beattie, J. R., et al., A Critical
Evaluation of Raman Spectroscopy for the Analysis of Lipids: Fatty Acid Methyl
Esters. Lipids 2004, 39, 407-419.
7. Beattie, J. R., et al., Classification
of adipose tissue species using raman spectroscopy. Lipids 2007, 42, (7),
8. Beattie, J. R., et al., Prediction of
adipose tissue composition using Raman spectroscopy: Average properties and
individual fatty acids. Lipids 2006, 41, (3), 287-294.
9. Beattie, J. R., et al., Multivariate
prediction of clarified butter composition using Raman spectroscopy. Lipids 2004, 39, (9), 897-906.
10. Bailey, G. F.; Horvat, R. J., Raman
Spectroscopy Analysis of the Cis/Trans Isomer Composition of Edible Vegetable
Oils. JAOCS 1972, 49, 494-498.
11. Butler, M., et al., Raman Spectral
Analysis of the 1300 cm-1 Region for Lipid and Membrane Studies. Chemistry and Physics of Lipids 1979, 29, 99-102.
12. Sadeghi-Jorabchi, H., et al.,
Determination Of The Total Unsaturation In Oils And Fats By Fourier Transform
Raman Spectroscopy. JAOCS 1990, 67, 483-486.
13. Sadeghi-Jorabchi, H., et al.,
Quantitative Analysis Of Oils And Fats By Fourier Transform Raman Spectroscopy.
Spectrochimica Acta 1991, 47A, 1449-1458.
14. Baeten, V., et al., Oil and fat
classification by FT-Raman spectroscopy. Journal
of Agricultural and Food Chemistry 1998,
46, (7), 2638-2646.
15. Chmielarz, B., et al., Studies On The
Double-Bond Positional Isomerization Process In Linseed Oil By UV, IR And
Raman-Spectroscopy. Journal of Molecular
Structure 1995, 348, 313-316.
16. da Silva, M. I. P., et al., Castor oil catalytic
hydrogenation reaction monitored by Raman spectroscopy. Materials Letters 2000,
45, (3-4), 197-202.
17. Lerner, B., et al. In Characterisation of Polyunsaturation in
Cooking Oils by the 910 FT-Raman., 2nd International on FT-IR,
Antwerp,Belgium, 1992; Vansant, E. F., Ed. University of Antwerp:
Antwerp,Belgium, 1992; pp 75-82.
18. Aparicio, R.; Baeten, V., Fats and Oils
authentication by FT-Raman. Oleagineux
Corps Gras Lipides 1998, 5, (4),
19. Davies, A. N., et al., Study of the use
of molecular spectroscopy for the authentication of extra virgin olive oils.
part I: Fourier transform Raman spectroscopy. Applied Spectroscopy 2000,
54, (12), 1864-1867.
20. Li-Chan, E., Developments in detection of
adulteration of olive oil. Trends in food
science & technology 1994,
21. Marigheto, N. A., et al., A comparison of
mid-infrared and Raman spectroscopies for the authentication of edible oils. Journal of the American Oil Chemists Society
1998, 75, (8), 987-992.
22. Baraga, J. J., et al., In-situ optical histochemistry of human
artery using near infrared Fourier transform Raman spectroscopy. Proccedings of the National Acedemy of
Sciences USA 1992, 89,
23. Frank, C. J., et al., Characterization of
Human Breast Biopsy Specimens with near-Ir Raman-Spectroscopy. Analytical Chemistry 1994, 66, (3), 319-326.
24. Engelsen, S. B., Explorative
spectrometric evaluations of frying oil deterioration. Journal of the American Oil Chemists Society 1997, 74, (12), 1495-1508.
25. Weldon, M. K.; Morris, M. D.,
Surface-enhanced Raman spectroscopic investigation of bacterial lipolysis in a
skin pore phantom. Applied Spectroscopy 2000, 54, (1), 20-23.