10 Chemometric Analysis of Raman and IR Spectra of Natural Dyes
297
studies focused on lycopene and β-carotene content in tomato products with NIR
spectrscopy in order to replace standard, time-consuming analytical methods. PLS
calibration, based on over one hundred spectra, was chosen. original spectra were
preprocessed by applying a multiplicative signal correction (mSC) or second derivative according to the algorithm proposed by Savitsky and golay. Prior analysis
the spectra were divided into three regions (4,000–6,000 cm
−1
, 6,000–8,000 cm
−1
and 8,000–10,000 cm
−1
). Calibrations were performed in each one, as well as in
combinations of them. mSC turned out in the described case the most effective
preprocessing technique as well as splitting the spectra into three distinct regions.
the results were perfectly comparable with those obtained with hPLC with the “R”
values over 0.998.
the mean values obtained by NIR spectroscopy and hPLC for eight carotenoids
(α-carotene, β-carotene, α-cryptoxanthin, β-cryptoxanthin, isolutein, lutein, violaxanthin, zeaxanthin) in over sixty maize varieties were equivalent [57, 59]. In order
to compare NIR and hPLC results, the modified PLS regression and the lowest
standard error of cross-validation were chosen. the correlation factor (R
2
) ranged
from 0.82 for lutein to 0.94 for zeaxanthin.
vis/NIR transmittance spectroscopy was successfully used to determine the total
levels of carotenoids and chlorophyll also in virgin oil [91]. to correct the signal
an initial smoothing technique combined with first derivative treatment was used.
Similarly to the former examples PLS regression was used to develop calibration
models. those models were successfully used to monitor on-line levels of the studied compounds during oil processing.
10.3.2 Flavonoids
the in situ detection of flavonoids is often disrupted by the overlapping the flavonoid marker bands with signals coming from components of biological matrix e.g.
lignin, cellulose.
In 1985, the RR microspectrometry technique was applied to the studies of skins
of mature Vitis vinifera “Pinot noir” berries and epidermal tissues of the flowers of
wild mallow ( Malva sylvestris) [92]. What is interesting, authors detected malvidin
3-glucoside in “Pinot noir” berries both in the quinonoidal base form (strong band
at 1653 cm
−1
) and in the flavylium form (features at 1,572, 1,600 and 1,648 cm
−1
).
the Raman mapping has been shown for green rooibos ( Aspalathus linearis)
where the aspalanthin, a dihydrochalcone, is the major flavonoid component [66].
From these results it can be seen that the aspalanthin is distributed heterogeneously
in the sample and the highest concentration was observed in the inner part of the
leaves. the Raman maps of aspalanthin distribution were obtained according to the
intensity of the band at 784 cm
−1
. the integration over the most intensive bands
in the range 1550 cm
−1
–1,670 cm
−1
ascribed to the phenyl rings and C = o stretching vibrations and 1,216 cm
−1
assigned to hydroxyl deformation and ring vibration gives incorrect results because these bands coincide to the signals of the other
297
studies focused on lycopene and β-carotene content in tomato products with NIR
spectrscopy in order to replace standard, time-consuming analytical methods. PLS
calibration, based on over one hundred spectra, was chosen. original spectra were
preprocessed by applying a multiplicative signal correction (mSC) or second derivative according to the algorithm proposed by Savitsky and golay. Prior analysis
the spectra were divided into three regions (4,000–6,000 cm
−1
, 6,000–8,000 cm
−1
and 8,000–10,000 cm
−1
). Calibrations were performed in each one, as well as in
combinations of them. mSC turned out in the described case the most effective
preprocessing technique as well as splitting the spectra into three distinct regions.
the results were perfectly comparable with those obtained with hPLC with the “R”
values over 0.998.
the mean values obtained by NIR spectroscopy and hPLC for eight carotenoids
(α-carotene, β-carotene, α-cryptoxanthin, β-cryptoxanthin, isolutein, lutein, violaxanthin, zeaxanthin) in over sixty maize varieties were equivalent [57, 59]. In order
to compare NIR and hPLC results, the modified PLS regression and the lowest
standard error of cross-validation were chosen. the correlation factor (R
2
) ranged
from 0.82 for lutein to 0.94 for zeaxanthin.
vis/NIR transmittance spectroscopy was successfully used to determine the total
levels of carotenoids and chlorophyll also in virgin oil [91]. to correct the signal
an initial smoothing technique combined with first derivative treatment was used.
Similarly to the former examples PLS regression was used to develop calibration
models. those models were successfully used to monitor on-line levels of the studied compounds during oil processing.
10.3.2 Flavonoids
the in situ detection of flavonoids is often disrupted by the overlapping the flavonoid marker bands with signals coming from components of biological matrix e.g.
lignin, cellulose.
In 1985, the RR microspectrometry technique was applied to the studies of skins
of mature Vitis vinifera “Pinot noir” berries and epidermal tissues of the flowers of
wild mallow ( Malva sylvestris) [92]. What is interesting, authors detected malvidin
3-glucoside in “Pinot noir” berries both in the quinonoidal base form (strong band
at 1653 cm
−1
) and in the flavylium form (features at 1,572, 1,600 and 1,648 cm
−1
).
the Raman mapping has been shown for green rooibos ( Aspalathus linearis)
where the aspalanthin, a dihydrochalcone, is the major flavonoid component [66].
From these results it can be seen that the aspalanthin is distributed heterogeneously
in the sample and the highest concentration was observed in the inner part of the
leaves. the Raman maps of aspalanthin distribution were obtained according to the
intensity of the band at 784 cm
−1
. the integration over the most intensive bands
in the range 1550 cm
−1
–1,670 cm
−1
ascribed to the phenyl rings and C = o stretching vibrations and 1,216 cm
−1
assigned to hydroxyl deformation and ring vibration gives incorrect results because these bands coincide to the signals of the other