Food Chemistry 181 (2015) 235–240
Contents lists available at ScienceDirect
Food Chemistry journal homepage: www.elsevier.com/locate/foodchem
Analytical Methods
Raman spectroscopy of white wines Coralie Martin a,b, Jean-Luc Bruneel a, François Guyon c, Bernard Médina c, Michael Jourdes d,e, Pierre-Louis Teissedre d,e, François Guillaume a,⇑ a
Université de Bordeaux, CNRS, ISM UMR 5255, 351 cours de la Libération, F-33405 Talence Cedex, France Advanced Track and Trace, ATT 99 avenue de la châtaigneraie, F-92500 Rueil Malmaison, France c Service Commun des Laboratoires, 3 Avenue du Dr A. Schweitzer, F-33608 Pessac, France d Université de Bordeaux, ISVV, EA4577 Unité de Recherche Oenologie, F-33140 Villenave D’ornon, France e INRA, ISVV, USC 1366 Œnologie, F-33140 Villenave d’Ornon, France b
a r t i c l e
i n f o
Article history: Received 24 July 2014 Received in revised form 15 January 2015 Accepted 14 February 2015 Available online 20 February 2015 Keywords: Raman spectroscopy Fluorescence spectroscopy White wines Phenolic composition Hydroxycinnamic acids Sugars
a b s t r a c t The feasibility of exploiting Raman scattering to analyze white wines has been investigated using 3 different wavelengths of the incoming laser radiation in the near-UV (325 nm), visible (532 nm) and near infrared (785 nm). To help in the interpretation of the Raman spectra, the absorption properties in the UV–visible range of two wine samples as well as their laser induced fluorescence have also been investigated. Thanks to the strong intensity enhancement of the Raman scattered light due to electronic resonance with 325 nm laser excitation, hydroxycinnamic acids may be detected and analyzed selectively. Fructose and glucose may also be easily detected below ca. 1000 cm 1. This feasibility study demonstrates the potential of the Raman spectroscopic technique for the analysis of white wines. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction Analytical methods are essential tools for wine quality control and authentication. Among all spectroscopy techniques currently used in this context, often complementary, there is a great need for low cost analytical tools that are small and light enough to be handled for field analysis. Optical methods based on the phenomenon of light absorption have experienced significant developments in recent years for the characterization of wines. These methods encompass absorption spectroscopy in the mid-infrared (MIR) and the near-infrared (NIR) for studying fundamental molecular vibrations and their harmonics (Bauer et al., 2008; Cozzolino, Dambergs, Janik, Cynkar, & Gishen, 2006; Cozzolino, McCarthy, & Bartowsky, 2012), absorption spectroscopy in the ultra-violet and visible (UV–vis) for probing electronic transitions (Acevedo, Jiménez, Maldonado, Domínguez, & Narváez, 2007; García-Jares & Médina, 1995; Harbertson & Spayd, 2006; Roig & Thomas, 2003; Urbano, Luque de Castro, Pérez, García-Olmo, & Gómez-Nieto, 2006). These techniques are well suited in an industrial context due to their ease of use, their measurement quickness, their relatively low financial cost and also because they can be small enough (miniaturized in a near future) for in situ operation. ⇑ Corresponding author. Fax: +33 (0)5 4000 3183. E-mail address: [email protected] (F. Guillaume). http://dx.doi.org/10.1016/j.foodchem.2015.02.076 0308-8146/Ó 2015 Elsevier Ltd. All rights reserved.
Surprisingly there has been very little research carried out on wines by means of spectroscopic techniques analyzing the emission of light. Indeed for wavelengths of light in the 260–1100 nm range, several phenomena will take place involving the electronic polarization of the molecules. Let us consider the interaction of a monochromatic electromagnetic radiation (laser) with molecules. If there is no absorption of the incoming radiation, elastic (Rayleigh) and inelastic (Raman) scattering of photons will occur. The spectral analysis of the Raman scattering provides information on molecular vibrations (Raman scattering effect is fully described in many books, see for instance (Dietzek, Cialla, Schmitt, & Popp, 2010)). If now the molecule absorbs the exciting radiation, two phenomena involving different mechanisms may take place if we exclude phosphorescence. The first one is fluorescence (Lakowicz, 2006) and the second is resonance Raman scattering (Dietzek et al., 2010). The resonance Raman spectrum will display maxima at the same positions to that of normal Raman scattering, but the vibrations coupled to the absorbing functional groups may be strongly intensified. To the best of our knowledge, only one paper has been published so far about Raman scattering of white wines (Meneghini et al., 2008) and none about resonance Raman scattering. A very few studies based on front face fluorescence spectroscopy for direct and global analyzes of wines have already been published (Airado-Rodríguez, Durán-Merás, Galeano-Díaz, & Wold, 2011; Dufour, Letort, Laguet, Lebecque, & Serra, 2006; Le Moigne et al.,
236
C. Martin et al. / Food Chemistry 181 (2015) 235–240
2008). However, fluorescence detection coupled to HPLC system have been extensively used to quantify flavanols (e.g. catechin, epicatechin. . .) and procyanidin dimer (e.g. procyanidin (GómezAlonso, García-Romero, & Hermosín-Gutiérrez, 2007; Silva, Ky, Jourdes, & Teissedre, 2012)). Molecular vibrations may be analyzed either in the MIR–NIR regions by probing the absorption of light by molecules, or by Raman and resonance Raman scattering of light in the visible range. Because the mechanisms of light-matter interaction are not the same for absorption and scattering, MIR–NIR and Raman spectroscopic techniques are complementary with distinct vibration selection rules. We will show that, for this reason, Raman spectroscopy possesses advantages for wine analysis. Also very small amount of sample is required (several lm3) as a Raman spectrometer may be coupled to a confocal microscope (a quite large majority of commercial Raman spectrometers combine Raman spectroscopy to optical microscopy). This method is a very effective tool in chemical analysis because it is non-destructive and usually does not require special preparation of the sample. Finally Raman spectrometers may be very compact and equipped with optical fibers allowing in situ measurements. For example mobile Raman spectrometers are efficient tools in the domains of arts and archeology (Colomban, 2012). The aim of this work is to investigate the potential of Raman spectroscopy for analyzing commercial wines. This preliminary study focuses on two samples of white wines (one dry and one medium) for establishing the potential of the technique. This study focus on white wines as their chemical composition is less complex than red wine one’s. Because the mechanisms at the origin of light emission depend on the absorption of the exciting radiation, we will first investigate the absorption of light in the UV–visible for the two samples of white wines. Then the emission spectra recorded using three wavelengths from the near-UV to the NIR for the laser excitation will be analyzed. As mentioned above, spectra where fluorescence emission and resonance Raman or normal Raman scattering take place are expected. 2. Materials and methods 2.1. Samples The white wines, one dry and one medium, that were investigated in this work originate from south-west of France. These two wines were chosen as visually their yellow color looked very similar and because their chemical composition must be significantly different. The first sample (sample #1) is a Bordeaux dry wine AOC ‘‘entre deux mers’’ vintage 2013 and the second (sample #2) sample is a medium wine AOC ‘‘côtes de Bergerac’’ vintage 2012. To help in the interpretation of the wine’s spectra, those of pure phenolic compounds or sugars dissolved in a wine model solution have been recorded. The used wine model solution was a hydroalcoholic solution (e.g. 12% ethanol) acidified with 5 g l 1 of tartaric acid with the pH adjusted at 3.5 (e.g. 3.5 is considered as the average wine acidity (Ribéreau-Gayon, Dubourdieu, et al., 2012; Ribéreau-Gayon, Glories, et al., 2012)) with a solution of NaOH (1 M). Phenolic acids and sugars with purity >99% were obtained commercially from Sigma–Aldrich. 2.2. Absorption spectroscopy UV–visible spectroscopy measurements were performed using a Lambda-650 UV–vis spectrophotometer (Perkin Elmer). Wine samples taken from freshly opened bottles were scanned in transmission mode (200–900 nm). Samples of white wines were placed in quartz cells of thickness 1 mm.
2.3. Emission spectroscopy All phenomena involving emission of light such as fluorescence, Raman scattering and resonance Raman scattering were recorded in the backscattering geometry by means of Raman spectrometers coupled to optical microscopes. Experiments using the 325 nm excitation wavelength of a He–Cd Laser were performed on the Raman spectrometer Labram-UV HR800 (Horiba Jobin Yvon) using a 4 UV-lens, experiments at 532 (frequency doubled Nd:YAG laser) and 785 nm (diode) were performed on the Xplora instrument (Horiba Jobin Yvon) using a 10 objective. The laser power at sample was 2 mW for 325 nm excitation wavelength, 13 mW for 532 nm and 50 mW for 785 nm. To analyze the emission of light at 325 nm excitation over a broad range of wavelengths, i.e. to analyze fluorescence, a 150 lines mm 1 diffraction grating was used. For Raman scattering a better spectral resolution in the range 2–5 cm 1 was needed and gratings with 2400 lines mm 1 for a 325 nm excitation wavelength, 1800 lines mm 1 at 532 nm and 1200 lines mm 1 at 785 nm were used. Samples of white wines and model solutions were placed in NMR glass tubes with overall diameter 4.97 mm and internal diameter 4.20 mm. The emission spectra were corrected for detector efficiency and could be therefore qualitatively compared to each other. However quantitative information on the absolute intensities of the emitted light cannot be provided in this study.
3. Results and discussion A difficulty to be overcome is related to the chemical nature of the wines with composition of about 12% of ethanol and 84% water. The other molecules include carboxylic acids (for example tartaric acid), sugars, glycerol and also polyphenols that represent a very small proportion of the total composition. As emphasized in the introduction of this paper, optical spectroscopy has been widely exploited for the analysis of wines and it is obvious that the obtained spectra will result from the superposition of all the optical responses of different molecules that make up the medium. A very large number of molecular species have been identified in wines, and two of these species alone represent about 96% of the total number of molecules. One may therefore easily understand that Raman signals are expected to be dominated by those of water and ethanol.
3.1. UV–visible absorption spectra In wines, mainly polyphenols will absorb light at wavelengths between 250 and 900 nm. The absorption features of these species are now well identified (Cerovic et al., 2002; Jurd, 1957). The UV– visible absorption spectra between 250 and 500 nm of the two samples of white wines are shown in Fig. 1a. There is no significant absorption around 500 nm, as expected for a white wine that does not contain anthocyanins known to have absorbance features at 267–275 and 475–545 nm. Hydroxycinnamic acids have absorbance maxima at 227–245 and 310–332 nm; benzoic acids show a single absorbance in the region of 235–305 nm and flavanols typically have maxima in the 250–270 and 350–390 nm regions. As white wines do not contain significant amounts of flavanols, the peaks around 263 nm may be therefore assigned mainly to the phenolic acids that are present in the white wines and the peak around 326 nm mainly to hydroxycinnamic acids. Of course these peaks are broad and overlap around 300 nm. Interestingly the two white wines have quite similar absorptions for wavelengths above 300 nm and the very weak tails of the spectra above 400 nm are responsible for their similar yellow color. In contrast,
237
C. Martin et al. / Food Chemistry 181 (2015) 235–240
(a)
263 nm
3.5
2.5 2.0 1.5 1.0
120 000
Intensity (counts/s)
Absorbance
3.0
273 nm
0.5
100 000 80 000
443 nm
60 000
437 nm
40 000
326 nm
300
(b)
438 nm
140 000
20 000
400 Wavelength (nm)
400
500
450
500 Wavelength (nm)
550
600
Fig. 1. (a) UV–visible absorption spectra (optical path 1 mm) of sample #1 (maximum at 273 nm) and sample #2 (263 nm). (b) Laser induced fluorescence spectra with 325 nm excitation of sample #1 (443 nm), sample #2 (438 nm) and caffeic acid model solution (437 nm). For convenience, the intensity of the spectrum of caffeic acid solution (200 mg l 1) has been multiplied by a factor of 10.
Ethanol CC stretching
O-H stretching
Raman Intensity
(b)
Raman Intensity
(a)
C-H stretching
1 000
2 000 Raman shift (cm-¹)
3 000
1 000
2 000
3 000
Raman shift (cm-¹)
Fig. 2. Raman spectra (including fluorescence) of (a) the dry white wine sample #1 and (b) the medium white wine sample #2 for three wavelengths of the excitation laser source: 325 nm (top), 532 nm (middle) and 785 nm (bottom). For comparing the spectra, the intensities were normalized versus that of the band assigned to the C–C stretching vibration of ethanol near 880 cm 1.
the absorption properties below 300 nm are totally different suggesting significant differences in their chemical composition.
could not be reached from the analysis of the UV–visible absorption spectra displayed in Fig. 1a where absorbance of the two wine samples around 325 nm is similar.
3.2. Laser induced fluorescence spectra 3.3. Raman and resonance Raman spectra The next step of the analysis is to investigate qualitatively fluorescence of the wine samples. The emission spectra of the wines shown in Fig. 1b were recorded with an excitation wavelength of 325 nm, i.e. at the maximum of absorption of hydroxycinnamic acids (Fig. 1a). The maxima around 440 nm are due to fluorescence emission of the white wines. These maxima of fluorescence are congruent with those known for hydroxycinnamic acids (Airado-Rodríguez et al., 2011; Promkatkaew et al., 2014; Putschögl, Zirak, & Penzkofer, 2008). As a matter of fact the maximum of fluorescence of caffeic acid (representative of hydroxycinnamic acids) in wine model solution is also around 440 nm (Fig. 1b) and the model solution of gallic acid (representative of hydroxybenzoic acids) does not display significant fluorescence with 325 nm excitation. With the 532 excitation wavelength, a very weak fluorescence is detected compared to that displayed in Fig. 1b, suggesting trace concentration of absorbing molecules belonging perhaps to the family of flavonoids. At 785 nm there is no detectable fluorescence. We may conclude that by exciting samples #1 and #2 with a laser with 325 nm wavelength, part of the incoming light is absorbed by hydroxycinnamic acids and related compounds giving rise to fluorescence with maxima around 440 nm. It should be pointed out that the intensity at the maximum of the fluorescence differs significantly for the two wines, suggesting that their composition in hydroxycinnamic acids is probably different. Such conclusion
To introduce the discussion on the Raman scattering of white wines, the emission spectra of the dry white wine sample #1 within the 200–3600 cm 1 spectral window are shown in Fig. 2a and those of the medium white wine sample #2 in Fig. 2b. In Fig. 2a are also shown the main Raman peaks assigned to ethanol and water molecules. Note however that polyphenols, sugars and other molecules will also contribute, but very weakly, to the intensity of the bands assigned to C–H and O–H stretching vibrations between 2800 and 3600 cm 1. Obviously the ethanol concentration may be easily evaluated from the ratio of the integrated intensities of the bands assigned to the C–H and O–H stretching vibrations using a calibration procedure. As mentioned in Section 1, we found in literature only one paper dealing with Raman scattering of white wines. The paper by (Meneghini et al., 2008) investigated the Raman scattering (with 532 nm excitation) of two samples of white wines restricting their analysis to the region of C–H stretching vibrations between 2600 and 3100 cm 1 to estimate, via a calibration procedure on the intensities, the ethanol and sucrose concentrations. For an excitation with wavelength 325 nm, the strong fluorescence background due to the absorption of hydroxycinnamic acids and related compounds already discussed above is observed in Fig. 2. However, as the maximum of fluorescence is around 432 nm (Fig. 1b), corresponding to a Raman shift of 7600 cm 1,
C. Martin et al. / Food Chemistry 181 (2015) 235–240
(d) (c) (b)
1631
1606
To strengthen this interpretation, the Raman spectra at 325 and 532 nm of model solutions of the two representative members of phenolic acids, gallic acid and caffeic acid, have been recorded (shown in Fig. 4). It comes out that, when comparing Figs. 3 and 4, the new peaks appearing for white wines using the 325 nm laser excitation may be safely assigned to hydroxycinnamic acids. We note however that the Raman spectra of gallic acid at 325 and 532 nm (Fig. 4) are not strictly identical around 1600 cm 1. Presumably gallic acid absorbs weakly light at 325 nm so that the intensity of some Raman bands may be weakly enhanced due to pre-resonance effects. Although these Raman measurements are not yet quantitative, we believe that the great potential of Raman spectroscopy to detect and measure hydroxycinnamic acids in wines is now established. It is also certainly possible to identify the main species of cinnamic acids (caffeic, p-coumaric, ferulic, etc.) present in the wine samples because their resonant Raman spectra should in principle not be identical. For example the shape and intensity of the peaks around 1600 cm 1 in Fig. 3 for the dry and the medium wine samples are significantly different. The detailed analysis of UV–visible absorption, fluorescence and Raman scattering of hydroxycinnamic acids combining experimental and time-dependent density functional theory (TD-DFT) approaches will be published in a near future. Other differences between the Raman spectra of samples #1 and #2 are observed below 1000 cm 1. Indeed, several lines of weak intensity are detected in the spectra of the medium wine and this regardless of the wavelength of excitation, which are not visible in the spectra of the dry white wine. In Fig. 5 are displayed the Raman spectra of samples #1 and #2 below 1000 cm 1 with 532 nm excitation. Indeed an excess of sugar in the medium wine compared to that in the dry one is expected. For comparison the Raman spectra of fructose and glucose model
(a) 800
1 000
1 200
1 400
1 600
Raman shift (cm-¹)
1 200
1 400
Raman shift (cm-¹)
1 600
1637
1 000
(b)
1607
Fig. 4. Raman spectra of gallic acid model solution recorded with excitations at 532 nm (a) and 325 nm (c), caffeic acid model solution at 532 nm (b) and 325 nm (d).
(d) Raman Intensity
Raman intensity
800
1173
(a)
Raman intensity
879
the fluorescence within the Raman spectral window (below ca. 3600 cm 1) is even weaker, allowing the analysis of the Raman peaks. For the 532 nm excitation wavelength, the fluorescence background is also weak allowing the analysis of the Raman scattering over the full spectral range, up to 4000 cm 1. For the 785 nm excitation wavelength, no fluorescence is observed but the analysis of the Raman scattering is limited as the detector of the spectrometer is not efficient for wavelengths larger than 980 nm (Raman shift 2500 cm 1). Fig. 2 summarizes several important phenomena. First of all, the Raman scattering intensity due to C–H bonds is comparable to that due to O–H bonds. In contrast the MIR absorption is essentially due to water molecules so that the differences between the MIR spectra of the two wine samples are very weak (see Fig. S1 in supplementary information). Second, the spectra recorded with 325 nm excitation differ significantly from those obtained with 532 and 785 nm excitations, through the appearance of new lines, probably because of resonance Raman scattering. In particular the peaks around 1600 cm 1 for 325 nm excitation are very intense. There are several methods to remove the fluorescence background and we used a fit by a polynomial in the present work. The fluorescence background may be easily estimated as it is quite close to a straight line providing that the spectrum is analyzed over a sufficiently narrow spectral window. Raman spectra corrected for fluorescence of the wine samples at the three excitation wavelengths are displayed within the 800–2000 cm 1 range in Fig. 3. As mentioned above, the spectra recorded with 325 nm excitation differ significantly compared to those recorded using 532 and 785 nm excitation wavelengths. Whereas the Raman spectra at 532 and 785 nm are almost identical in Fig. 3, the most spectacular change at 325 nm is the evidence of two strong lines around 1600 cm 1 and several weaker ones at lower wavenumbers. Clearly these new enhanced Raman signatures are due to resonance Raman scattering that may be assigned to molecular species absorbing at 325 nm. Remembering first that the absorption maxima at 325 nm in Fig. 1a and second that the maxima of fluorescence around 440 nm (Fig. 1b) after excitation at 325 nm may be assigned to hydroxycinnamic acids, it is very likely that this family of molecules is at the origin of the resonance Raman scattering observed on the spectra shown in Figs. 2 and 3.
Raman intensity
238
(c) (b) (a)
800
1 000
1 200
1 400
1 600
Raman shift (cm-¹)
Fig. 3. Zoom of the Raman spectra of Fig. 2 corrected for fluorescence of (a) the dry wine sample #1 and (b) the medium wine sample #2 with 325 nm (top), 532 nm (middle) and 785 nm (bottom) excitation wavelengths.
400
500
600
700
800
Raman shift (cm-¹)
Fig. 5. Raman spectra (785 nm excitation) of (a) the dry white wine sample #1, (b) the 40 mg l 1 glucose model solution, (c) the 40 mg l 1 fructose model solution and (d) the medium wine sample #2.
C. Martin et al. / Food Chemistry 181 (2015) 235–240
solutions (40 g l 1) are also shown in Fig. 5 (assignments of Raman peaks of sugars may be found in literature (Mathlouthi & Luu, 1980; Mathlouthi & Vinh Luu, 1980; Vasko, Blackwell, & Koenig, 1972)). Clearly there is no sugar in the dry white wine sample and a quite large excess of fructose relative to glucose in the medium wine. This comparison is quite spectacular and in agreement with the expected content of sugar in a medium white wine (
Contents lists available at ScienceDirect
Food Chemistry journal homepage: www.elsevier.com/locate/foodchem
Analytical Methods
Raman spectroscopy of white wines Coralie Martin a,b, Jean-Luc Bruneel a, François Guyon c, Bernard Médina c, Michael Jourdes d,e, Pierre-Louis Teissedre d,e, François Guillaume a,⇑ a
Université de Bordeaux, CNRS, ISM UMR 5255, 351 cours de la Libération, F-33405 Talence Cedex, France Advanced Track and Trace, ATT 99 avenue de la châtaigneraie, F-92500 Rueil Malmaison, France c Service Commun des Laboratoires, 3 Avenue du Dr A. Schweitzer, F-33608 Pessac, France d Université de Bordeaux, ISVV, EA4577 Unité de Recherche Oenologie, F-33140 Villenave D’ornon, France e INRA, ISVV, USC 1366 Œnologie, F-33140 Villenave d’Ornon, France b
a r t i c l e
i n f o
Article history: Received 24 July 2014 Received in revised form 15 January 2015 Accepted 14 February 2015 Available online 20 February 2015 Keywords: Raman spectroscopy Fluorescence spectroscopy White wines Phenolic composition Hydroxycinnamic acids Sugars
a b s t r a c t The feasibility of exploiting Raman scattering to analyze white wines has been investigated using 3 different wavelengths of the incoming laser radiation in the near-UV (325 nm), visible (532 nm) and near infrared (785 nm). To help in the interpretation of the Raman spectra, the absorption properties in the UV–visible range of two wine samples as well as their laser induced fluorescence have also been investigated. Thanks to the strong intensity enhancement of the Raman scattered light due to electronic resonance with 325 nm laser excitation, hydroxycinnamic acids may be detected and analyzed selectively. Fructose and glucose may also be easily detected below ca. 1000 cm 1. This feasibility study demonstrates the potential of the Raman spectroscopic technique for the analysis of white wines. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction Analytical methods are essential tools for wine quality control and authentication. Among all spectroscopy techniques currently used in this context, often complementary, there is a great need for low cost analytical tools that are small and light enough to be handled for field analysis. Optical methods based on the phenomenon of light absorption have experienced significant developments in recent years for the characterization of wines. These methods encompass absorption spectroscopy in the mid-infrared (MIR) and the near-infrared (NIR) for studying fundamental molecular vibrations and their harmonics (Bauer et al., 2008; Cozzolino, Dambergs, Janik, Cynkar, & Gishen, 2006; Cozzolino, McCarthy, & Bartowsky, 2012), absorption spectroscopy in the ultra-violet and visible (UV–vis) for probing electronic transitions (Acevedo, Jiménez, Maldonado, Domínguez, & Narváez, 2007; García-Jares & Médina, 1995; Harbertson & Spayd, 2006; Roig & Thomas, 2003; Urbano, Luque de Castro, Pérez, García-Olmo, & Gómez-Nieto, 2006). These techniques are well suited in an industrial context due to their ease of use, their measurement quickness, their relatively low financial cost and also because they can be small enough (miniaturized in a near future) for in situ operation. ⇑ Corresponding author. Fax: +33 (0)5 4000 3183. E-mail address: [email protected] (F. Guillaume). http://dx.doi.org/10.1016/j.foodchem.2015.02.076 0308-8146/Ó 2015 Elsevier Ltd. All rights reserved.
Surprisingly there has been very little research carried out on wines by means of spectroscopic techniques analyzing the emission of light. Indeed for wavelengths of light in the 260–1100 nm range, several phenomena will take place involving the electronic polarization of the molecules. Let us consider the interaction of a monochromatic electromagnetic radiation (laser) with molecules. If there is no absorption of the incoming radiation, elastic (Rayleigh) and inelastic (Raman) scattering of photons will occur. The spectral analysis of the Raman scattering provides information on molecular vibrations (Raman scattering effect is fully described in many books, see for instance (Dietzek, Cialla, Schmitt, & Popp, 2010)). If now the molecule absorbs the exciting radiation, two phenomena involving different mechanisms may take place if we exclude phosphorescence. The first one is fluorescence (Lakowicz, 2006) and the second is resonance Raman scattering (Dietzek et al., 2010). The resonance Raman spectrum will display maxima at the same positions to that of normal Raman scattering, but the vibrations coupled to the absorbing functional groups may be strongly intensified. To the best of our knowledge, only one paper has been published so far about Raman scattering of white wines (Meneghini et al., 2008) and none about resonance Raman scattering. A very few studies based on front face fluorescence spectroscopy for direct and global analyzes of wines have already been published (Airado-Rodríguez, Durán-Merás, Galeano-Díaz, & Wold, 2011; Dufour, Letort, Laguet, Lebecque, & Serra, 2006; Le Moigne et al.,
236
C. Martin et al. / Food Chemistry 181 (2015) 235–240
2008). However, fluorescence detection coupled to HPLC system have been extensively used to quantify flavanols (e.g. catechin, epicatechin. . .) and procyanidin dimer (e.g. procyanidin (GómezAlonso, García-Romero, & Hermosín-Gutiérrez, 2007; Silva, Ky, Jourdes, & Teissedre, 2012)). Molecular vibrations may be analyzed either in the MIR–NIR regions by probing the absorption of light by molecules, or by Raman and resonance Raman scattering of light in the visible range. Because the mechanisms of light-matter interaction are not the same for absorption and scattering, MIR–NIR and Raman spectroscopic techniques are complementary with distinct vibration selection rules. We will show that, for this reason, Raman spectroscopy possesses advantages for wine analysis. Also very small amount of sample is required (several lm3) as a Raman spectrometer may be coupled to a confocal microscope (a quite large majority of commercial Raman spectrometers combine Raman spectroscopy to optical microscopy). This method is a very effective tool in chemical analysis because it is non-destructive and usually does not require special preparation of the sample. Finally Raman spectrometers may be very compact and equipped with optical fibers allowing in situ measurements. For example mobile Raman spectrometers are efficient tools in the domains of arts and archeology (Colomban, 2012). The aim of this work is to investigate the potential of Raman spectroscopy for analyzing commercial wines. This preliminary study focuses on two samples of white wines (one dry and one medium) for establishing the potential of the technique. This study focus on white wines as their chemical composition is less complex than red wine one’s. Because the mechanisms at the origin of light emission depend on the absorption of the exciting radiation, we will first investigate the absorption of light in the UV–visible for the two samples of white wines. Then the emission spectra recorded using three wavelengths from the near-UV to the NIR for the laser excitation will be analyzed. As mentioned above, spectra where fluorescence emission and resonance Raman or normal Raman scattering take place are expected. 2. Materials and methods 2.1. Samples The white wines, one dry and one medium, that were investigated in this work originate from south-west of France. These two wines were chosen as visually their yellow color looked very similar and because their chemical composition must be significantly different. The first sample (sample #1) is a Bordeaux dry wine AOC ‘‘entre deux mers’’ vintage 2013 and the second (sample #2) sample is a medium wine AOC ‘‘côtes de Bergerac’’ vintage 2012. To help in the interpretation of the wine’s spectra, those of pure phenolic compounds or sugars dissolved in a wine model solution have been recorded. The used wine model solution was a hydroalcoholic solution (e.g. 12% ethanol) acidified with 5 g l 1 of tartaric acid with the pH adjusted at 3.5 (e.g. 3.5 is considered as the average wine acidity (Ribéreau-Gayon, Dubourdieu, et al., 2012; Ribéreau-Gayon, Glories, et al., 2012)) with a solution of NaOH (1 M). Phenolic acids and sugars with purity >99% were obtained commercially from Sigma–Aldrich. 2.2. Absorption spectroscopy UV–visible spectroscopy measurements were performed using a Lambda-650 UV–vis spectrophotometer (Perkin Elmer). Wine samples taken from freshly opened bottles were scanned in transmission mode (200–900 nm). Samples of white wines were placed in quartz cells of thickness 1 mm.
2.3. Emission spectroscopy All phenomena involving emission of light such as fluorescence, Raman scattering and resonance Raman scattering were recorded in the backscattering geometry by means of Raman spectrometers coupled to optical microscopes. Experiments using the 325 nm excitation wavelength of a He–Cd Laser were performed on the Raman spectrometer Labram-UV HR800 (Horiba Jobin Yvon) using a 4 UV-lens, experiments at 532 (frequency doubled Nd:YAG laser) and 785 nm (diode) were performed on the Xplora instrument (Horiba Jobin Yvon) using a 10 objective. The laser power at sample was 2 mW for 325 nm excitation wavelength, 13 mW for 532 nm and 50 mW for 785 nm. To analyze the emission of light at 325 nm excitation over a broad range of wavelengths, i.e. to analyze fluorescence, a 150 lines mm 1 diffraction grating was used. For Raman scattering a better spectral resolution in the range 2–5 cm 1 was needed and gratings with 2400 lines mm 1 for a 325 nm excitation wavelength, 1800 lines mm 1 at 532 nm and 1200 lines mm 1 at 785 nm were used. Samples of white wines and model solutions were placed in NMR glass tubes with overall diameter 4.97 mm and internal diameter 4.20 mm. The emission spectra were corrected for detector efficiency and could be therefore qualitatively compared to each other. However quantitative information on the absolute intensities of the emitted light cannot be provided in this study.
3. Results and discussion A difficulty to be overcome is related to the chemical nature of the wines with composition of about 12% of ethanol and 84% water. The other molecules include carboxylic acids (for example tartaric acid), sugars, glycerol and also polyphenols that represent a very small proportion of the total composition. As emphasized in the introduction of this paper, optical spectroscopy has been widely exploited for the analysis of wines and it is obvious that the obtained spectra will result from the superposition of all the optical responses of different molecules that make up the medium. A very large number of molecular species have been identified in wines, and two of these species alone represent about 96% of the total number of molecules. One may therefore easily understand that Raman signals are expected to be dominated by those of water and ethanol.
3.1. UV–visible absorption spectra In wines, mainly polyphenols will absorb light at wavelengths between 250 and 900 nm. The absorption features of these species are now well identified (Cerovic et al., 2002; Jurd, 1957). The UV– visible absorption spectra between 250 and 500 nm of the two samples of white wines are shown in Fig. 1a. There is no significant absorption around 500 nm, as expected for a white wine that does not contain anthocyanins known to have absorbance features at 267–275 and 475–545 nm. Hydroxycinnamic acids have absorbance maxima at 227–245 and 310–332 nm; benzoic acids show a single absorbance in the region of 235–305 nm and flavanols typically have maxima in the 250–270 and 350–390 nm regions. As white wines do not contain significant amounts of flavanols, the peaks around 263 nm may be therefore assigned mainly to the phenolic acids that are present in the white wines and the peak around 326 nm mainly to hydroxycinnamic acids. Of course these peaks are broad and overlap around 300 nm. Interestingly the two white wines have quite similar absorptions for wavelengths above 300 nm and the very weak tails of the spectra above 400 nm are responsible for their similar yellow color. In contrast,
237
C. Martin et al. / Food Chemistry 181 (2015) 235–240
(a)
263 nm
3.5
2.5 2.0 1.5 1.0
120 000
Intensity (counts/s)
Absorbance
3.0
273 nm
0.5
100 000 80 000
443 nm
60 000
437 nm
40 000
326 nm
300
(b)
438 nm
140 000
20 000
400 Wavelength (nm)
400
500
450
500 Wavelength (nm)
550
600
Fig. 1. (a) UV–visible absorption spectra (optical path 1 mm) of sample #1 (maximum at 273 nm) and sample #2 (263 nm). (b) Laser induced fluorescence spectra with 325 nm excitation of sample #1 (443 nm), sample #2 (438 nm) and caffeic acid model solution (437 nm). For convenience, the intensity of the spectrum of caffeic acid solution (200 mg l 1) has been multiplied by a factor of 10.
Ethanol CC stretching
O-H stretching
Raman Intensity
(b)
Raman Intensity
(a)
C-H stretching
1 000
2 000 Raman shift (cm-¹)
3 000
1 000
2 000
3 000
Raman shift (cm-¹)
Fig. 2. Raman spectra (including fluorescence) of (a) the dry white wine sample #1 and (b) the medium white wine sample #2 for three wavelengths of the excitation laser source: 325 nm (top), 532 nm (middle) and 785 nm (bottom). For comparing the spectra, the intensities were normalized versus that of the band assigned to the C–C stretching vibration of ethanol near 880 cm 1.
the absorption properties below 300 nm are totally different suggesting significant differences in their chemical composition.
could not be reached from the analysis of the UV–visible absorption spectra displayed in Fig. 1a where absorbance of the two wine samples around 325 nm is similar.
3.2. Laser induced fluorescence spectra 3.3. Raman and resonance Raman spectra The next step of the analysis is to investigate qualitatively fluorescence of the wine samples. The emission spectra of the wines shown in Fig. 1b were recorded with an excitation wavelength of 325 nm, i.e. at the maximum of absorption of hydroxycinnamic acids (Fig. 1a). The maxima around 440 nm are due to fluorescence emission of the white wines. These maxima of fluorescence are congruent with those known for hydroxycinnamic acids (Airado-Rodríguez et al., 2011; Promkatkaew et al., 2014; Putschögl, Zirak, & Penzkofer, 2008). As a matter of fact the maximum of fluorescence of caffeic acid (representative of hydroxycinnamic acids) in wine model solution is also around 440 nm (Fig. 1b) and the model solution of gallic acid (representative of hydroxybenzoic acids) does not display significant fluorescence with 325 nm excitation. With the 532 excitation wavelength, a very weak fluorescence is detected compared to that displayed in Fig. 1b, suggesting trace concentration of absorbing molecules belonging perhaps to the family of flavonoids. At 785 nm there is no detectable fluorescence. We may conclude that by exciting samples #1 and #2 with a laser with 325 nm wavelength, part of the incoming light is absorbed by hydroxycinnamic acids and related compounds giving rise to fluorescence with maxima around 440 nm. It should be pointed out that the intensity at the maximum of the fluorescence differs significantly for the two wines, suggesting that their composition in hydroxycinnamic acids is probably different. Such conclusion
To introduce the discussion on the Raman scattering of white wines, the emission spectra of the dry white wine sample #1 within the 200–3600 cm 1 spectral window are shown in Fig. 2a and those of the medium white wine sample #2 in Fig. 2b. In Fig. 2a are also shown the main Raman peaks assigned to ethanol and water molecules. Note however that polyphenols, sugars and other molecules will also contribute, but very weakly, to the intensity of the bands assigned to C–H and O–H stretching vibrations between 2800 and 3600 cm 1. Obviously the ethanol concentration may be easily evaluated from the ratio of the integrated intensities of the bands assigned to the C–H and O–H stretching vibrations using a calibration procedure. As mentioned in Section 1, we found in literature only one paper dealing with Raman scattering of white wines. The paper by (Meneghini et al., 2008) investigated the Raman scattering (with 532 nm excitation) of two samples of white wines restricting their analysis to the region of C–H stretching vibrations between 2600 and 3100 cm 1 to estimate, via a calibration procedure on the intensities, the ethanol and sucrose concentrations. For an excitation with wavelength 325 nm, the strong fluorescence background due to the absorption of hydroxycinnamic acids and related compounds already discussed above is observed in Fig. 2. However, as the maximum of fluorescence is around 432 nm (Fig. 1b), corresponding to a Raman shift of 7600 cm 1,
C. Martin et al. / Food Chemistry 181 (2015) 235–240
(d) (c) (b)
1631
1606
To strengthen this interpretation, the Raman spectra at 325 and 532 nm of model solutions of the two representative members of phenolic acids, gallic acid and caffeic acid, have been recorded (shown in Fig. 4). It comes out that, when comparing Figs. 3 and 4, the new peaks appearing for white wines using the 325 nm laser excitation may be safely assigned to hydroxycinnamic acids. We note however that the Raman spectra of gallic acid at 325 and 532 nm (Fig. 4) are not strictly identical around 1600 cm 1. Presumably gallic acid absorbs weakly light at 325 nm so that the intensity of some Raman bands may be weakly enhanced due to pre-resonance effects. Although these Raman measurements are not yet quantitative, we believe that the great potential of Raman spectroscopy to detect and measure hydroxycinnamic acids in wines is now established. It is also certainly possible to identify the main species of cinnamic acids (caffeic, p-coumaric, ferulic, etc.) present in the wine samples because their resonant Raman spectra should in principle not be identical. For example the shape and intensity of the peaks around 1600 cm 1 in Fig. 3 for the dry and the medium wine samples are significantly different. The detailed analysis of UV–visible absorption, fluorescence and Raman scattering of hydroxycinnamic acids combining experimental and time-dependent density functional theory (TD-DFT) approaches will be published in a near future. Other differences between the Raman spectra of samples #1 and #2 are observed below 1000 cm 1. Indeed, several lines of weak intensity are detected in the spectra of the medium wine and this regardless of the wavelength of excitation, which are not visible in the spectra of the dry white wine. In Fig. 5 are displayed the Raman spectra of samples #1 and #2 below 1000 cm 1 with 532 nm excitation. Indeed an excess of sugar in the medium wine compared to that in the dry one is expected. For comparison the Raman spectra of fructose and glucose model
(a) 800
1 000
1 200
1 400
1 600
Raman shift (cm-¹)
1 200
1 400
Raman shift (cm-¹)
1 600
1637
1 000
(b)
1607
Fig. 4. Raman spectra of gallic acid model solution recorded with excitations at 532 nm (a) and 325 nm (c), caffeic acid model solution at 532 nm (b) and 325 nm (d).
(d) Raman Intensity
Raman intensity
800
1173
(a)
Raman intensity
879
the fluorescence within the Raman spectral window (below ca. 3600 cm 1) is even weaker, allowing the analysis of the Raman peaks. For the 532 nm excitation wavelength, the fluorescence background is also weak allowing the analysis of the Raman scattering over the full spectral range, up to 4000 cm 1. For the 785 nm excitation wavelength, no fluorescence is observed but the analysis of the Raman scattering is limited as the detector of the spectrometer is not efficient for wavelengths larger than 980 nm (Raman shift 2500 cm 1). Fig. 2 summarizes several important phenomena. First of all, the Raman scattering intensity due to C–H bonds is comparable to that due to O–H bonds. In contrast the MIR absorption is essentially due to water molecules so that the differences between the MIR spectra of the two wine samples are very weak (see Fig. S1 in supplementary information). Second, the spectra recorded with 325 nm excitation differ significantly from those obtained with 532 and 785 nm excitations, through the appearance of new lines, probably because of resonance Raman scattering. In particular the peaks around 1600 cm 1 for 325 nm excitation are very intense. There are several methods to remove the fluorescence background and we used a fit by a polynomial in the present work. The fluorescence background may be easily estimated as it is quite close to a straight line providing that the spectrum is analyzed over a sufficiently narrow spectral window. Raman spectra corrected for fluorescence of the wine samples at the three excitation wavelengths are displayed within the 800–2000 cm 1 range in Fig. 3. As mentioned above, the spectra recorded with 325 nm excitation differ significantly compared to those recorded using 532 and 785 nm excitation wavelengths. Whereas the Raman spectra at 532 and 785 nm are almost identical in Fig. 3, the most spectacular change at 325 nm is the evidence of two strong lines around 1600 cm 1 and several weaker ones at lower wavenumbers. Clearly these new enhanced Raman signatures are due to resonance Raman scattering that may be assigned to molecular species absorbing at 325 nm. Remembering first that the absorption maxima at 325 nm in Fig. 1a and second that the maxima of fluorescence around 440 nm (Fig. 1b) after excitation at 325 nm may be assigned to hydroxycinnamic acids, it is very likely that this family of molecules is at the origin of the resonance Raman scattering observed on the spectra shown in Figs. 2 and 3.
Raman intensity
238
(c) (b) (a)
800
1 000
1 200
1 400
1 600
Raman shift (cm-¹)
Fig. 3. Zoom of the Raman spectra of Fig. 2 corrected for fluorescence of (a) the dry wine sample #1 and (b) the medium wine sample #2 with 325 nm (top), 532 nm (middle) and 785 nm (bottom) excitation wavelengths.
400
500
600
700
800
Raman shift (cm-¹)
Fig. 5. Raman spectra (785 nm excitation) of (a) the dry white wine sample #1, (b) the 40 mg l 1 glucose model solution, (c) the 40 mg l 1 fructose model solution and (d) the medium wine sample #2.
C. Martin et al. / Food Chemistry 181 (2015) 235–240
solutions (40 g l 1) are also shown in Fig. 5 (assignments of Raman peaks of sugars may be found in literature (Mathlouthi & Luu, 1980; Mathlouthi & Vinh Luu, 1980; Vasko, Blackwell, & Koenig, 1972)). Clearly there is no sugar in the dry white wine sample and a quite large excess of fructose relative to glucose in the medium wine. This comparison is quite spectacular and in agreement with the expected content of sugar in a medium white wine (
