Rapid

lable at ScienceDirect

Food Control 48 (2015) 56e61

Contents lists avai

Food Control

journal homepage: www.elsevier .com/locate/ foodcont

Rapid analysis of glucose, fructose and sucrose contents of commercialsoft drinks using Raman spectroscopy

Kerem Ilaslan a, Ismail Hakki Boyaci a,b,*, Ali Topcu a,b

aHacettepe University, Faculty of Engineering, Department of Food Engineering, Beytepe Campus, 06800 Ankara, Turkeyb Food Research Center, Hacettepe University, Beytepe, 06800 Ankara, Turkey

a r t i c l e i n f o

Article history:Received 19 August 2013Received in revised form2 January 2014Accepted 2 January 2014Available online 11 January 2014

Keywords:Raman spectroscopyHPLCGlucoseSucroseFructosePLS

* Corresponding author. Food Research Center, H06800 Ankara, Turkey. Tel.: þ90 312 297 61 46; fax:

E-mail address: [email protected] (I.H. Boyaci)

0956-7135/$ e see front matter � 2014 Elsevier Ltd.http://dx.doi.org/10.1016/j.foodcont.2014.01.001

a b s t r a c t

The objective of this study was to quantify glucose, fructose, and sucrose in commercial soft drinks byRaman spectroscopy as a fast and low-cost technique. For the calibration, dilutions in the range of 0e12%(w/w) were prepared in water for each of glucose, fructose, and sucrose. The Raman spectrum for eachdilution was obtained. Calibration models were formed and curves were plotted by using the fullspectrum of Raman data. The partial least squares (PLS) regression method was used to carry out thespectroscopic data analysis. The contents of the sugars in the soft drinks were predicted depending uponthe calibration models by PLS. The slope of regression values of glucose, fructose, and sucrose were 0.967,0.992, and 1.008 and the coefficient of determination (R2) values were 0.913, 0.998 and 0.993 for vali-dation, respectively. A high-performance liquid chromatography (HPLC) method was used to verify theefficiency of the Raman method. The coefficient of determination values between the HPLC and thepredicted values of glucose, fructose and sucrose were determined as 0.913, 0.968 and 0.910, respectively.The results of this work provide a rapid method for evaluating the quantitative analysis of glucose,fructose, and sucrose in soft drinks.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Carbohydrates are main sources of energy for essential meta-bolic processes and are important constituents of cellular sub-stances. Aside from the beneficial features of carbohydrates, excessconsumption of carbohydrates causes health problems (Moros,Inon, Garrigues, & de la Guardia, 2005). Taking into account thisfact, carbohydrate values, which are stated on the label of softdrinks, increase in importance (Amendola, Iannilli, Restuccia,Santini, & Vinci, 2004).

Carbohydrates (mono-, di and polysaccharides) are one of themajor class of organic compounds found in nature (Zhbankov,Andrianov, & Marchewka, 1997). A number of analytical tech-niques, including high-performance liquid chromatography (HPLC)and isotopic methods have been used for specifying the structureand properties of carbohydrates (Gestal, et al., 2004). HPLC is acommon method used for analysing glucose in hydrolysates.Chromatographic methods are inherently low-throughput tech-niques that require tedious sample preparation and a relatively

acettepe University, Beytepe,þ90 312 299 21 23..

All rights reserved.

long analysis time for each analysis. Enzymatic methods for sugaranalysis are specific, rapid, and reproducible; nevertheless theyrequire single determinations for each compound, which is time-consuming and expensive. In contrast, vibrational methods arenon-destructive, easy to apply, rapid, and do not require samplepre-treatment. Raman, mid-infared (MIR), and near-infrared (NIR)spectrometry are novel and useful alternatives to the classicalmethods mentioned above. (Gestal et al., 2004; Rodriguez-Saona,Fry, McLaughlin, & Calvey, 2001; Shih & Smith, 2009).

In recent years, Raman spectroscopy, a branch of vibrationalspectroscopy, has risen in importance (Das & Agrawal, 2011).Raman spectroscopy is a powerful tool for investigating the struc-ture of mono- and disaccharides and is increasingly used in biology.The Raman affect is described as inelastic scattering of incidentlight from a sample and frequency shift by energy of its charac-teristic molecular vibrations. The scattered light is collected,dispersed in a monochromator and then detected by a sensor. Thevibrational spectrum has the complimentary information as theinfrared (IR) vibrational spectrum, but they have a few importantdistinctions. Raman scattering arises from the differences in thepolarisability or shape of the electron distribution in the moleculeas it vibrates, while infrared absorption requires a change of theintrinsic dipole moment with the molecular vibration. Water has aminor scattering effect and does not interfere with Raman

K. Ilaslan et al. / Food Control 48 (2015) 56e61 57

scattering from solutes in aqueous solutions such as soft drinks. Forthese reasons, Raman spectroscopy is a suitable and reliablemethod for quantitative analysis of soft drinks (Arboleda &Loppnow, 2000; Beaten & Aparicio, 2000; Kneipp, Kneipp, Itzkan,Dasari, & Feld, 1999).

Data-mining techniques such as chemometrics were used toprocess data of raw spectra that were extracted from the vibrationalmethods. The term “chemometrics” is generally used to indicatethe use of multivariatemathematical techniques and/or statistics toderive chemical information from data. These techniques are usedto display the differences between apparently similar spectra byusing the correlation between the collected data and changes in thevariables of interest. Raw data acquired from Raman spectroscopyneeds to be pre-processed before processing data with chemo-metric methods. Pre-processing (i.e. signal processing) techniquesused on data to make an analysis more stable and/or accurateinclude interrogation of simple band metrics (width, position, area,etc.) and least-squares analyses for calibration curves and kineticstudies (Ozbalci, Boyaci, Topcu, Kadilar, & Tamer, 2013; Shaver,2001). Partial least-squares regression is one of the mostcommonly used chemometric methods. A large amount of collecteddata on a spectrum can be reduced to principle components thatreflect the variance. The choice of proper spectral range and thenumber of variables employed in the model are the main factorsthat underlie the success of this method (Ozbalci et al., 2013).

In this work, we have focused on carbonated beverages in orderto measure the amount of carbohydrates: glucose, fructose andsucrose by using HPLC and Raman spectroscopy with chemometricmethods. Although, the initial costs of Raman and HPLC equip-ments were expensive, the analysis of Raman spectrometer wasrelatively cheaper than HPLC with an advantage of a short responsetime without no sample pretreatment and the additional cost ofchemicals. The developed method as a fast and economic way todetect the amount of glucose, fructose and sucrose in soft drinks byusing Raman spectroscopy and chemometric methods. The devel-oped Raman method was also validated by the conventionalmethod, i.e. HPLC.

2. Materials and methods

2.1. Chemicals and samples

High purity (99.5%) D-(þ)-glucose and D-(�)-Fructose was pro-vided by SigmaeAldrich (Steinheim, Germany). Biochemical gradeD-(þ)-sucrose was purchased from Acros Organics (Geel, Belgium).The chemicals were used without any further treatment. Seventeencommercial soft-drink samples were analysed. The soft drinksamples (fruit-flavoured mineral water, soda, orangeade andschorle (made from diluting apple juice with carbonated water)were purchased from local supermarkets in Turkey.

2.2. Instrumentation

The Raman measurements were carried out by a DeltaNuExaminer Raman microscope (DeltaNu Inc., Laramie, WY, USA),with a 785 nm laser source, motorized microscope with a stagesample holder and a CCD detector. Each sample was manipulatedwith the built-in “automatic baseline correction” functions of thesoftware.

The HPLC measurements were carried out by SpectraSystemHPLC (ThermoFinnigan Inc., CA, USA) integrated with an auto-sampler including temperature control for the column (Spec-traSystem AS3000), a degasser system (SpectraSystem SCM1000), aquaternary gradient pump (SpectraSystem P4000) and a refractiveindex detector (Shodex RI-101, Showa Denko, NY, USA). A personal

computer with a software package for system control and dataacquisition (ChromQuest 4.2.34) was used for the HPLC analyses.

2.3. Sample preparation and sugar analysis

In the Raman experiment, three different concentrations in therange of 0e10% (w/w) of glucose, fructose, and sucrose wereprepared for the calibration of sugars under the room conditions.Each set included 25 different dilutions. The ranges of sugarcontents were determined from the carbohydrate values on thelabel of soft drinks. The soft drink samples were degassed in anultrasonic bath for 15 min. After the degassing process, the soft-drink samples were analysed by Raman spectroscopy withoutany dilution. The soft drink samples contain some additives suchas colour, aroma, citric acid etc. beside the sugars. To demonstrateif any of these additives cause a matrix effect on the spectra of thesoft drink samples, plain mineral water samples with combina-tions of colour, aroma, and citric acid were also prepared andanalyzed by Raman spectroscopy at the same conditions. Theparameters of the Raman instrument were as follows: 75 mWlaser power, 30 s acquisition time, and 200 ml sample volume in aglass Raman cuvette.

The soft-drink samples were also evaluated by SpectraSystemHPLC (ThermoFinnigan Inc., CA, USA) for determining sugar con-tents. Before the HPLC experiment, the soft-drink samples werediluted 20 times by mass to reach the desired concentration forHPLC. The analysis of glucose, fructose, and sucrose were per-formed by HPLC with a refractive index detector (Shodex RI-101,Showa Denko, NY, USA) isocratically at 0.6 ml/min flow rate at80 �C with a 300� 7.8 mm i.d. cation exchange column (Rezex RCMcolumn, Ca2þ, 8 mm, Torrance, CA). For the HPLC column, deionisedwater was used as the mobile phase and injection volume was20 ml. The external standard method was used for quantification ofsugars. The experimental processes are shown in Fig. 1.

2.4. Chemometric data analysis

The full range (200e2000 cm�1) of the Raman spectra was usedto form calibration and validation data sets of the PLS models.Before forming the model, the data that was collected from Ramanmeasurements were pre-processed by baseline, first derivative,second derivative, normalization, smoothing, and mean centrefunctions to achieve succeeding models. Among all the pre-processing operations applied, the baseline function was deter-mined as suitable for our data set. From the 25 dilutions, 20 of themwere separated for calibration and five of them (20% of all samples)were used for validation. The Raman data were loaded to the xcomponent section and the concentrations were loaded to the ycomponent section for both calibration and validation. Venetianblinds were chosen as a cross-validation method. This method wassimple and easy to implement, and generally safe to use if therewere relatively many objects in a random order. The calibrationcurves were plotted. Root-Mean-Square Error of Cross-Validation(RMSECV), Root-Mean-Square Error of Prediction (RMSEP), slopeof regression and coefficient of determination (R2) values werecalculated. The suitable latent variables (LVs) were determinedfrom the correlation between the number of latent variables withRMSECV values in the PLS calibration model. The Raman spectro-scopic data of the collected soft-drink samples was loaded to the xcomponent section of validation for prediction. The PLS analyseswere carried out by the PLS toolbox (Eigenvector Research, Inc.,Wenatchee,Washington, USA) of MATLab R2010a (TheMathWorks,Inc., Natick, Massachusetts, USA). The accuracy and precision of themethod were determined. The student’s t-test was used on HPLCand Raman data.

Fig. 1. A scheme of the experimental processes; (a) calibration and (b) soft drink experiments.

K. Ilaslan et al. / Food Control 48 (2015) 56e6158

3. Results and discussion

3.1. Raman spectra of samples

Raman spectra of commercial soft drinks (mineral water,schlore, orangeade and soda) and Aqueous solutions of glucose,fructose and sucrose were collected and are shown in Fig. 2.Glucose, fructose, and sucrose have strong bands that reflect thespecific fingerprint spectra. The disaccharide sucrose is consisted ofglucose and fructose molecules so the Raman peaks are very closeto these monosaccharides but a little shifted. Structural changes ofthe two monosaccharides in the sucrose molecule are the mainreason of the changes in the peaks (Soderholm, Roos, Meinander, &Hotokka, 1999).

3.1.1. GlucoseIn the range of 200e500 cm�1, d(CeCeC), d(CeCeO), d(CeO)

and s(CeC) were the main skeletal vibrational motions of glucose(Korolevich, Zhbankov, & Sivchik, 1990; Mathlouthi & Luu, 1980;

Fig. 2. The spectra of sugars and soft drinks; (a) schlore, (b) mineral water, (c) or-angeade and (d) soda.

Wells & Atalla, 1990). The bands at 415 and 437 cm�1 were d(C2eC1eO1) bending vibration in a- and b-glucose, respectively. Astrong band at 523 cm�1 showed the skeletal vibration of glucose(Mathlouthi & Luu, 1980). The bands at 838 and 856 cm�1 wereassigned to the n(CeC), and d(C1eH1) vibrations (Goodacre,Radovic, & Anklam, 2002; Mathlouthi & Luu, 1980). Between thebands of 820 and 950 cm�1, n(CeO), d(CeCeH), n(CeC) and d(CeCeO) vibrations were specific to glucose and fructose (Soderholmet al., 1999). The 1106 cm�1 band was assigned to the d(CeOeC)angle-bending model (Kacurakova & Mathlouthi, 1996).

3.1.2. FructoseThe bands at 314 and 353 cm�1 were assigned to the d(CeCeC)

ring vibration in the pyranoid and furanoid forms of fructose(Mathlouthi & Luu, 1980). The strong band at 631 cm�1 was relatedto ring deformation. The band at 709 cm�1 was related to theskeletal vibration of fructose (Tipson, 1987). The band at 800 cm�1

in fructopyranose could be attributed to the n(CeC) vibration. Thestrong band at 870 cm�1 was related to CeOeC cyclic alkyl ethers.The bands at 1028 and 1054 cm�1 could be assigned to the n(CeO)vibration in the pyranoid and furanoid rings. The band at 1074 cm�1

was CeOeC cyclic alkyl ethers (Goodacre et al., 2002; Mathlouthi &Luu, 1980).

3.1.3. SucroseThe band at 419 cm�1 was related to the d(CeCeO) ring vibra-

tion in the pyranoid ring of fructose and dominated the spectralrange between 300 and 500 cm�1 (Hineno,1977; Mathlouthi & Luu,1980). The band at 544 cm�1 originated from the a-glycosidic bondof C1 on the glucosyl subunit. The two bands at 744 and 800 cm�1

were related to the n(CeC) vibration of fructopyranose and fruc-tofuranose, respectively. These bands were shifted as a result of theeffect of glycosidic bond with a glucose ring (Mathlouthi & Luu,1980). The band at 1127 cm�1 originated from CeOH deformation(Goodacre et al., 2002).

3.1.4. Soft drinkIn this study, the main peaks of glucose, fructose, and sucrose

have seen on the spectra of the soft drinks. As described in Fig. 2,the intense bands at 627, 707, 870, 819 and 1071 cm�1 in the softdrinks were related to the fructose molecule. The bands at 517, 849,913 and 1124 cm�1 originated from the glucose molecule. The

Fig. 3. The comparison of the actual and predicted sugar content of aqueous solutions. The calibration data: (a) glucose, (b) fructose and (c) sucrose; validation data: (d) glucose, (e)fructose and (f) sucrose.

K. Ilaslan et al. / Food Control 48 (2015) 56e61 59

bands at 745 and 796 cm�1 were specific for the sucrose molecule.The sugar content of the soft drinks could be easily followed on thespecific peaks of the Raman spectroscopy.

The soft drink samples may contain additives such as colour,aroma, andcitric acid etc. besides thesugars. Therefore, thespectraofthe samples (soda containing no sugar with combinations of colour,aroma, and citric acid) were obtained to comparewith the spectra ofaromatized sodawith sugar to show if these additives cause amatrix

effect on the spectra of the soft drink samples. (Fig. S3). As shown inFig. S3, the main peaks of the aroma in soda with no sugar were 761and 798 cm�1; the citric acid was 752 and 900 cm�1. The food col-ouring showed no significant peak in this state of Raman spec-trometer. When we looked at the spectrum of aromatized soda, themain peaks of the aroma and citric acid had no distinct peak on thespectra of soft drinks. It was deduced from these results, the matrixeffect of the soft drinks could be eliminated in this study.

Fig. 4. A comparison of the results of HPLC and predicted PLS values for real samples. (a) glucose, (b) fructose and (c) sucrose.

K. Ilaslan et al. / Food Control 48 (2015) 56e6160

3.2. Data analysis

In the first step of the work, the full spectra of glucose, fructoseand sucrose were obtained from the Raman spectroscopy. In orderto analyse the acquired Raman spectral data, we needed a moresophisticated technique, such as chemometric methods, becausethe Raman spectra included very extensive data and the dissimi-larity of the Raman peaks were not distinct. For these reasons, PLSregression was applied. Some of the pre-processing operationswere applied to the raw Raman spectral data. First and second or-der derivatives such as mean centre, smoothing and baselinefunctions were tried individually or simultaneously. The baselinefunction gave us the most accurate results, so this function wasapplied as a pre-processing treatment. The coefficient of determi-nation and the slope of regression equation of the calibrationgraphs were higher after the baseline function than the other pre-processing operations. In Fig. S2, the raw data of Raman spectros-copy were plotted after some of the pre-processing operationsgiven as baseline, first derivative, second derivative, normalization,smoothing and mean centre.

RMSECV and RMSEP values were indicators that helped us tochoose the best fitting latent variables. For glucose, fructose andsucrose, separately the LVs were 13, 11 and 12. The RMSECV valuesof glucose, fructose and sucrose were 0.290, 0.179 and 0.512, andRMSEP values were 0.363, 0.162 and 0.303, respectively. The cor-relations between the numbers of latent variables and RMSECV

values are plotted in Fig. S1a, 1b and 3c. Calibration curves of thesugar samples were plotted with the calibration and the validationdata of PLS. The slopes of regression equations of glucose, fructoseand sucrose were 0.997, 0997 and 0.998 for the calibration and0.967, 0.992 and 1.008 for the validation, respectively. The coeffi-cient of determination values of glucose, fructose and sucrose were0.999 for the calibration and 0.913, 0.998 and 0.993 for the vali-dation, respectively. The calibration curves of the sugar samples aregiven for calibration data in Fig. 3aec, and for validation data inFig. 3def. The prediction results of the validation were comparableto the measured values, and thus the R2 values were high enough,as expected.

In the second step, the 17 soft-drink samples were analysed byRaman spectroscopy and HPLC. The measured HPLC (x-axis) andthe predicted PLS (y-axis) results were plotted to achieve a linearmodel (regression equation), of which the intercept value was zero.The graphs were plotted for making a comparison between themeasured and predicted values of sugar samples. The plottedgraphs are given in Fig. 4aec. The slope of regression values ofglucose, fructose and sucrose were 0.993, 0.989 and 1.000,respectively. The coefficient of determination values of glucose,fructose and sucrose were 0.913, 0.968 and 0.910, respectively. Therelative standard deviation (RSD %) and bias were applied for theprecision and the accuracy of the method. RSD and bias of themeasured/predicted values of the method were found to be 1.077%and�0.76, respectively. The limit of detection (LOD) and the limit of

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17Soft Drink Samples

To

ta

l S

ug

ar C

on

te

nts

(g

/1

00

ml)

The amount of carbohydrate on the labelThe results of HPLC The predicted values of Raman spectroscopy

Fig. 5. A comparison between the concentration of carbohydrate (g/100 ml) for thesoft drink samples.

K. Ilaslan et al. / Food Control 48 (2015) 56e61 61

quantification (LOQ) were determined as 0.2 g/100 ml and 0.7 g/100 ml, respectively. The soft drink samples were evaluated atdifferent days and no significant difference is found (p < 0.05) thus,the repeatability of the Raman method was high. We could inferfrom these results that the PLS methods have a high predictioncapability. Fructose had higher R2 (0.968) than both glucose (0.913)and sucrose (0.910) (Fig. 4), which was mainly caused by the spe-cific and distinct peaks of fructose when compared with glucoseand sucrose.

The time taken for Raman analysis (30 s) was lower than that ofother chemical and enzymatic methods. The pre-processing step ofsoft-drink samples did not include any chemical treatment ordestructive practices, so new samples could be analysed rapidlywith this method (Delfino et al., 2011). The amount of carbohydrate(g/100 ml) on the label was compared with the results of HPLC andthe predicted values of PLS to demonstrate the efficiency of ourmethod, which is shown as a bar graph (Fig. 5). The predictedvalues of PLS were relatively higher than the label values and theHPLC results. The Student’s t-test demonstrated the accuracy of themethod, since therewere no significant differences (p> 0.05) whencompared with the results of HPLC and the values on the labels(Fig. 5).

4. Conclusion

In this paper, our main purpose was to determine glucose, fruc-tose and sucrose content in soft drinks. The milestone of this workwas to predict the amount of sugars byPLSmethodwith the spectraldata of Raman spectroscopy and to demonstrate that the resultswere in accord with the results of the HPLC method and the carbo-hydrate amounts indicated on the labels. We observed that theRaman system with the PLS method is a faster and cheaper tech-nique and showed higher performance than HPLC system which ismore time consuming. These advantagesmake Raman spectroscopycombined with chemometric methods useful for the analysis ofsugar contents in soft drinks. In the literature, therewere fewRamanstudies about thequantitative analysis of sugars in soft drinks andnostudies for detecting more than one sugar in a single sample; hencethis study contributes to the literature in this respect.

Acknowledgements

This study was financially supported by Republic of Turkey,Ministry of Science, Industry and Technology (SAN-TEZ PROJECTNO: 01597.STZ.2012-2).

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.foodcont.2014.01.001.

References

Amendola, C., Iannilli, I., Restuccia, D., Santini, I., & Vinci, G. (2004). Multivariatestatistical analysis comparing sport and energy drinks. Innovative Food Scienceand Emerging Technologies, 5, 263e267.

Arboleda, P. H., & Loppnow, G. R. (2000). Raman spectroscopy as a discovery tool incarbohydrate chemistry. Analytical Chemistry, 72, 2093e2098.

Beaten, V., & Aparicio, R. (2000). Edible oils and fats authentication by Fouriertransform Raman spectrometry. Biotechnology, Agronomy, Society and Environ-ment, 4, 196e203.

Das, R. S., & Agrawal, Y. K. (2011). Raman spectroscopy: recent advancements,techniques and applications. Vibrational Spectroscopy, 57, 163e176.

Delfino, I., Camerlingo, C., Portaccio, M., Della Ventura, B., Mita, L., Mita, D. G., et al.(2011). Visible micro-Raman spectroscopy for determining glucose content inbeverage industry. Food Chemistry, 127(2), 735e742.

Gestal, M., Gomez-Carracedo, M. P., Andrade, J. M., Dorado, J., Fernandez, E.,Prada, D., et al. (2004). Classification of apple beverages using artificial neuralnetworks with previous variable selection. Analytica Chimica Acta, 524, 225e234.

Goodacre, R., Radovic, B. S., & Anklam, E. (2002). Progress toward the rapidnondestructive assessment of the floral origin of European honey usingdispersive Raman spectroscopy. Applied Spectroscopy, 56, 521e527.

Hineno, M. (1977). Ir-spectra and normal vibrations of beta-D-glucopyranose. Car-bohydrate Research, 56, 219e227.

Kacurakova, M., & Mathlouthi, M. (1996). FTIR and laser-Raman spectra of oligo-saccharides in water: characterization of the glycosidic bond. CarbohydrateResearch, 284, 145e157.

Korolevich, M. V., Zhbankov, R. G., & Sivchik, V. V. (1990). Calculation of absorption-band frequencies and intensities in the IR-spectrum of alpha-D-glucose in acluster. Journal of Molecular Structure, 220, 301e313.

Kneipp, K., Kneipp, H., Itzkan, I., Dasari, R. R., & Feld, M. S. (1999). Ultrasensitivechemical analysis by Raman spectroscopy. Chemical Reviews, 99, 2957e2976.

Mathlouthi, M., & Luu, D. V. (1980). Laser-raman spectra of D-glucose and sucrose inaqueous-solution. Carbohydrate Research, 81, 203e212.

Moros, J., Inon, F. A., Garrigues, S., & de la Guardia, M. (2005). Determination of theenergetic value of fruit and milk-based beverages through partial-least-squaresattenuated total reflectance-Fourier transform infrared spectrometry. AnalyticaChimica Acta, 538, 181e193.

Ozbalci, B., Boyaci, I. H., Topcu, A., Kadilar, C., & Tamer, U. (2013). Rapid analysis ofsugars in honey by processing Raman spectrum using chemometric methodsand artificial neural networks. Food Chemistry, 136, 1444e1452.

Rodriguez-Saona, L. E., Fry, F. S., McLaughlin, M. A., & Calvey, E. M. (2001). Rapidanalysis of sugars in fruit juices by FT-NIR spectroscopy. Carbohydrate Research,336, 63e74.

Shaver, M. J. (2001). Chemometrics for Raman spectroscopy (Vol. 7). The United Statesof America: Marcel Dekker Inc.

Shih, C. J., & Smith, E. A. (2009). Determination of glucose and ethanol after enzy-matic hydrolysis and fermentation of biomass using Raman spectroscopy.Analytica Chimica Acta, 653, 200e206.

Soderholm, S., Roos, Y. H., Meinander, N., & Hotokka, M. (1999). Raman spectra offructose and glucose in the amorphous and crystalline states. Journal of RamanSpectroscopy, 30, 1009e1018.

Tipson, R. S. (1987). Advances in carbohydrate chemistry and biochemistry. InR. S. Tipson, & D. Horton (Eds.), Vibrational spectra of carbohydrates (p. 77).London: Academic Press, Inc.

Wells, H. A., & Atalla, R. H. (1990). An Investigation of the vibrational-spectra ofglucose, galactose and mannose. Journal of Molecular Structure, 224, 385e424.

Zhbankov, R. G., Andrianov, V. M., & Marchewka, M. K. (1997). Fourier transform IRand Raman spectroscopy and structure of carbohydrates. Journal of MolecularStructure, 437, 637e654.