Detection of flavour release from pectin gels using electronic noses

130 downloads 45577 Views 322KB Size Report
An electronic nose device developed in our laboratory was used to follow the flavour release of a multi-component essence encapsulated in pectin gels .... 0.25mm i.d. × 0.25 μm film, Supelco). ..... one, it is observed that the registered intensity is lower for ..... royal gala apple aroma using electronic nose technology-potential.
Sensors and Actuators B 101 (2004) 28–38

Detection of flavour release from pectin gels using electronic noses Mar´ıa Eugenia Monge a , Donatella Bulone b , Daniela Giacomazza b , Delia L. Bernik a , R. Mart´ın Negri a,∗ a

b

Departamento de Qu´ımica Inorgánica, Anal´ıtica y Qu´ımica F´ısica, Instituto de Qu´ımica F´ısica de Materiales, Ambiente y Energ´ıa (INQUIMAE), Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Ciudad Universitaria, Pabellón 2, (C1428EHA) Buenos Aires, Argentina Istituto di BioFisica-Sezione di Palermo, Consiglio Nazionale delle Ricerche (CNR), Via U. La Malfa 153, Palermo I-90146, Italy Received 13 November 2003; received in revised form 3 February 2004; accepted 10 February 2004 Available online 27 April 2004

Abstract An electronic nose device developed in our laboratory was used to follow the flavour release of a multi-component essence encapsulated in pectin gels. Gelation kinetics of high-methoxylated pectin was followed by dynamic rheological measurements in presence and absence of flavour. A transition from a non-stabilized structure to a fully stabilized gel was observed at about 100 min since mixing the gel components. Flavour release through different days after encapsulation was analyzed following the time evolution of the individual gas sensor’s signals of the e-nose and also the modification of the fingerprints associated to the whole response of the sensors array. The correlation with rheological measurements indicate that the e-nose is able to detect the gel formation and to follow the flavour release changes through aging of the material using principal component analysis (PCA). Artificial neural networks (ANN) analysis was also a successful approach to classify the samples in three categories related with the flavour composition and concentration in the gas phase. © 2004 Elsevier B.V. All rights reserved. Keywords: Electronic nose; Encapsulation; Flavour release; Gels; Multivariate analysis; Pectin; Rheology

1. Introduction The release of volatile essences from gels is an issue of high relevance in food technology and food colloids science, particularly in the case of a multi-component flavouring essence encapsulated in a biocompatible gel. By the encapsulation of flavouring essences in defined conditions it is possible to avoid rapid evaporation, to protect the essence from oxidation and, in some cases, to have a controlled release of the flavour [1]. Since the last decade, there is an increasing interest for developing different strategies to follow the flavour release to the air phase in simple, non-destructive manners. A central issue is related to the investigation of the influence of gels structural properties on the flavour release. In reference to the methods for analyzing the flavour release, the most popular are based on static headspace, followed by gas chromatography with flame ionization detection (GC/FID) or gas chromatography combined with mass spectrometry (GC/MS). These methods are used for ∗ Corresponding author. Tel.: +54-11-4576-3378; fax: +54-11-4576-3341. E-mail address: [email protected] (R.M. Negri).

0925-4005/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2004.02.019

separation, quantification and/or identification of the components, sampling at a fixed time during flavour release [2,3]. This classical methodology allows the accurate analysis of volatile compounds and determines the chemical nature of the components of flavour samples. These methods have been used to study the flavour release of model flavours of relatively simple composition, [4,5]. However, real flavours are complex mixtures typically of more than 20 compounds with different properties. The complete analysis of an essence would consist in the separation, identification and quantification of each aroma compound in the mixture. This makes the interpretation of the resulting data set a complicated task [6]. Moreover, following the changes in the flavour release pattern in real-time would become an exhausting duty. It has been demonstrated that the use of the electronic nose methodology has become a powerful tool as an odour sensing and recognizing system [6–9]. The electronic nose is a device composed by an array of gas sensors, with non specific responses that has pattern recognition ability from multivariate data analysis [7]. The information is obtained from the sensors array responses by pattern recognition techniques, such as principal component analysis (PCA)

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

or artificial neural networks (ANN). This methodology is promising for real-time, in-line, in situ determinations and non-destructive sensing, especially useful in the food industry. On the other hand, the study of the influence of gel structural properties on flavour release is a field of research in continuous progress in food chemistry, as the methods for analyzing rheological properties become accessible [3]. Thus, the aim of the present work is to explore the capabilities of a simple method based on the electronic nose methodology to investigate the kinetics of flavour release of a multi-component essence and the influence of the encapsulation media. We used a portable electronic nose device, developed at INQUIMAE, which does not require headspace injection or separation of the components. This device was applied in previous works to studies of fish freshness and discrimination of fragrances [8,9]. We used gels based on natural high-methoxylated pectin, one of the most important gelling polysaccharides in food technology, as the encapsulation media. This hydrocolloid has the ability of thickening or gelling in aqueous systems under specific conditions [10]. Rheological measurements were performed in order to study the influence of the viscoelastic properties of the gel on the flavour release.

2. Materials and methods 2.1. Materials Commercial citrus high-methoxylated pectin (HMP) type 105 rapid set (origin Brazil; provided by Rosenfeld, Argentina), was used as purchased. The tutti-frutti essence was provided by International Flavours and Fragrances, Argentina. Analytical grade sucrose, potassium citrate, citric acid and sodium benzoate and Milli Q water were used in the gel preparation. 2.2. Gas chromatography combined with mass spectrometry (GC/MS) An analysis of the liquid essence was performed with GC/MS in order to identify its main components. A Shimadzu GC-17A gas chromatograph with mass detection in the range 40-450 mass/charge range was used with an SPB1 fused silica capillary column (30 m × 0.25 mm i.d. × 0.25 ␮m film, Supelco). It was used as carrier gas (column pressure was 80 kPa, column flow was 1.3 ml/min, split ratio: 7). The Injector temperature was 220 ◦ C and the interface temperature was 250 ◦ C. The column temperature program was 70 ◦ C, hold for 5 min and 15 ◦ C/min to 250 ◦ C. The main component detected in the tutti-frutti essence was limonene. The presence of 1-hexene, 2,3-dimethylpentane, 2-phenantrenol, pentanoic acid, cyclohexane, ethanol and 2-methylpentane was also detected.

29

The percentage of limonene is estimated in about 80% in this series of compounds. The proportion of the other components were estimated within the following intervals: (8 ± 4)% for 1-hexene, (4 ± 2)% for 2,3-dimethylpentane and less than 2% for each of the other compounds of the series. 2.3. Gel preparation The procedure for gel preparation was the following. First, the pectin powder was mixed with 10% of the total sucrose. This mixture was dissolved in a potassium citrate buffer 12 mM and heated under stirring up to 80 ◦ C for 10 min. At that moment, the rest of the sucrose was added up to the desired final sucrose concentration (62% (w/w)). The solution was stirred for 10 more minutes at 80 ◦ C. Then, the system was left at room temperature until reaching approximately 40 ◦ C. At this point, 200 ␮l of tutti-frutti essence and a few microliters of 50% citric acid solution were added to 4 ml of the sucrose–pectin mixture and stirred with a vortex for 30 s in a closed recipient. Finally, the closed recipients with the mixture were stored at a controlled temperature of 33 ◦ C using a circulating water bath. Five gel’s replicates of the same weights were prepared by this procedure for each experiment. The final concentration of pectin in the obtained gels was 0.4% (w/w). The refractive index was measured in each experiment to check the sugar content of the solutions (before reaching the gel point), expected at about 62◦ Brix. A portable Arcano refractometer Model Ref 107 0–90%Brix was used. 2.4. Sucrose solutions Two hundred microlitres of tutti-frutti essence were added to 4 ml of 62% (w/w) sucrose solutions prepared in citrate buffer at 33 ◦ C in absence of pectin. The sucrose concentration was controlled with the refractometer. 2.5. Rheological measurements Rheological measurements under low-amplitude oscillatory shear were performed on a controlled stress AR 1000 (TA Instrument, UK) rheometer using a double cylinder geometry (rotor outer radius 21.96 mm, rotor inner radius 20.38 mm, stator outer radius 20.00 mm, cylinder immersed height 59.50 mm, gap 500 ␮m). Measurements were done at a frequency of 0.5 Hz and a strain of 4 × 10−3 . The hot solution was loaded into the geometry set which was previously thermostatized. The cylinder–cylinder upper gap was coated with silicon oil to avoid loss of solvent. The temperature was set using a Haake DC 30 circulating bath. Elastic (G ) and viscous (G ) moduli were monitored as function of time for a period of 24 h. After this period of time, mechanical spectra were recorded at frequencies ranging from 0.02 to 30 Hz.

30

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

2.6. Electronic nose

2.7. Flavour release determinations

An electronic nose device, developed at the University of Buenos Aires and described in previous works [8,9], was used for the detection of the flavours released to the air from the gels. The device consists of an array of non-specific gas sensors and a sample chamber. In the present case, 10 polycrystalline tin dioxide based commercial sensors, from Figaro were used. The identification codes of the sensors are the following: TGS 831 (sensor 1), TGS 813 (sensor 2), TGS 825 (sensor 3), TGS 832 (sensor 4), TGS 880 (sensor 5), TGS 826 (sensor 6), TGS 816 (sensor 7), TGS 842 (sensor 8), TGS 823 (sensor 9) and TGS 800 (sensor 10). The whole set of sensors’ signals is referred as a “fingerprint”. The electronic nose methodology allows to obtain and to discriminate the fingerprints of complex mixtures under different conditions. The flask with the sample was placed into the chamber of the e-nose for each measurement. We started to record the sensor’s signal immediately after closing the chamber during a time referred as τ, which corresponds to the steady state situation. The values of the sensor’s signals at τ were used for the analysis after the subtraction of a base line [8]. Therefore, a set of 10 signals {S1 , . . . , S10 } was obtained in each measurement. Each signal is indicative of the electrical conductance change in the respective sensor due to the presence of the flavour in the chamber (Fig. 1). The relative humidity inside the chamber was always measured, (values between 30 and 50%), but the measured values were not used for the multivariate analysis.

The recipients containing the gels with the encapsulated flavours were kept closed and stored at 33 ◦ C in a circulating water bath. Three hours after encapsulation, one of the samples was opened and measured with the e-nose. This elapsed time was necessary in order to assure gel formation and stabilization (see rheological analysis of the gels). Along the consecutive days, the other replicates were opened and measured. Once a given replicate was opened, it was kept opened and stored at 33 ◦ C to be able to follow the flavour release kinetics in the following days. For example, the third replicate was opened in day 2 and measured during days 2–5. From now on the number of days that the sample was left opened at 33 ◦ C will be referred as D. In this way, we studied both the influence of gel aging on flavour release in a closed recipient and the flavour release from an opened recipient as a function of D. A measurement consists of an “e-nose” determination performed day after day during a week. That is, a vector {S1 , . . . , S10 } was obtained every day for each sample. The fingerprints were highly reproducible between different replicates. We use the term “free essence” to refer to the commercial pure tutti-frutti flavour from now on. Several controls were performed. First, e-nose measurements of the pectin gel without loading with the essence were performed in order to discard the possibility of a background signal due to the other components of the gel, in particular to water. No sensors’ signals were obtained for the encapsulation media (the unloaded gel), that is the pectin gel does not “smell”. This result indicates that the contribution of the gel to the water humidity does not

Fig. 1. Signals of the sensors as function of measurement’s time, since the introduction of a tutti-frutti essence sample, up to reaching steady state. Each curve corresponds to a different sensor of the electronic nose.

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

interfere in the tutti-frutti flavour detection with this e-nose device. Controls using free limonene and encapsulated limonene were performed as limonene is the main component of the tutti-frutti essence. The array pattern of replicated and control samples, like the free essence, encapsulated essence and free limonene, were recorded in different days before and after performing the release experiments, in order to detect the possible influence of sensors drift through the days. No differences were detected, indicating that the drift of the sensors is negligible in the situations described in the present work. 2.8. Data analysis: principal component analysis and artificial neural networks The e-nose methodology uses multivariate data analysis in order to discriminate between groups of signals and, as a consequence, to discriminate between different samples. For a given sample it is possible to analyze the kinetics of the flavour release by comparing both the changes of the individual sensor signals and the changes of the fingerprints provided by the array. There are several well-known methods of multivariate data analysis, but perhaps the most popular is principal component analysis (PCA), a linear unsupervised method that is useful for a graphic visualization and data discrimination. PCA was performed with the whole group of obtained data. Thus, for every data group {S1 , . . . , S10 }, the so-called principal components, {PC1 , PC2 , . . . , PC10 }, were calculated. The important point is that, when analyzing the total set of {PC1 , PC2 , . . . , PC10 } groups, more than 96% of the total data variance resulted contained in the group defined by the vectors {PC1 , PC2 }. Data were not normalized. We also used artificial neural network (ANN) analysis in order to discriminate between different samples. ANN is a non-parametric method and does not require linearity of the sensor responses with the concentration of the volatile components in the working range as PCA does. It was used a back propagation neural network, composed by a one hidden layer. The number of neurons in the hidden layer and the number of epochs were optimized in the training step in 5 and 996, respectively. It was used 36 of the 44 measurements (80%) as exemplars for training; 4 measurements (10%) were used as a cross-validation data set and other 4 measurements (10%) were used as the testing data set. The outputs were defined as (1 0 0) for the free tutti-frutti essence, (0 1 0) for the flavour release from pectin gels at D = 0 and (0 0 1) for the flavour release from gels at 1 ≤ D ≤ 5. The two principal components, PC1 and PC2 , associated to the release experiment were used as input vectors for the artificial neural network. The momentum rate used was 0.7 and the step size 0.1.

31

3. Results and discussion 3.1. Rheological analysis of the gels The determination of gel point or gel time is always difficult and ambiguous [11] as the definition of gel itself is a crucial and still disputed question in literature. Many criteria have been proposed to answer this question, but they are largely dependent on the chosen time scale [12]. In the present work we follow the criterion used by many authors, that is, to consider a gel as a system bearing G larger than G at the chosen strain and frequency value. Accordingly, the gelation time is defined as the time at which G equals G . It is worth noting that even systems macroscopically fluid can exhibit G larger than G , as observed already in pectin at high temperatures and concentrations [13] or in entangled polymeric solution at high frequencies [12]. Thus, the above criterion should be adopted by considering that the crossover between G and G marks the appearance of a gel. The values of G and G are shown in Fig. 2a as function of time. It is observed that after the first 10 min, G is always larger than G , indicating a noticeable solid-like response since the very early stage of the measurement. The fast initial growth of G and G is followed by a slower increase, reflecting the network rearrangement. These results indicate that the gel formation occurs suddenly since cooling the reagents. The system continues evolving up to the formation of a stable network, at times longer than 100 min. The derivative with time of G , dG /dt, is represented as a function of time in Fig. 2b in order to better appreciate the structure development rate [14]. The latter displays a maximum in dG /dt at about 6 min and reaches a steady value in 100 min (Fig. 2b). Typical time for thermal equilibration with our experimental setting is 5–6 min. So, the first three points of Fig. 2 reflect the combined effects of gel setting and temperature decrease. These two effects are interdependent because the gel formation is also a consequence of temperature change. If the first three points are omitted in Fig. 2b, the derivative’s maximum disappears and only the decreasing part of the curve is visible. This is equivalent to say in terms of the description reported by Lopes da Silva et al. [14] that the gel formation is so fast that the first stage can not be resolved. Therefore, the decreasing of dG /dt over a much longer time reflects the change in the gel structure or structural evolution as a rearrangement of polymer chains up to a full-stabilized structure. The √ dependence of G , G and the complex viscosity (η∗ = G2 + G2 /ω) with frequency is shown in Fig. 3. The flatness spectrum of G and the slope value close to −1 obtained for the log–log dependence of η* with frequency, are typical of gelled systems [13] and solid-like samples. We know from previous experiments [15] that pectin solutions in low concentrations appears as a system of branched

32

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

Fig. 2. (a) Storage (G ) and loss (G ) modulus, as function of time (b) Derivative of the elastic modulus, dG /dt, as function of time up to 100 min after mixing. T = 33 ◦ C. ω = 0.5 Hz. The results for samples with and without flavour are shown. The line at 5 min indicates the time required for thermal equilibration.

clusters, so it is reasonable that at higher concentrations the system behaves as a transient network, at least from a mechanical point of view. Gels without flavour appear to have a larger elastic component at times longer than 20 min, in the whole range of considered frequencies. Besides, the value of the maximum of dG /dt in presence of the flavour (Fig. 2b) is smaller than the value obtained without flavour. These results indicate that the essence decreases both the degree of polymer chains entanglement and also the rate of gel formation up to a fully stabilized structure. Similar results have been obtained previously with hydrophilic polymers [16]. The value of G and the time position of the maximum in dG /dt are different in the presence of tutti-frutti essence. We have observed this behaviour in several replicates.

3.2. Analysis of the sensors’ signals and fingerprints A fast increase of the signals due to the release of the flavour from the top of the gel–air interface towards the sensors was observed after introducing each sample in the e-nose chamber (Fig. 1). The signals increase until the flavour saturates the chamber at a certain time, referred as τ (see Section 2). This time was about 20 min for the samples that were measured the same day of preparation (D = 0), while it took 40 min for the samples that were left opened for more than 24 h (D ≥ 1). It was observed that the signals {S1 , . . . , S10 } decrease as function of D for all the samples considered (Fig. 4). Briefly, the signals decrease day after day and it takes more time to saturate the e-nose chamber, as the flavour content of the gels decreased progressively.

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

33

Fig. 3. Frequency sweep: G , G and the complex viscosity (η*) in the range from 0.01 to 30 Hz. The measurements were performed 24 h after the start of gelation, that is, for a stabilized gel in presence of flavour. T = 33 ◦ C; (䊉) G ; (䊊) G ; (䉲) η*.

The group of signals {S1 , . . . , S10 } shown in Fig. 4 constitutes a pattern or fingerprint that is characteristic of a particular sample at a given day D. The fingerprint of a measurement can be represented as a radar or a bar plot of the signals {S1 , . . . , S10 } obtained. For example, in the radar plot, each vertex corresponds to one sensor. Fig. 5a shows the radar plots obtained for the free tutti-frutti essence.

Fig. 5b shows the pattern at the first day of release (D = 0) of the encapsulated essence. Fig. 5c shows the fingerprint of the tutti-frutti essence dissolved in a 62% (w/w) sucrose solution (not gelled). It is clear from Fig. 5 that both the intensities and the fingerprints are different for the free essence, the encapsulated essence and the essence in sucrose solution. Comparing the

Fig. 4. Signals of the array of sensors correspondent to a particular sample of tutti-frutti essence encapsulated in pectin gel, for different days (D) after encapsulation. The set of five bars associated to each sensor represents the signals for 0 ≤ D ≤ 4, consecutively.

34

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

Fig. 5. Radar plot of different samples having the same amount of essence. Each vertex of the radar corresponds to one sensor of the array. (a) Free essence; (b) essence encapsulated in pectin gel at D = 0; (c) essence dissolved in 62% (w/w) sucrose solution; (d) free limonene (note that the scale is different in (d)).

results obtained for the free essence and the encapsulated one, it is observed that the registered intensity is lower for the latter and the odour pattern is also modified. As mentioned in Section 2 (flavour release determinations), free and encapsulated limonene were also measured because this is the main component of the essence. These experiments not only constitute some of the controls for sensors drift, but also support the general interpretation of the data. In particular we observed that the fingerprints (radar plots) of “free” and encapsulated limonene are the same, with the only difference of lower signals for the encapsulated limonene. Of higher relevance is the observation that the radar plot of limonene is completely different from that corresponding to the “free” tutti-frutti essence and to the encapsulated tutti-frutti essence (Fig. 5), indicating that the other components of the tutti-frutti essence have a non negligible contribution to its fingerprint. The results of the release through different days are in agreement with a differential release of some compounds of the essence, due to different factors such as changes in

partition coefficients and solubilization of the essence components. This conclusion is validated by the different fingerprint observed for limonene. As a matter of fact, the PCA point in Fig. 7 associated to free limonene is clearly discriminated from those of the tutti-frutti essence (free and encapsulated). The intensity of the fingerprint corresponding to the essence dissolved in the sucrose solution is almost intermediate between those of the free essence and the encapsulated tutti-frutti essence pattern. This shows the important effect of the media on the release (see also Roberts et al. [4]). In the case of the essence in sucrose solution, the results reflect the interaction of the flavour compounds with water and sucrose molecules. The effect of sucrose concentration in the flavour release is a subject of interest at present. Hansson et al. [2] have shown that an increase in the concentration of sucrose increased the release of the most polar flavour compounds due to a “salting-out” effect of sucrose and that the release changed when pectin is added in no gelling conditions.

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

35

Fig. 6. Signal of a particular sensor of the array (sensor number 4, in volts) as a function of D, the number of days that an encapsulated essence ( ) or the essence in 62% (w/w) sucrose solution (䉫) were exposed to the air at controlled temperature. The results of different replicates are shown.

The amount of flavour released from the gel increases with temperature (data not shown). Although this is probably due to the increase of the vapour pressure of the individual components of the essence, the analysis of the temperature effects is not an easy subject. In fact, the structural properties of the pectin gels are very much dependent on temperature as we have observed (see Rheological analysis of the gels) and also reported by Lopes da Silva et al. [14]. 3.3. Release through different days As we have discussed, there is a very fast release of the flavour from the gel, clearly shown in Fig. 4. This situation is also illustrated when considering the signals of the sensor that provides the highest signal, sensor number 4 (Fig. 6). There is a very fast decrease of the signals during the first 24 h. The signals decrease more than 50% from D = 0 to D = 1 for the open system. The fact that a mixture of compounds is encapsulated in the gels determines the change of the fingerprint and the rate of the release during aging. In addition, Fig. 6 shows a clear difference of the release kinetics between the essence encapsulated in the gel and the essence dissolved in sucrose solutions. This result illustrates again the influence of the media on the flavour release. 3.4. Principal component analysis and artificial neural network analysis The whole data obtained for tutti-frutti with the e-nose were analyzed by PCA. The two first principal components set contains the 96.6% of the data variance, so an effective dimensional reduction was achieved. The plot of PC1 versus PC2 is shown in Fig. 7 where each point is associated to one measurement.

Different regions can be distinguished in the PC1 –PC2 plane. In the direction of decreasing PC1 , the first region with full circles (䊉) corresponds to the free essence. The following points (䉫) correspond to the flavour release from sucrose solutions. The full squares correspond to measurements done at 25, 90, 140 and 190 min after the addition of tutti-frutti flavour to the pectin–sugar–buffer mixture at 33 ◦ C. The first two points (t = 25 and 90 min) correspond to the essence encapsulated in a “non-stabilized” gel, as discussed in rheological analysis of the gels. This region is placed in the PC1 –PC2 plane between that correspondent to the essence in sucrose solution at D = 0 and the encapsulated one in a “full-stabilized” structure. The points corresponding to different days of release (D) from “stabilized” gels are placed in different sub-areas. The decrease of PC1 from 1.2 to −0.6 is correlated with the decrease of the essence concentration in the system, starting from the free essence, following with the essence dissolved in sucrose solution and finishing with the release from gels. In the same way, the decrease of PC2 values is parallel to an increasing matrix effect, starting without the presence of a matrix, following with sucrose solutions and finishing with gels that change from a “non-stabilized” structure to a “fully stabilized” one. This matrix effect influences the degree of retention of the essence components. Therefore, the PCA analysis clearly shows the influence of the media on the odour. A second PCA analysis was performed using only the data associated to the release of encapsulated tutti-frutti essence in “fully-stabilized” pectin gels but including the free essence (Fig. 8). Clearly the points correspondent to the first day of release (D = 0) can be discriminated from those associated to the following days as well as from the tutti-frutti essence. In the same way as in Fig. 7, the PC1 –PC2 points associated to measurements performed at different

36

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

Fig. 7. PCA analysis. The data correspondent to the free essence, the essence in a 62% (w/w) sucrose solution, the essence encapsulated in the gel and free limonene were considered for the analysis. (䊉) Free essence; (䉫) essence in 62% (w/w) sucrose solution; (䊏) essence in gel, released during D = 0, at different times after gel manufacturing that belong to “non-stabilized” structures (see the text); ( ) essence in gel, released at different days (0 ≤ D ≤ 5); ( ) free (non encapsulated) limonene.

Fig. 8. PCA analysis using exclusively the results obtained at different D for samples of free essence and encapsulated in gel. (䊉) Free essence; ( ) encapsulated, D = 0; (䉱) encapsulated, 1 ≤ D ≤ 5.

days move monotonically towards lower PC1 values, through aging. The first principal component is again associated to the concentration of the flavour in the system whereas the second is related to the presence of a matrix as encapsulating media. The results show that PCA provides a good graphical description of the release behaviour in different samples and that this kind of analysis can be used to follow the flavour release through different days. The PCA also

shows that after introducing the flavour in the sample, the release during the first 100 min has different characteristics from the release at longer times. This reflects the evolution of the flavour release pattern along the stabilization of the gel properties (compare Fig. 7 with Fig. 2). Artificial neural network analysis was also performed to test the power of the electronic nose for discriminating samples. The two principal components PC1 and PC2 obtained

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38 Table 1 Artificial neural network results Expected outputs 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1

0 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0

ANN results 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 0

1.00 0.00 0.00 0.01 −0.03 −0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.91

0.01 1.05 1.05 1.05 1.04 1.05 0.07 0.06 0.06 0.06 0.06 0.05 0.06 0.05 0.05 0.17

−0.06 −0.05 −0.05 −0.05 −0.03 −0.05 0.92 0.93 0.93 0.94 0.94 0.94 0.94 0.94 0.94 −0.06

The outputs were defined as (1 0 0) for the free essence; (0 1 0) for the flavor released from gels at D=0; (0 0 1) for flavor released at 1 ≤ D ≤ 5.

for the experiment shown in Fig. 8 were used as inputs for the ANN. The samples were classified into 3 categories, which are the outputs of the network: free tutti-frutti essence, flavour release from gels at D = 0 and flavour release from gels at 1 ≤ D ≤ 5 (see data analysis in Section 2). The outputs of the network were defined as (1 0 0) for the free essence; (0 1 0) for the flavour released from gels at D = 0; (0 0 1) for flavour released at 1 ≤ D ≤ 5. These three groups can be clearly distinguished in the PCA. In this way, the outputs of the network are related to the presence of the encapsulation media and the aging effect on the flavour release. A replication of the experiment shown in Fig. 8 was used as a completely independent data set. These are unknown samples for the built neural network in order to test its potential for classifying the data set in the three mentioned categories. Therefore, a PCA was firstly performed on the results of the replication experiment and then the PC1 and PC2 values were introduced as inputs in the network in order to be classified. The results of Table 1, referred to this data set, show that the network was able to classify correctly 100% of the samples.

4. Conclusions The correlation of e-nose measurements with the rheological measurements show that the e-nose is able to detect changes in flavour release indicative of structural changes in the encapsulation media, in our case gel formation and gel stabilization. The fingerprints of the flavour released from a gel are clearly discriminated from the free essence fingerprint and from those associated to flavour release from sucrose solutions at matched concentrations. The signal decrease with gel aging is probably due to a decrease of the flavour activity in the gel phase. We demonstrated that

37

PCA allows following the flavour release through different days. It is expected an increase of the amounts of future studies in this field using the e-nose methodology because it also allows appreciating the influence of the rheological properties of the system on the release kinetics.

Acknowledgements RMN and DLB are members of the Consejo Nacional de Investigaciones Cient´ıficas y Técnicas (CONICET, Argentina). MEM has a fellowship of the Agencia Nacional de Promoción Cient´ıfica y Tecnológica (ANPCYT, Argentina). The work was supported by the University of Buenos Aires (Programación UBACyT 2001-2002, Proyecto X171), Fundación Antorchas and ANPCYT (Proyecto BID-1201 OC-AR PICT 4438, Argentina). International Flavours & Fragrances Argentina is acknowledged for providing the essence used in this work. We thank Dr. V. Martorana (CNR) for his constant support and encouragement.

References [1] B.L. Zeller, F.Z. Saleeb, R.D. Ludescher, Trends in development of porous carbohydrate food ingredients for use in flavour encapsulation, Trends Food Sci. Technol. 9 (1999) 389–394. [2] A. Hansson, J. Andersson, A. Leufvén, The effect of sugars and pectin on flavour release from a soft drink-related model system, Food Chem. 72 (2001) 363–368. [3] A. Hansson, A. Leufvén, K. Pehrson, B. Stenlöf, Multivariate analysis of the influence of pectin, white syrup, and citric acid on aroma concentration in the headspace above pectin gels, J. Agric. Food Chem. 50 (2002) 3803–3809. [4] D.D. Roberts, J.S. Elmore, K.R. Langley, J. Bakker, Effects of sucrose Guar Gum and carboxymethylcellulose on the release of volatile flavour compounds under dynamic conditions, J. Agric. Food Chem. 44 (1996) 1321–1326. [5] Y.M. Gunning, P.A. Gunning, E.K. Kemsley, R. Parker, S.G. Ring, R.H. Wilson, A. Blake, Factors affecting the release of flavour encapsulated in carbohydrate matrixes, J. Agric. Food Chem. 47 (1999) 5198–5205. [6] H. Young, K. Rossiter, M. Wang, M. Miller, Characterization of royal gala apple aroma using electronic nose technology-potential maturity indicator, J. Agric. Food Chem. 47 (1999) 5173–5177. [7] J.W. Gardner, P. Bartlett, Electronic Noses, Oxford University Press, Oxford, 1999. [8] M. O’Connell, G. Valdora, G. Peltzer, R.M. Negri, A practical approach for fish freshness determinations using a portable electronic nose, Sens. Actuators B 80 (2001) 149–154. [9] A. Branca, P. Simonian, M. Ferrante, E. Novas, R.M. Negri, Electronic nose based discrimination of a perfumery compound in a fragrance, Sens. Actuators B 92 (2003) 222–227. [10] P.M. Gilsenan, R.K. Richardson, E.R. Morris, Thermally reversible acid-induced gelation of low-methoxy pectin, Carbohydr. Polym. 41 (2000) 339–349. [11] S.B. Ross-Murphy, Incipient behavior of gelatin gels, Rheol. Acta 30 (1991) 401–411. [12] G.M. Kavanagh, S.B. Ross-Murphy, Rheological characterisation of polymer gels, Prog. Polym. Sci. 23 (1998) 533–562.

38

M.E. Monge et al. / Sensors and Actuators B 101 (2004) 28–38

[13] V. Evangeliou, R.K. Richardson, E.R. Morris, Co-gelation of high methoxy pectin with oxidised starch or potato maltodextrin, Carbohydr. Polym. 42 (2000) 233–243. [14] J.A. Lopes da Silva, M.P. Gonçalves, M.A. Rao, Kinetics and thermal behaviour of the structure formation process in HMP/sucrose gelation, Int. J. Biol. Macromol. 17 (1995) 25.

[15] D. Bulone, V. Martorana, C. Xiao, P.L. San Biagio, Role of sucrose in pectin gelation: static and dynamic light scattering experiments, Macromolecules 35 (2002) 8147–8151. [16] N. Francois, A.M. Rojas, M.E. Daraio, D.L. Bernik, Dynamic rheological measurements and drug release kinetics in swollen scleroglucan matrices, J. Control. Rel. 90 (2003) 355–362.