Predicting Vapor-Phase Concentrations for the Assessment of the

Jun 30, 2017 - To attain this goal, the experimental Henry's law constant (H) of each odorant in each studied fragrance system (containing one, two, t...
0 downloads 6 Views 2MB Size
Article pubs.acs.org/IECR

Predicting Vapor-Phase Concentrations for the Assessment of the Odor Perception of Fragrance Chemicals Diluted in Mineral Oil Patrícia Costa,*,† Miguel A. Teixeira,† Gabriel Mestre,§ Luísa Carneiro,† José Miguel Loureiro,† and Alírio E. Rodrigues† †

Laboratory of Separation and Reaction Engineering−Laboratory of Catalysis and Materials (LSRE−LCM), Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias s/n, 4200-465 Porto, Portugal § IUT Lyon, Villeurbanne, Auvergne-Rhône-Alpes 69622, France S Supporting Information *

ABSTRACT: In this study, the Henry’s law methodology is applied to predict the release of odorants present in single and multicomponent fragrance mixtures diluted in mineral oil, a simplified matrix used in cosmetic products. To attain this goal, the experimental Henry’s law constant (H) of each odorant in each studied fragrance system (containing one, two, three, or four odorants) was first evaluated by plotting their liquid phase and experimental vapor phase concentrations assessed by headspace gas chromatography. From that point, the H value of each odorant in the multicomponent fragrance system was predicted from its corresponding Hexp in the single fragrance component system. The theoretical vapor-phase concentrations were also calculated using the activity coefficients for vapor−liquid equilibria by applying the thermodynamic UNIFAC model. The odor intensity and character of the studied fragrance systems were assessed through the Stevens’s power law and Strongest Component models (psychophysical models). This study confirmed that the headspace concentrations and odor intensity of each odorant present in a multicomponent fragrance mixture dissolved in mineral oil can be efficiently predicted from its corresponding H determined when present alone in the simplified matrix, for low concentrations. Also, comparing both methodologies, UNIFAC and Henry’s law, it was concluded that Henry’s law is a better predictive model for the vapor−liquid equilibria, showing lower deviations from the experimental data. Therefore, the proposed predictive mathematical model can be attractive for the assessment of sensory quality of multicomponent fragrance systems in early formulation stages.



INTRODUCTION The creation of a successful scented product is not only a simple procedure enclosed in magic and artistic creativity, as it seems, but a function of multivariables including performance, character, stability, consumer liking, fit to market, marketing, price, and so forth. Furthermore, the headspace composition above a scented product, that is, the fragrance fraction available in the air to be perceived by the consumer, is different depending on the physicochemical properties of each odorant (e.g., saturated vapor pressure, molecular weight, boiling point, octanol−water partition coefficient) and on odorant−odorant and odorant−matrix interactions and psychophysical properties of the odorants (odor threshold (OT), dose response curve/ power law exponent). Some authors have also reported that this perceptual continuum is a function of the biophysical interactions occurring from the odorant−olfactory receptor level to signal encoding at the neuronal level. Considering cosmetic products, a high number of ingredients are included in their composition, namely, water, oils, silicones, surfactants, polymers, organic solvents, acid and alkali salts, color pigments, proteins, plant extracts, preservatives, and antioxidants, just to mention a few.1 This being said, it is easy to imagine the high number of possible intermolecular interactions © XXXX American Chemical Society

between fragrances and the remaining chemicals. Depending on the product end-use (shampoo, body lotion, moisturizer, deodorant, etc.), we will find different chemicals in the product’s base composition and, therefore, different molecular interactions. Also, we can add more variables to this complex equation by including the viscosity of the mixture, polarity, solubility, pH, temperature, concentration of each chemical, and packaging type.2 For all of these reasons, the formulation process encompasses a prior reflection period to select the most suitable fragrance raw materials to be used in a certain formulation product. At this stage of the process, different questions arise: Considering a whole product line, how will the performance of a selected fragrance be in a body lotion, a shampoo, or a shower gel? Will it be soluble in all of these bases? And applying the same fragrance concentration, in which of them will the odor be more intense? Is it possible to predict its perceived odor despite the complexity of the formulations? And how can we improve the performance of a fragrance? The answers to these Received: Revised: Accepted: Published: A

April 28, 2017 June 29, 2017 June 30, 2017 June 30, 2017 DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

increase or decrease the intensity of a strong odor. Apart from its easiness for prediction of odor intensity, the Strongest Component model also gives assurance that the intensity of a mixture will not be grossly larger than the odor intensity of its strongest odorant. Comparing different odor perception models available in the literature, the Strongest Component model appears as the second or third best model in correlation with olfactory panellists’ evaluations.8,27 Finally, as has been proved in previous studies carried out by our research group, the Strongest Component model allows calculation of the odor intensity and character (the dominant odor) of different fragrance mixtures.4,28 Therefore, in the present study, we intend to extend our methodology using Henry’s law together with psychophysical models (Stevens’s power law and Strongest Component model) to predict the gas concentrations and olfactory perception of complex fragrance systems diluted in mineral oil (a simplified matrix widely used in typical cosmetic formulations). For this purpose, different fragrances, alone and in mixtures, were diluted in mineral oil, and their gas concentrations were calculated by headspace gas chromatography (HS-GC) at equilibrium conditions. These vapor-phase concentrations were then compared with those predicted using Henry’s law from single fragrance mixtures as well as those predicted by the UNIFAC groupcontribution method. Afterward, the vapor-phase concentrations were converted into a measure of olfactory perception (odor intensity and character) using the psychophysical models of the Stevens’s power law and Strongest Component, respectively.

questions are neither simple nor straightforward, requiring scientific knowledge from different fields. In this context, some relevant scientific works have been published reporting the performance of fragrances from liquid mixtures,3−8 textiles,9−12 and skin.13−16 The methodologies proposed by R&D departments from the fragrance industry and academics to describe fragrance diffusion from different matrixes and substrates mostly report experimental measurements due to the difficulty of developing theoretical models to express these complex phenomena. Recently, our research group proposed a new approach for predicting the gas concentrations of odorants to be used in the calculation of the olfactive perception of multicomponent fragrance mixtures from dipropylene glycol (very useful in toiletry formulations) using Henry’s law.17 This methodology assumes that the evaporation of each fragrance component present in a multicomponent fragrance mixture and dissolved in a certain matrix could be estimated from its corresponding Henry’s law constant (H) determined when present alone in the matrix for low concentrations and considering that molecular interactions between odorants (odorant−odorant interaction) are negligible versus their molecular interactions with the matrix (odorant−matrix interaction). The obtained results showed a strong linear relationship between experimental H for single fragrances and experimental H for binary, ternary, and quaternary fragrance mixtures, encouraging the application of this methodology to other matrixes. Furthermore, from the predicted and experimental gas concentrations, we also predicted the perceived odor using the psychophysical models Stevens’s power law and Strongest Component. The Stevens’s power law defines that the perceived sensation is proportional to the stimulus magnitude raised to an exponent. The odor intensity is obtained through the predicted or experimental gas concentration of an odorant, while the OTs and exponents (n) are obtained from databases available in the literature.18−20 In the OT databases, it is possible to find a compilation of three types of OTs: (a) detection thresholds (i.e., the minimum concentration of an odorant that can be detected by humans); (b) recognition thresholds (i.e., if a stimulus is both detected and recognized or identified); and (c) nondefined thresholds (which these databases do not label as “detection” or “recognition”). It is important to mention that in the present study the detection and nondefined thresholds were used to increase the available data. Another important aspect to discuss here are the average values or ranges of thresholds that are reported in these databases without discussing parameters with impact on the accuracy of the olfactory data, namely, the measurement techniques involved in the original references, individual physiological differences, or psychological factors.21−23 Although the Stevens’s power law is widely used to assess the odor intensity of odorants, the acquired results should be discussed considering inherent limitations as it does not consider possible perceptual multicomponent interactions occurring at the nose level, which can eventually influence the perceived odor intensity (suppression or synergy) of the fragrance mixture.24,25 The Strongest Component model has been applied to account for the odor intensity and quality of fragrance mixtures as it presents several advantages in comparison with other models including the Additivity model, Euclidean additivity model, Vectorial model, U model, and UPL2 model.8,26,27 From all, the Strongest Component is probably the simplest model. It does not reflect (as well as others) common observations in olfaction, like weaker odors with the capacity to



MATERIALS AND METHODS Chemicals. R-(+)-Limonene (CAS no. 5989-27-5, purity ≥ 98%) was obtained from Sigma-Aldrich. (±)-Linalool (CAS no. 78-70-6, purity > 97% GC) and (−)-α-pinene (CAS no. 778526-4, purity > 98%) were obtained from Fluka. Benzyl acetate (CAS no. 140-11-4, purity > 99.5%) was purchased from Merck. Linalyl acetate (CAS no. 115-95-7, purity > 97%, FCC) was supplied by Aldrich. p-Cymene (CAS no. 99-87-6, purity > 97%) was obtained from Alfa Aesar. White mineral oil was obtained from Fisher Scientific. All reagents were used as received without further purification. Fragrance System Preparation. Liquid mixtures containing one, two, three, or four fragrance chemicals (Table 1) at different concentrations and diluted in mineral oil were prepared gravimetrically in 4 mL vials using an Adam Equipment balance model AAA250L with a precision of ±0.2 mg. The homogeneity of the fragrance systems was guaranteed using a vortex. In order to have each fragrance chemical present at least one time in each binary, ternary, and quaternary fragrance mixture, different combinations were prepared, as displayed in Table S1, using a simple Design of Experiments. Due to the huge application of limonene in cosmetic and toiletries formulations, it was used in all of the combinations. After preparation of the fragrance mixtures, two replicates of each mixture (1 mL) were placed in 20 mL closed-cap headspace vials and allowed to equilibrate for at least 24 h, at a controlled room temperature (23 ± 1 °C). The schematic representation of the experimental procedure is illustrated in Figure S1. Experimental Measurement Vapor-Phase Concentration. The vapor-phase concentration of each component was quantified by HS-GC using a Varian CP-3800 equipped with a split/splitless injector, flame ionization detector, and capillary B

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Table 1. Properties of the Odorants and the Simplified Matrix, Including the Molecular Formula, Molecular Weight (MW), Vapour Pressure at 23 °C (Psat), Logarithmic Values of Octanol/Water Partition Coefficients (log P), OT, Number of Distinct OTs Averaged for Each Component (N), OT Variability (SD), and Olfactory Power Law Exponent (n) compound

molecular formulaa

MW (g/mol)a

Psat (Pa)

log Pa

OTd(mg/m3)

N

SD

ne

limonene α-pinene linalyl acetate linalool ethyl acetate benzyl acetate p-cymene mineral oil

C10H16 C10H16 C12H20O2 C10H18O C4H8O2 C9H10O2 C10H14 C8H18

136.23 136.23 196.29 154.25 88.11 150.18 134.22 114.23

205.4b 513.4b 14.8c 22.1b 12425.6a,b,c 21.9b 200.0a,b 1880.00g

4.57 4.83 3.93 2.97 0.73 1.96 4.10

0.619 0.240 0.061 0.009 20.4 0.332 0.392 odorless

8 5 2 10 14 2 2

19.0 46.6 0.0370 1.89 350 77.8 0.420

0.37 0.49 0.35f 0.35f 0.51 0.38 0.35f

a From PubMed.gov, U.S. National Library of Medicine National Institutes of Health.29 bFrom the DIPPR 801 Database.30 cFrom the Chemspider Database of Chemical Structures and Property Predictions, Royal Society of Chemistry.31 dFrom refs 18 and 19. OTs, namely, the odor detection thresholds and nondefined thresholds (i.e., those that the compilation does not label as “detection” or “recognition”), were collected from the van Gemert (1999, 2003) databases, which compile OTs from different publications. As a result of the large variability found in literature data (for the same odorant, this difference can be in terms of order of magnitude), a geometric mean value was used in order to minimize variabilities between data. For OT data presented as a range of values, the mean of the two extremes was considered. For limonene, α-pinene, and linalool, the considered OTs were those reported in the databases for limonene, (+)-limonene, α-pinene, (−)-α-pinene, linalool, (+)-linalool, and (−)-linalool. eFrom ref 20. f The median power law exponent value in the compilation of data from ref 20. As in olfaction, the majority of the odorants seem to generate power functions with exponents smaller than unity; the median value suggested by this database is 0.35 whenever the exponent is not known. The compiled data has a weighting coefficient attributed to each author based on statistics, followed by averaging and standardization of the exponent values.20,23,32 g From The Good Scents Company (vapor pressure at 25 °C).33

Table 2. Theoretical (Htheor) and Experimental (Hexp) Henry’s Law Constants, Absolute Relative Deviation between Htheor and Hexp (|δH|, %), Together with the F Value and Maximum Liquid Concentrations (Cliquid,max, g/L) for the Single Fragrance Systems Diluted in Mineral Oil Theoretical

Hexpa

H

single fragrance mixtures limonene α-pinene linalyl acetate linalool benzyl acetate p-cymene ethyl acetate

Experimental

theor

(1.4 (3.3 (2.0 (1.2 (1.0 (1.7 (2.3

± ± ± ± ± ± ±

0.1) 0.3) 0.1) 0.1) 0.1) 0.2) 0.1)

× × × × × × ×

−5

10 10−5 10−6 10−5 10−5 10−5 10−3

(4.46 ± 0.02) (6.0 ± 0.1) (3.98 ± 0.01) (5.4 ± 0.3) (1.47 ± 0.03) (2.86 ± 0.01) (4.71 ± 0.04)

F × × × × × × ×

−5

10 10−5 10−6 10−5 10−5 10−5 10−3

7172 4131 1987 306 565 64656 234

Cliquid,max (g/L)b

|δH| (%)

± ± ± ± ± ± ±

69 45 51 79 32 42 52

173.64 84.88 62.25 16.570 57.92 95.63 11.45

0.02 0.02 0.01 0.001 0.04 0.02 0.01

Values are expressed as mean ± standard deviation (n = 2). p value < 0.05. bThe maximum liquid concentration of each fragrance component where Henry’s law is valid. a

estimated from its corresponding H value when present alone exp in the simplified matrix (Hi,s ) (eq 2)

column Chrompack CP-Wax 52 CB, 50 m length, 0.25 mm i.d., and 0.2 μm film thickness. The injector and detector ports were set at 240 and 250 °C, respectively. The injection volume for the gas phase was 0.4 mL, and the split ratio was 10:1. Gas sampling and injection were performed using a gastight syringe from SGE (Australia) installed in an automatic headspace sampler HT250D by HTA SrL. The carrier gas used was helium (He N60) with a constant flow rate of 1 mL/min. Experimental Henry’s Law Constant Assessment. From the experimental vapor-phase concentration, it was possible to calculate the Henry’s law constant for each odorant i (Hiexp) present in each fragrance system diluted in mineral oil, at a certain liquid concentration, following eq 1 Hiexp

=

Cigas

Cigas Ciliquid

liquid Cigas,pred = Hiexp ,H ,s × Ci

(2)

Vapor-Phase Concentration Prediction. The gas concentration of odorant present in each fragrance mixture was also predicted from a vapor−liquid equilibria (VLE) calculation using the UNIFAC (UNIversal Functional Activity Coefficient) method to evaluate the activity coefficients. In this case, the gas,pred concentrations of the odorants in the headspace (Ci,UNIFAC ) were predicted based on the modified Raoult’s law for describing the VLE (eqs 3 and 4)34

yP = xiγiPisat i Cigas,pred ,UNIFAC =

(1)

Ciliquid

where and are the experimental vapor- and liquidphase concentrations of component i, respectively (g/L). The vapor-phase concentration of each fragrance component i (Cigas,pred ) in multicomponent fragrance systems was also ,H

(3)

yM P i i RT

= xiγi

MiPisat RT

(4)

where xi and yi represent the liquid and gas mole fractions of component i, respectively, Mi is its molecular mass, γi is the activity coefficient in the liquid phase, Pisat is the vapor pressure of pure component i, R is the universal gas constant, and T is C

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Table 3. Theoretical (Htheor) and Experimental (Hexp) Henry’s Law Constants, Absolute Relative Deviation between Htheor and Hexp (|δH|, %), Together with the F Value and Maximum Liquid Concentrations (Cliquid,max, g/L) for the Binary, Ternary and Quaternary Fragrance Systems Diluted in Mineral Oil

Values are expressed as mean ± standard deviation (n = 2). p value < 0.05. bThe maximum liquid concentration of each odorant where Henry’s law is valid.

a

D

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 1. Experimental gas-phase concentration of limonene as a function of its liquid phase concentration in all of the studied mixtures in mineral oil.

vapor-phase concentrations using the Stevens’s power law,35 according to eq 6

the absolute temperature. UNIFAC parameters used to calculate the activity coefficients are listed in Table S2, and the activity coefficients of all odorants present in the liquid phases of the studied fragrance systems are displayed in Tables S3−S6. Theoretical Henry’s Law Constant Estimation. From eq 4, one can calculate the theoretical Henry’s law constant (Hitheor ) as eq 5 Hitheor =

⎛ C gas ⎞ni ψi = ⎜ i ⎟ ⎝ OTi ⎠

where is the concentration of the component i in the gas phase, OTi is the odor threshold (odor detection threshold and nondefined OT) of component i measured in air (both in units of mass or mole per volume), and ni is the power law exponent for each component i. Once the odor intensities for each odorant in the fragrance systems are obtained, it is possible to assess their odor quality using the Strongest Component model.26 It considers that the odor character of a mixture of N odorants is governed by the one with the highest odor intensity (eq 7)

γiPisat RTC T

(6)

Cigas

(5)

Ci(g/L) ⎞ ⎛ ⎟. where, CT is the total liquid concentration in mol ⎜C T = ∑ M L ⎝ i(g/mol) ⎠ theor values for single fragrance systems are listed in The H Table 2. Perceived Odor Evaluation. The odor intensities above the liquid fragrance systems were calculated from the

ψmix = max(ψi ) E

∀ i = 1, ..., N

(7)

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 2. Experimental gas-phase concentrations of α-pinene (A), linalool (B), p-cymene (C), linalyl acetate (D), benzyl acetate (E), and ethyl acetate (F) as a function of their liquid-phase concentrations in mineral oil for all of the mixtures studied.

Comparison of Predicted Vapor-Phase Concentrations or Odor Intensities Using UNIFAC and Henry’s Law Models. The average relative deviations (ARDH, %) between the experimental and calculated vapor-phase concentrations or odor intensities using Henry’s law were evaluated following eq 8 NP

ARDH (%) =

100 ∑ NP 1

|Aiexp

− Aipred | Aiexp

concentrations or odor intensities using the UNIFAC method were calculated according eq 9 NP

ARDUNIFAC (%) =



|Aiexp − Aipred | 100 ∑ NP 1 Aiexp

(9)

RESULTS AND DISCUSSION Theoretical and Experimental Henry’s Law Constant. Single Fragrance Systems on a Mineral Oil Simplified Matrix. The Htheor and Hexp values for single fragrance-systems (i.e., containing one odorant diluted in mineral oil) together with the relative deviations between them, the statistical parameter F value, and maximum liquid concentrations of each fragrance component where Henry’s law is valid (Cliquid,max) are presented in Table 2. Analyzing the theoretical data, it was found that the orders of magnitude of the Htheor values are

(8)

where NP corresponds to the number of experimental data points and Aiexp and Aipred are the experimental and predicted vapor-phase concentrations or odor intensities of component i, respectively, from the experimental H for single fragrance exp systems (Hi,s ). The average relative deviations (ARDUNIFAC, %) between the experimental ( Aiexp) and predicted ( Aipred ) vapor-phase F

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

experimental and theoretical H (|δH| = 100|Hexp − Htheor|/Hexp) for each single fragrance mixture are presented in Table 3. Results are very good for some odorants (e.g., 1%) and not so much for others (e.g., 94%). Moreover, the high absolute deviations observed for some cases might not be so significant considering the magnitude of the gas-phase concentrations, which are highly diluted in air due to the low range of liquid concentrations, low vapor pressure, and high molecular weight. We can even point out this as a limitation of the proposed methodology. In addition, the vapor pressure of some components considered in the present study can differ by several orders of magnitude depending on the database.36 Other factor that can contribute to the reported deviations is the activity coefficient value calculated using the VLE methodology; here, we can obtain overestimated values because the UNIFAC method does not accurately predict the vapor-phase concentration of odorants when in the presence of specific associative interactions. Finally, the experimental quantitative errors associated with component weighing and HS-GC analysis should also be considered. In the present study, the Henry’s law constants of odorants diluted in mineral oil were successfully predicted using two simple methodologies: the UNIFAC model and static HS-GC. However, it should be mentioned that Henry’s law constants can also be efficiently attained through dynamic approaches, such as, for example, using retention indices. Furthermore, we also found studies focusing on the solute−solvent intractions using gas chromatographic activity coefficients.38 In analogy to our previous study,17 no correlations were observed among the maximum liquid concentrations of each odorant where Henry’s law is valid (Cliquid,max) and its vapor pressure and molecular weight. However, it was concluded that the addition of one or more molecules in a certain system can affect the liquid concentration upper-limit value for Henry’s law validity (Figures 1 and 2). For instance, inspecting Figure 1, where the gas-phase concentration of limonene as a function of its liquid concentration in all studied systems is represented, it can be noticed that linalyl acetate and linalool account for the increase of the limonene gas-phase concentration in the ternary fragrance systems and that the Cliquid,max of limonene reduces to about half from single to binary, ternary, and quaternary fragrance systems (Tables 2 and 3).

linked with the molecular weight and vapor pressure of the components. The results are in accordance with the findings reported by our research group using the same odorants and combinations (qualitative and quantitative), but it diluted in a different simplified matrix, dipropylene glycol.17 In what concerns the Hexp values for single fragrance systems (Table 2 and Figure S2), the obtained data revealed that below a certain liquid concentration a linear relationship exists between the liquid and vapor concentration of each studied component in the liquid mixture. Except for linalool, the relationship reported above between the magnitude of Htheor values and the molecular weight and vapor pressure of odorants was also experimentally verified. Comparing the Hexp values of odorants when mixed in mineral oil and dipropylene glycol,17 some differences were found in terms of orders of magnitude. In this context, the Hexp value of α-pinene was the most affected, with values decreasing 2 orders of magnitude when diluted in mineral oil exp exp −5 −3 (Hmineral and Hdipropylene oil = 6.045 × 10 glycol = 1.497 × 10 ). This is probably due to the high affinity between this component and mineral oil. Binary, Ternary, and Quaternary Fragrance Systems on a Mineral Oil Simplified Matrix. Regarding the systems containing two, three, and four odorants diluted in mineral oil (Table 3 and Figures 1 and 2), a linear relationship was found between the experimental liquid and gas concentrations of each odorant below a certain liquid concentration, as observed for the single fragrance systems. If we compare the obtained data with those reported for dipropylene glycol observed under the same experimental conditions and for the same range of liquid concentrations,17 we can confirm that the effect of the simplified matrix on the release of odorants was higher between molecules able to express the same type of interactions. For instance, the more hydrophobic odorants, limonene, α-pinene, and p-cymene, showed more affinity with mineral oil (nonpolar molecule), thus considerably decreasing their release. For the above-mentioned nonpolar components, from Tables S3−S6, it is possible to find lower γ values in the presence of mineral oil (ranging from 1.06 to 1.40) as compared with those obtained when diluted in dipropylene glycol (ranging from 7.27 to 18.68) (Tables S7−S9).17 This means that the odorants have less affinity with dipropylene glycol and, consequently, they will be more expelled from the liquid solution to the headspace (higher vapor-phase concentrations of odorants). The effect of cosmetic matrixes on the release of fragrance chemicals has been reported in the literature. For instance, Costa et al.2 studied the release of natural fragrance components, including limonene, p-cymene, and linalyl acetate, from three cosmetic matrixes (dipropylene glycol, glycerine, and skin lotion) using dynamic headspace analysis with gas chromatography. In the presence of glycerine, authors reported an initial fast release of these molecules, in contrast with that observed when mixed with dipropylene glycol and skin lotion where lower gas concentrations were detected for these molecules. This was probably due to the high affinity between these molecules and dipropylene glycol as this simplified matrix is less hydrophilic than glycerine. Therefore, on the basis of the obtained data, and as already reported by other authors, we conclude that the vapor-phase concentration of each odorant will be different depending on the solvent in which it is diluted.2,14,36,37 The absolute relative deviations between the

Figure 3. Odor intensity as a function of their liquid concentrations for the single fragrance systems diluted in mineral oil. G

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 4. Odor intensity as a function of their liquid concentrations for the binary fragrance systems diluted in mineral oil.

obtained data for single fragrance systems, we found α-pinene and ethyl acetate as the most intense odorants as a consequence of their high volatility and psychophysical exponents n and, in the particular case of ethyl acetate, low molecular weight. For this reason, and according to the Strongest Component model, the odor space of all fragrance mixtures where α-pinene and ethyl acetate are found is dominated by these molecules. As a consequence of the affinity of limonene, α-pinene, and p-cymene to mineral oil, their odor intensities are lower as compared with those when diluted in dipropylene glycol. This highlights the influence of the matrix on the release of components, as previously noticed by other authors,2,37 and reinforces the relevance of the matrix choice depending on the fragrance raw materials to be used in a certain formulation as well as their final application. Predicted Vapor-Phase Concentrations and Odor Intensities. As mentioned previously, the vapor-phase concentration of each odorant was predicted using Henry’s law and UNIFAC models and, ultimately, converted into olfactive

In an attemp to validate the proposed Henry’s law model, the Hexp values for single fragrance systems were plotted against the Hexp for binary, ternary, and quaternary fragrance systems (Figure S3). Data showed a good linear relashionsip between the Hexp values for single fragrance systems and the Hexp values for binary (r2 = 0.999), ternary (r2 = 0.998), and quaternary (r2 = 0.997) fragrance systems. For the binary and quaternary mixtures, the vapor phase of an odorant in multicomponent fragrance systems can be accurately predicted from its measured H value when alone in the same simplified matrix. Perceived Odor. After calculating the gas concentration of each odorant above the liquid mixture, it is possible to predict the respective odor intensity using the psychophysical model Stevens’s power law (Figures 3−6). Albeit the inherent methodology constraints, as it does not consider possible interactions at the olfactory physiological level between odorants in complex mixtures,24,25 the Stevens’s power law model is considered adequate for modeling fragrance perception, as attested by previous scientific works.17,28,39,40 On the basis of the H

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 5. Odor intensity as a function of their liquid concentrations for the ternary fragrance systems diluted in mineral oil. Figure 6. Odor intensity as a function of their liquid concentrations for the quaternary fragrance systems diluted in mineral oil.

perception through the psychophysical model Stevens’s power law. The average relative deviations (ARD, %) between experimental and predicted gas concentrations or odor intensities are displayed in Table 4. For some fragrance combinations, the ARD values were very low using the Henry’s law approach; this is the case of the binary fragrance systems limonene + α-pinene and limonene + ethyl acetate with ARD values lower than 7.5. Comparing both predictive methods, the ARD values for the gas concentrations using Henry’s law were lower than those obtained using the UNIFAC method (average ARDs were 31.4 and 49.9 for Henry’s law and UNIFAC, respectively). These values were slightly higher than those obtained for the equivalent fragrance combinations diluted in dipropylene glycol (average ARDs were 17.2 and 40.5 for Henry’s law and UNIFAC, respectively).17 Ideally, validation of the odor intensities of the multicomponent mixtures diluted in mineral oil should be performed by a sensory panel. Moreover, in its absence, one way to bring some proof of validation in terms of perceived character could be through the use of the Just Noticeable Differences concept,41 that is, the minimal difference that can be detected between two similar stimuli. On the basis of that, we can consider our results acceptable

from the odor perception point of view because for some odorants only a noteworthy alteration in the headspace concentration (25−33%)42 modifies the sensory quality of a fragrance mixture.43,44 We concluded that the odor intensities of the binary fragrance systems were calculated with lower relative errors using the gas concentrations obtained using the Henry’s law approach in comparison to those from the UNIFAC model. On balance, we can point out Henry’s law, together with the psychophysical model Stevens’s power law, as an efficient tool to predict the odor intensities of multicomponent fragrance systems diluted in mineral oil. In fact, we are dealing with a methodology that, in a simple and relatively fast way, gives us relevant information about the odor space and intensity of a certain fragrance mixture.



CONCLUSIONS This work proposes Henry’s law for predicting the vapor-phase concentration of single, binary, ternary, and quaternary fragrance systems diluted in mineral oil. The UNIFAC group-contribution method was also applied to predict the vapor-phase equilibria I

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

α-pinene. Comparing both Henry’s law and UNIFAC models, the odor perception was more accurately calculated using the gas concentrations obtained through the former approach. On the basis of these findings, along with those previously reported for another simplified matrix (dipropylene glycol), we propose the extension of this attractive methodology (due to their accuracy and simplicity) to other simplified matrixes and real fragranced cosmetic formulations.

Table 4. Average Relative Deviations (ARD, %) between Experimental and Predicted Gas Concentrations or Odor Intensities Using the UNIFAC Method and ARD (%) between Experimental and Predicted Gas Concentrations or Odor Intensities Obtained from the Experimental Henry’s Law from Single Fragrance Mixturesa ARDUNIFAC Cgas limonene α-pinene linalyl acetate linalool benzyl acetate p-cymene ethyl acetate limonene α-pinene limonene linalyl acetate limonene linalool limonene benzyl acetate limonene p-cymene limonene ethyl acetate limonene α-pinene linalyl acetate limonene p-cymene linalool limonene benzyl acetate ethyl acetate limonene α-pinene p-cymene linalyl acetate limonene α-pinene p-cymene linalool limonene p-cymene benzyl acetate ethyl acetate average a

ARDH ψ

Cgas

Single Fragrance Systems 72.2 37.9 − 62.0 40.2 − 56.3 25.2 − 80.2 43.2 − 49.1 22.7 − 27.4 11.1 − 72.3 51.1 Binary Fragrance Systems 72.8 38.9 5.2 49.6 28.8 5.2 80.3 45.5 29.4 45.4 19.3 20.7 75.6 44.4 13.0 73.5 37.2 15.3 69.9 36.6 22.9 22.8 9.7 39.4 67.6 34.3 8.5 32.3 11.3 74.0 69.2 35.5 6.2 54.4 33.0 7.5 Ternary Fragrance Systems 65.8 37.4 54.3 63.5 43.5 63.4 62.5 45.9 26.5 87.2 53.3 55.3 65.2 30.9 33.4 63.8 28.3 44.4 82.7 48.1 18.8 32.1 14.0 78.3 57.1 35.1 53.1 Quaternary Fragrance Systems 57.5 27.3 36.4 88.2 64.9 77.6 13.0 4.3 115.7 23.3 7.4 177.4 58.5 27.8 38.5 89.2 66.5 79.5 9.9 3.4 82.0 53.2 23.7 112.0 67.1 33.8 11.1 6.1 2.2 82.5 36.2 15.8 36.8 51.3 30.8 9.2 49.9 25.3 31.4



ψ − − − − − −

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.iecr.7b01802. Schematic representation of the experimental procedures, experimental gas-phase concentration of fragrance components as a function of its liquid-phase concentration in mineral oil, correlation between the experimental Henry’s law constant for single fragrance systems containing one odorant and mineral oil and the experimental Henry’s law constant for multicomponent fragrance systems, description of the fragrance systems and total fragrance concentration percentages tested, UNIFAC groups and subgroups, number of groups, parameters of molecular van der Waals volumes and surface area, and liquid-phase and predicted molar compositions of the vapor phase of the fragrance mixtures (PDF)

2.0 2.6 12.4 7.0 5.8 5.0 8.6 13.1 3.1 20.8 2.2 3.7



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel.: +351 22 041 3688. Fax: +351 22 508 1674.

25.2 38.9 9.7 25.8 13.3 18.6 7.3 28.1 24.2

ORCID

Patrícia Costa: 0000-0001-9246-5611 Alírio E. Rodrigues: 0000-0002-0715-4761 Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was financially supported by Project POCI-01-0145FEDER-006984 − Associate Laboratory LSRE-LCM funded by FEDER through COMPETE2020, Programa Operacional Competitividade e Internacionalizaçaõ (POCI) − and by national funds through FCT, Fundaçaõ para a Ciência e a Tecnologia. P.C. and M.A.T. acknowledge their postdoctoral grants from the Fundaçaõ para a Ciência e a Tecnologia (SFRH/BPD/93108/2013 and SFRH/BPD/76645/2011, respectively).

12.1 52.0 30.8 42.9 12.8 54.0 26.2 29.9 4.0 23.4 12.6 4.6 12.1



REFERENCES

(1) Iwata, H.; Shimada, K. Formulas, Ingredients and Production of Cosmetics; Springer, 2013. (2) Costa, P.; Velasco, C. V.; Loureiro, J. M.; Rodrigues, A. E. Effect of Cosmetic Matrices on the Release and Odour Profiles of the Supercritical CO2 Extract of Origanum majorana L. Int. J. Cosmet. Sci. 2016, 38, 364−374. (3) Mata, V. G.; Gomes, P. B.; Rodrigues, A. E. Perfumery Ternary Diagrams (PTD): a New Concept Applied to the Optimization of Perfume Compositions. Flavour Fragrance J. 2005, 20, 465−471. (4) Teixeira, M. A.; Rodríguez, O.; Mata, V. G.; Rodrigues, A. E. Perfumery Quaternary Diagrams for Engineering Perfumes. AIChE J. 2009, 55, 2171−2185.

Values are presented as geometric averages.

of the studied fragrance systems. It was concluded that the H values were linked to the vapor pressure and molecular weight of each odorant; the highest H values were found for the most volatile molecules with the lowest molecular weights. The results highlight the relevance of odorant−matrix interactions on the odor perception of multicomponent fragrance systems. The highest odor intensity values were obtained by ethyl acetate and J

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

(28) Teixeira, M. A.; Rodríguez, O.; Rodrigues, A. E.; Selway, R. L.; Riveroll, M.; Chieffi, A. Prediction Model for the Odor Intensity of Fragrance Mixtures: A Valuable Tool for Perfumed Product Design. Ind. Eng. Chem. Res. 2013, 52, 963−971. (29) PubMed.gov, US National Library of Medicine National Institutes of Health. http://pubchem.ncbi.nlm.nih.gov/ (Accessed May 2015). (30) DIPPR 801 Database. http://www.aiche.org/dippr/eventsproducts/801-database (Accessed May 2015). (31) Chemspider Database of Chemical Structures and Property Predictions. http://www.chemspider.com/Default.aspx (Accessed May 2015). (32) Teixeira, M. A.; Rodríguez, O.; Rodrigues, A. E. The Perception of Fragrance Mixtures: A Comparison of Odor Intensity Models. AIChE J. 2010, 56, 1090−1106. (33) The Good Scents Company. http://www.thegoodscentscompany. com/ (Accessed June 2015). (34) Fredenslund, A.; Jones, R. L.; Prausnitz, J. M. GroupContribution Estimation of Activity Coefficients in Nonideal Liquid Mixtures. AIChE J. 1975, 21, 1086−1099. (35) Stevens, S. S. On the Psychophysical Law. Psychol. Rev. 1957, 64, 153−181. (36) Cometto-Muñ iz, J. E.; Cain, W. S.; Abraham, M. H. Quantification of Chemical Vapors in Chemosensory Research. Chem. Senses 2003, 28, 467−477. (37) Neuner-Jehle, N.; Etzweiler, F. The Measuring of Odors. In Perfumes - Art, Science and Technology.; Muller, P. M., Lamparsky, D, Eds.; Blackie Academic & Professional: London, 1994; pp 153−212. (38) Kováts, E. S.; Fóti, G.; Dallos, A. Solute−Solvent Interaction Parameters by Gas Chromatography. J. Chromatogr. A 2004, 1046, 185−202. (39) Teixeira, M. A.; Rodríguez, O.; Mota, F. L.; Macedo, E. A.; Rodrigues, A. E. Evaluation of Group-Contribution Methods To Predict VLE and Odor Intensity of Fragrances. Ind. Eng. Chem. Res. 2011, 50, 9390−9402. (40) Teixeira, M. A.; Barrault, L.; Rodríguez, O.; Carvalho, C. C.; Rodrigues, A. E. Perfumery Radar 2.0: A Step Toward Fragrance Design and Classification. Ind. Eng. Chem. Res. 2014, 53, 8890−8912. (41) Cain, W. Differential Sensitivity for Smell: “Noise” at the Nose. Science 1977, 195, 796−798. (42) Gamble, E. Applicability of Weber’s law to Smell. Am. J. Psychol. 1898, 10, 82−142. (43) Le Berre, E.; Béno, N.; Ishii, A.; Chabanet, C.; Etiévant, P.; Thomas-Danguin, T. Just Noticeable differences in Component Concentrations Modify the Odor Quality of a Blending Mixture. Chem. Senses 2008, 33, 389−395. (44) Muller, P.; Lamparsky, D. Perfumes - Art, Science and Technology. Springer-Science+Business Media, B. V.: The Netherlands, 1994.

(5) Spedding, P. L.; Grimshaw, J.; O’Hare, K. D. Abnormal Evaporation Rate of Ethanol from Low Concentration Aqueous Solutions. Langmuir 1993, 9, 1408−1413. (6) Ferrero Vallana, F. M.; Girling, R. P.; Nimal Gunaratne, H. Q.; Holland, L. A. M.; McNamee, P. M.; Seddon, K. R.; Stonehouse, J. R.; Todini, O. Ionic Liquids as Modulators of Fragrance Release in Consumer Goods. New J. Chem. 2016, 40, 9958−9967. (7) Ferrero Vallana, F. M.; Holland, L. A. M.; Seddon, K. R.; Todini, O. Delayed Release of a Fragrance from Novel Ionic Liquids. New J. Chem. 2017, 41, 1037−1045. (8) Teixeira, M. A.; Rodríguez, O.; Gomes, P.; Mata, V.; Rodrigues, A. E. Perfume Engineering: Design, Performance & Classification; Elsevier: Oxford, U.K., 2013. (9) Stora, T.; Escher, S.; Morris, A. The Physicochemical Basis of Perfume Performance in Consumer Products. Chimia 2001, 55, 406− 412. (10) Martel, B.; Morcellet, M.; Ruffin, D.; Vinet, F.; Weltrowski, L. Capture and Controlled Release of Fragrances by CD Finished Textiles. J. Inclusion Phenom. Mol. Recognit. Chem. 2002, 44, 439−442. (11) Miro Specos, M. M.; Escobar, G.; Marino, P.; Puggia, C.; Defain Tesoriero, M. V.; Hermida, L. Aroma Finishing of Cotton Fabrics by Means of Microencapsulation Techniques. J. Ind. Text. 2010, 40, 13− 32. (12) Haefliger, O. P.; Jeckelmann, N.; Ouali, L.; León, G. Real-Time Monitoring of Fragrance Release from Cotton Towels by Low Thermal Mass Gas Chromatography Using a Longitudinally Modulating Cryogenic System for Headspace Sampling and Injection. Anal. Chem. 2010, 82, 729−737. (13) Vuilleumier, C.; Flament, I.; Sauvegrain, P. Headspace Analysis Study of Evaporation Rate of Perfume Ingredients Applied onto Skin. Int. J. Cosmet. Sci. 1995, 17, 61−76. (14) Behan, J. M.; Perring, K. D. Perfume Interactions with Sodium Dodecyl Sulphate Solutions. Int. J. Cosmet. Sci. 1987, 9, 261−268. (15) Cortez-Pereira, C. S.; Baby, A. R.; Kaneko, T. M.; Velasco, M. V. R. Sensory Approach to Measure Fragrance Intensity on the Skin. J. Sens. Stud. 2009, 24, 871−901. (16) Baydar, A.; McGee, T.; Purzycki, K. Skin Odor Value Technology for Fragrance Performance Optimization. Perfume Flavor. 1995, 20, 45−53. (17) Costa, P.; Teixeira, M. A.; Lièvre, Y.; Loureiro, J. M.; Rodrigues, A. E. Modeling Fragrance Components Release from a Simplified Matrix Used in Toiletries and Household Products. Ind. Eng. Chem. Res. 2015, 54, 11720−11731. (18) van Gemert, L. J. Compilation of Odor Threshold Values in Air and Water; BACIS: The Netherlands, 1999. (19) van Gemert, L. J. Compilations of Odour Threshold Values in Air, Water and Other Media; Oliemans Punter & Partners BV: The Netherlands, 2003. (20) Devos, M.; Rouault, J.; Laffort, P. Standardized Olfactory Power Law Exponents; Editions Universitaires de Dijon: France, 2002. (21) Chastrette, M. Data Management in Olfaction Studies. SAR and QSAR in Environmental Research 1998, 8, 157−181. (22) Cain, W. S.; Schmidt, R. Can We Trust Odor Databases? Example of t- and n-Butyl Acetate. Atmos. Environ. 2009, 43, 2591− 2601. (23) Teixeira, M. A.; Rodríguez, O.; Rodrigues, A. E. Perfumery Radar: A Predictive Tool for Perfume Family Classification. Ind. Eng. Chem. Res. 2010, 49, 11764−11777. (24) Laing, D. G. Perceptual Odour Interactions and Objective Mixture Analysis. Food Qual. Prefer. 1994, 5, 75−80. (25) Jinks, A.; Laing, D. G. The Analysis of Odor Mixtures by Humans: Evidence for a Configurational Process. Physiol. Behav. 2001, 72, 51−63. (26) Cain, W. S.; Schiet, F. T.; Olsson, M. J.; de Wijk, R. A. Comparison of Models of Odor Interaction. Chem. Chem. Senses 1995, 20, 625−637. (27) Laffort, P.; Dravnieks, A. Several Models of Suprathreshold Quantitative Olfactory Interaction in Humans Applied to Binary, Ternary and Quaternary Mixtures. Chem. Senses 1982, 7, 153−174. K

DOI: 10.1021/acs.iecr.7b01802 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX