- Email: [email protected]
Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling
Journal of Food Engineering xxx (2016) 1e9
Contents lists available at ScienceDirect
Journal of Food Engineering journal homepage: www.elsevier.com/locate/jfoodeng
Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling M. Teles dos Santos a, *, I.S. Viana b, J.N.R. Ract c, G.A.C. Le Roux a ~o Paulo, Av. Prof. Luciano Gualberto, 380, 05508 900, Sa ~o Paulo, Brazil Department of Chemical Engineering, University of Sa ~o Paulo, Av. Prof. Lineu Prestes, 748, 05508 900, Sa ~o Paulo, Brazil Chemistry Institut, University of Sa c ~o Paulo, Av. Prof. Lineu Prestes, 580, 05508 900, Sa ~o Paulo, Brazil Faculty of Pharmaceutical Sciences, University of Sa a
b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 November 2015 Received in revised form 28 January 2016 Accepted 30 March 2016 Available online xxx
In order to evaluate thermal properties of fats and oils for product design, palm stearin, canola oil and fully hydrogenated soybean oil blends were studied. The fatty acids composition (by gas chromatography), the regiospecific distribution of saturated, monounsaturated and polyunsaturated fatty acids (by Nuclear Magnetic Resonance), the softening point and thermal transitions (by differential scanning calorimetry) were measured experimentally. Iodine value and saponification value were calculated using experimental fatty acids composition. Thermodynamic modeling (Solid-liquid Equilibrium) and computational simulations of solid-liquid transitions were used to predict the Solid Fat Content (SFC) of blends with all possible mass fraction of each oil at 0 C and 25 C and to predict changes in heat capacity of the mixtures in the whole melting range. The computational predictions were able to identify the correlation between the amount of saturated fatty acids and melting profile, offering quantitative insights for the whole ternary diagram. The experimental DSC curves, in average, showed more peaks than the predicted curves, due to the use of equilibrium hypothesis by the model and the presence of kinetics factor in experimental DSC. © 2016 Elsevier Ltd. All rights reserved.
Keywords: Vegetable oil Fats Solid-liquid equilibrium DSC Fatty acids
1. Introduction The solid-liquid phase behavior of vegetable oils influences many characteristics of fatty foods, including organoleptic properties and flavor release. Chocolate is one of the products largely influenced by this phase behavior. The final melting point of cocoa butter (CB) occurs above typical room temperature but below mouth temperature, making it solid under ambient conditions but melting in the mouth, releasing desired sensory properties, such as cooling sensation (Torbica et al., 2006). However, there are many concerns about CB shortage of supply and increasing prices. Some efforts have been made to overcome that, seeking for CB alternatives (Lipp and Anklam, 1997; Jahurul et al., 2014). A large number of natural and modified fats and oils have been studied to check their suitability to reproduce physical properties close to that of cocoa butter. These studies deal mainly with polymorphism in crystallization (Wille and Lutton, 1966; Bricknell and Hartel, 1998; ve rend et al., 2010; Miyasaki et al., 2015), Loisel et al., 1998; Le Re
* Corresponding author. E-mail address: [email protected] (M. Teles dos Santos).
kinetics (Los et al., 2002; Foubert et al., 2002) and physical measurements, such as cloud point, solid fat content, brittleness and hardness (Gregersen et al., 2015). Choose molecules or mixtures components and their ratios in order to match desired properties can be also viewed as a Computational Product Design problem, faced by other areas of study, such as polymers, solvents, fuels and lubricants (Satyanarayana et al., 2009; Chemmangattuvalappil et al., 2009; Yunus et al., 2014; Heintz et al., 2014). However, in order to use these computational approaches to narrow the search space for new mixtures using vegetable oils, it is necessary to be able to compute physical properties in solid-liquid mixtures. The nature of the solideliquid equilibrium of fats and oils is complex and imposes several challenges to describe in details the multiple solid phases that can be formed (Hjorth et al., 2014; Maximo et al., 2014). In this work we aim to estimate the solid-liquid transitions of fats and oils treating vegetable oils as triacylglycerol mixtures and predicting the distribution of such compounds in solid and liquid phases according to the thermodynamic principle of minimum Gibbs free energy. We have showed how computational tools can be useful to predict SFC of binary blends of vegetable oils, with and without interesterification (Teles dos Santos et al., 2013). This work
http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029 0260-8774/© 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
2
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
aims to compare computational predictions with experimental data for a system composed by canola oil (CO), palm stearin (PS) and fully hydrogenated soybean oil (FHSO). 2. Materials and experimental methods 2.1. Materials Formulations were prepared using palm stearin (PS) from Agropalma (Brazil), fully hydrogenated soybean oil (FHSO) from Cargill (Brazil) and canola oil (CO) from Bunge (Brazil). All reagents used in this study were of analytical grade. 2.2. Experimental methods 2.2.1. Blends preparation Fat blends formulated with CO, PS, and FHSO were prepared at different ratios, according to Table 1. Blends 1e3 represent the original components, blends 4e6 are binary blends, and blends 7e10 are ternary mixtures. The mixtures were prepared after complete melting of the fats at 70 C, kept under stirring for 10 min to ensure complete homogenization and stored under refrigeration at 5 C. 2.2.2. Fatty acids composition Fatty acid (FA) composition was determined on a Varian GC gas chromatograph (430 GC, Varian Chromatograph Systems, USA), equipped with a CP 8412 auto injector. The Galaxie software was used for quantification and identification of peaks. Injections were performed on a 100 m fused silica capillary column (ID ¼ 0.25 mm) coated with 0.2 mm of polyethylene glycol (SP-2560, Supelco, USA) using helium as the carrier gas at an isobaric pressure of 37 psi; linear velocity of 20 cm/s; the makeup gas was helium at 29 mL/ min at a split ratio of 1:50; volume injected: 1.0 mL. The injector temperature was set at 250 C and the detector temperature at 280 C. The oven temperature was initially held at 140 C (5 min) and programmed to increase to 240 C at a rate of 4 C/min, and then held isothermally for 30 min. Fatty acids (FAs) in the triacylglycerols (TAG) of the blends were converted into fatty acid methyl esters (FAME) by saponification with 0.5 mol/L methanolic sodium hydroxide, followed by refluxing with a solution of ammonium chloride and sulphuric acid in methanol, according to the methodology proposed by Menezes et al. (2013). Qualitative FA composition of the samples was determined by comparing the retention times of the peaks with those of the respective standards of the fatty acids. Quantitative composition was accomplished by area normalization, expressed as mass percent. Samples were run in triplicate and values were averaged.
Table 1 Mass fraction of the prepared blends. PS: palm stearin, CO: canola oil, FHSO: fully hidrogenated soybean oil. Blend
1 2 3 4 5 6 7 8 9 10
Mass fraction PS
CO
FHSO
1 0 0 1/2 1/2 0 1/3 2/3 1/6 1/6
0 1 0 1/2 0 1/2 1/3 1/6 2/3 1/6
0 0 1 0 1/2 1/2 1/3 1/6 1/6 2/3
2.2.3. Regiospecific distribution of fatty acids The regiospecific distribution of fatty acids on the triacylglycerol backbones was determined by proton-decoupled 13C NMR (Vlahov, 2008; Standal et al., 2009; Wang et al., 2010). Lipid samples of 250 mg were dissolved in 0.5 mL of deuterated chloroform (CDCL3) using 5-mm NMR tubes. NMR spectra were recorded on a Bruker Avance DPX spectrometer operating at 300 MHz. The 13C spectra were acquired with a spectral width of 2332.090 Hz, pulse of 10.2 ms, and relaxation delay of 30 s. Determination of 13C was performed at a frequency of 75.8 MHz with a 5 mm multinuclear probe operating at 30 C. The results show the fatty acids composition in terms of saturated fatty acids, oleic acid (monounsaturated) and linoleic þ linolenic acids (polyunsaturated). This composition is further depicted in the two possible positions in the glycerol: the middle position (sn-2) and the terminal positions (sn1,3), identified by the stereo specific number (sn). 2.2.4. Iodine value and saponification value Iodine values and saponification values were calculated based on the fatty acid composition, according to the AOCS official methods Cd 1c-85 and AOCS Cd 3a-94, respectively (AOCS, 2009). 2.2.5. Softening point Softening point was determined using the standard open tube melting point method, according to the AOCS official method Cc 325 (AOCS, 2009). This analysis was performed in triplicate. 2.2.6. Differential scanning calorimetry (DSC) The thermal behavior curves were obtained using a differential scanning calorimetry (DSC) cell on a DSC 4000 Perkin Elmer (Perkin Elmer Corp., Norwalk, CT, USA), using nitrogen and sealed aluminum capsules, each containing a sample mass of 5e10 mg. The samples were initially kept at 80 C for 10 min, then cooled at 10 C/min up to 60 C, following an isothermal time of 30 min, and finally heated to 80 C at 5 C/min, according to the AOCS Official Method Cj 1e94 (AOCS, 2009). The calibration was performed with indium (temperature of fusion: 156.6 C and heat of fusion: 28.45 J/g). Curves were processed by Pyris software, and crystallization curves analyzed for the onset of crystallization (Tonset C), peak crystallization temperatures (Tpeak C) and crystallization enthalpies (DHc J/g). 3. Modeling and simulation 3.1. Solid-liquid equilibrium modeling Vegetable oils are formed mainly by triacylglycerols (TAGs), which are molecules formed by three fatty acids sterified to a glycerol structure. Due to their high molecular weight, such compounds tend to crystallize in 3 main polymorphic forms from liquid phase (Sato, 2001). (Fig. 1). The goal is to estimate, for a given temperature, the amount of TAGs present in the solid and liquid phases. This phase equilibrium problem was solved using the second order condition for phase equilibrium (minimum of Gibbs free energy). This leads to a nonlinear programming problem (NLP) searching for the minimization of the Gibbs free energy function (G), subject to linear material balance constraints. The problem can be stated as:
min GðnÞ ¼
np nc X X i¼1 j¼1
j j
ni mi ðnÞ ¼
np X
nj g j
(1)
j¼1
s.t:
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
3
0
msolidðjÞ i;0
¼
1 solidðjÞ T DHm;i @ solidðjÞ Tm;i solidðjÞ
1 1A T
(5)
solidðjÞ
Where DHm;i and Tm;i are, respectively, the melting enthalpy and melting temperature of TAG i in solid state j. 3.2. Excess Gibbs energy model The 2-sufixe Margules model was chosen for three main reasons: 1- it is suitable for mixtures formed by molecules with similar molar volume, shape and chemical nature (Prausnitz et al., 1999); 2- An experimental database in TAGs is available allowing to compute the model interaction parameters (Aij) (Wesdorp et al., 2005) and 3- it allows flexibility/simplicity required in the optimization step. The 2-sufixe Margules equation for multicomponent mixtures is given by:
Fig. 1. Solid-liquid transitions in triacylglycerols molecules.
gE ¼
nc X nc X
Aij xi xj
; where Aij ¼ 2qaij
(6)
i¼1 j¼iþ1
ni ¼
np X
nji
i ¼ 1…nc
(2)
j¼1
j
0 ni ni i ¼ 1…nc; j ¼ 1…np
(3)
Where nc and np are the number of different triacylglycerols j (TAGs) and the number of phases in the mixture, respectively; ni j and mi represent the number of mols and the chemical potential of TAG i in phase j, respectively and ni is the total number of mols of TAG i. The triacylglycerol molecules can crystallize in 3 main crystalline states: a, b0 and b. The decision variables are the number of mols of each component i in each phase j (nji ). Activity coefficients were used to deal with non-idealities in solid phases. The intensive Gibbs energy for a phase j (g j ) is then computed using an Excess Gibbs free energy model to compute the activity coefficients, according to Equation (4):
gj ¼
nc nc X X j j j j j j xi mi ¼ > g j ¼ xi mi;0 þ RT ln gi xi i¼1
(4)
i¼1
gji
xji
Where and are the activity coefficient and molar fraction of component i on phase j, respectively, and mji;0 is the chemical potential of pure component i at the same conditions (T,P) of the mixture. For the reference state of pure liquid phase (j ¼ liquid), chemical potential is zero. For solid phases (j ¼ solid), the chemical potential of a pure component i in the solid state j in the temperature of the mixture (T) is given by:
The parameter q is a measure of molecular size in the considered pair (i,j) and xi is the molar fraction of TAG i. The parameters aij are related to interactions between TAGs i and j (Prausnitz et al., 1999). The necessary binary interaction parameters (Aij) are calculated using correlations with the fatty acids chain and position in triacylglycerols i and j (Wesdorp et al., 2005). 3.3. Melting temperature and melting enthalpy For each TAG, 6 values are required: 3 values of melting enthalpies and 3 values of melting points (each one corresponding to a solid-liquid transition, see Fig. 1). A program, developed in FORTRAN 90, includes a set of available experimental data. Due to the high number of TAGs that can be formed from few fatty acids, it is common that experimental data is not always available. In such cases, group contribution models are used (Zeberg-Mikkelsen and Stenby, 1999; Wesdorp et al., 2005). 3.4. Solution approach The NLP optimization problem was implemented in GAMS (v.23) (Rosenthal, 2008), using a Generalized Reduced Gradient Method (CONOPT 3) solver. This optimization program was then coupled (using batch files) with the main program written in FORTRAN 90, which handles the calculation of interaction parameters, melting temperature, melting enthalpy and the generation of triacylglycerols from fatty acids data. This last one is done by random distribution of fatty acids in the glycerol, generating all possible
Table 2 Experimental fatty acids composition of the blends. Fatty acid composition (% mass) Fatty acid
PS Blend 1
CO Blend 2
FHSO Blend 3
PS:CO (1/2,1/2) Blend 4
PS:FHSO (1/2,1/2) Blend 5
CO:FHSO (1/2,1/2) Blend 6
PS:CO:FHSO (1/3,1/3,1/3) Blend 7
PS:CO:FHSO (2/3,1/6,1/6) Blend 8
PS:CO:FHSO (1/6,2/3,1/6) Blend 9
PS:CO:FHSO (1/6,1/6,2/3) Blend 10
C12:0 C14:0 C16:0 C18:0 C18:1 C18:2 C18:3 Total
0.2 1.1 55.7 5.4 31.5 6.0 0.0 100.0
0.0 0.0 4.7 2.6 67.5 18.0 7.1 100.0
0.0 0.0 12.2 87.8 0.0 0.0 0.0 100.0
0.1 0.6 30.5 4.0 49.3 11.9 3.5 100.0
0.1 0.6 34.2 45.9 15.9 3.3 0.0 100.0
0.0 0.0 8.2 45.5 33.7 9.2 3.4 100.0
0.1 0.4 24.4 32.1 32.8 8.0 2.2 100.0
0.1 0.5 26.0 45.4 21.8 5.2 1.0 100.0
0.0 0.3 14.8 17.0 50.3 13.1 4.6 100.0
0.2 0.7 39.9 18.8 32.4 6.9 1.1 100.0
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
4
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
TAGs by combinatorial analysis, as described in details in our previous work (Teles dos Santos et al., 2013). When a particular polymorphic form is evaluated in the solid state, the number of mols of all TAGs in the other two polymorphic forms is set to zero in the optimization problem. All results of this work refers to the b0 solid state. The results from the optimization step are the number of mols of each TAG in each phase (solid and liquid). Thus, once the number of mols of each TAG in each phase is determined, one can compute the SFC for a given the temperature (SFC(T)) and estimate the changes in heat capacity (Cp), according to Eqs. (7) and (8), respectively, where Mi is the molar mass of TAG i and GE the extensive Excess Gibbs energy of the mixture.
Pnp1 Pnc j i¼1 ni Mi j¼1 SFCðTÞ ¼ P np Pnc j i¼1 ni Mi j¼1
app
Cp
¼ Cp þ
np nc vni vGE X X þ DHmji j vT vT j¼1 i¼1
(7)
(8)
For further details, a previous work of the authors can be consulted (Teles dos Santos et al., 2012).
4. Results 4.1. Experimental results 4.1.1. Chemical composition (fatty acids) Table 2 shows the fatty acids composition of the blends. As expected, the fat with higher degree of saturation is FHSO (100%), followed by PS (62.5%) and canola oil (7.3%). As discussed in the following steps, the degree of unsaturation (fatty acids with double bonds) will direct influence the physical properties of the blends, due to the lower melting points of triacylglycerols formed by unsaturated fatty acids, which has bended chains and lower crystalline packing. The amount of saturated fatty acids must be viewed with caution. Firstly, the triacylglycerols are the chemical compounds
responsible for the bulk physical properties, such as melting points, not individual fatty acids. Thus, the position of fatty acids in the triacylglycerols plays an important role (isomerism). TAGs with the same fatty acids can show different properties. For example, the TAG C16:0-C18:1-C16:0(palmitic-oleic-palmitic acid) has higher melting point than the TAG C16:0- C16:0-C18:1(palmitic-palmiticoleic acid). Also, blends with similar bulk fractions of saturated/ unsaturated FAs can have different fatty acids profile and, as consequence, different properties. For example, blend 7 and 10 have similar amount of saturated fatty acids (57.0% and 59.6%, respectively). However, as one can see from Table 2, the major saturated fatty acid in blend 7 is stearic acid - C18:0 (32.1%), while for blend 10 palmitic acid is the most important FA (39.9%). For these reasons, the position of fatty acids in the glycerol structure and the chemical composition expressed in triacylglycerols must be further analyzed. 4.1.2. Regiospecific distribution of fatty acids Fig. 2 shows the regiospecific distribution of fatty acids in glycerol for the blend 6, 9 and 4. One can note that, for blends 6 and 9, the partition of saturated and unsaturated fatty acids is similar regardless the glycerol position. In the other hand, this partition is quite different for blend 4. For the latter, amount of saturated FA in the middle position is about 19%, while in the terminal positions they account for more than 45%. The nonrandom distribution of fatty acids in the glycerol leads to deviations in the computational predictions, as the models use random distribution of fatty acids to predict TAG composition of the starting materials (PS, CO, FHSO). According to Buchgraber et al. (2004), the non random distribution found in vegetable oils is due to the biosynthetic pathways that plant cells use to synthesize TAGs. 4.1.3. Physical properties Table 3 shows the physical properties for each blend and for the pure oils. We can note from Table 3 that the higher the amount of unsaturations, the higher the iodine value and the lower the softening point.
Table 3 Softening point, iodine value and saponification value of the blends.
Fig. 2. Mass fraction (%) of saturated, monounsaturated and polyunsaturated fatty acids in the two positions in the glycerol for blend 6, 9 and 4.
Blend
Softening point ( C)
Iodine value
Saponification value
PS (blend 1) CO (blend 2) FHSO (blend 3) PS:CO (1/2,1/2) (blend 4) PS:FHSO (1/2,1/2) (blend 5) CO:FHSO (1/2,1/2) (blend 6) PS:CO:FHSO (1/3,1/3,1/3) (blend 7) PS:CO:FHSO (2/3,1/6,1/6) (blend 8) PS:CO:FHSO (1/6,2/3,1/6) (blend 9) PS:CO:FHSO (1/6,1/6,2/3) (blend 10)
51.10 NM 67.93 46.47
37.51 107.97 0.00 72.09
2.10 2.00 2.00 2.05
62.57
19.35
2.05
63.57
53.79
2.00
60.97
47.68
2.03
63.27
42.65
2.06
55.07
78.01
2.01
52.37
30.41
2.03
NM: not measured.
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
5
Fig. 3. Differential scanning calorimetry (DSC) for fully hydrogenated soybean oil (FHSO), palm stearin (PS) and canola oil (CO).
4.1.4. DSC The temperatures where the main phase transitions occur are observed by differential scanning calorimetry (DSC). Fig. 3 shows the experimental DSC for the pure samples. The final solid-liquid transitions is about 10 C for canola oil, 50 C for palm stearin and above 60 C for fully hydrogenated soybean oil. This is in agreement with the amount of saturated fatty acids in each sample. Exothermic peaks are observed in the PS and FHSO DSC, indicating recrystallization. One can also note that sharp peaks are observed for CO and FHSO, indicating that in these samples almost all TAGs melt in a narrow temperature range, while for PS, different peaks are observed in a temperature range from 20 C to 60 C. From the experimental DSC, one can observe that the melting
occur mainly in lower temperatures for CO (30 C to - 10 C), higher temperatures for FHSO (60 C) and over a broad temperature range for PS (10 Ce50 C), following the relative amount of saturated/unsaturated fatty acids indicated by Table 2. This potentially allows a large diversity of melting behaviors' of formulations using these fats. Fig. 4 shows binary blends and Fig. 5 a ternary blend, all with equal amounts of the starting oils. It can be observed that the addition of CO or PS does not significantly changes the final melting point of FHSO if they are in the same proportion. However, the addition of CO decreases the melting point of PS (Fig. 4). In the ternary blend showed in Fig. 5, two temperature ranges of
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
6
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
Fig. 4. Differential scanning calorimetry (DSC) for the 1:1 blends canola oil (CO) - fully hydrogenated soybean oil (FHSO), palm stearin (PS) - canola oil (CO) and palm stearin (PS) fully hydrogenated soybean oil (FHSO).
phase transitions can be clearly distinguished: one ranging from 20 C to 5 C and other from 40 C to 60 C, the latter showing two sharply separated peaks (at 49 C and 56 C). From 0 C to 20 C, an exothermic peak indicating recrystallization is observed. As observed for binary blends, the final melting point of FHSO is not significantly affected by the presence of the other two oils (PS and CO), indicating a weak interaction among triacylglycerols from FHSO and PS/CO. Indeed, the predicted TAG composition for FHSO is very different from that of PS/CO (see Supplementary Data). FHSO is composed mainly by only two saturated fatty acids (see Table 2), while the system PS/CO contains a wide variety of unsaturated TAGs.
4.2. Computational predictions Computational predictions for solid-liquid transitions for binary blends are shown in Figs. 6e8. The simulations used also blend compositions not studied experimentally (20/80 and 80/20), in order to evaluate computationally how the fraction of each oil in the blend affects the phase transitions. The model detected the influence of the more saturated fat in the final melting point. As the amount of the more saturated fat is increased (FHSO in Figs. 6 and 7 and PS in Fig. 8), the final melting point of the blend increases as well. Comparing the experimental DSC and the computationally
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
7
Fig. 5. Differential scanning calorimetry (DSC) for the blend palm stearin (PS) - canola oil (CO) - fully hydrogenated soybean oil (FHSO) 1:1:1.
Fig. 6. Simulated differential scanning calorimetry (eq-DSC) for the blend fully hydrogenated soybean oil (FHSO) - canola oil (CO) at different fractions.
Fig. 7. Simulated differential scanning calorimetry (eq-DSC) for the blend fully hydrogenated soybean oil (FHSO) - palm stearin (PS) at different fractions.
predicted phase transitions for the blend CO - FHSO 1/1 (Figs. 4 and 6), one can observe that: - The final melting temperature was correctly predicted (60 C). - The larger peak was correctly identified by the models (around 56 C).
- In the simulated curve (Fig. 6), several peaks were predicted in the range 30 C to 10 C, which correspond to the low temperatures experimental phase transitions observed in Fig. 4. - Excluding the exothermic peak, the experimental curve did not show peaks between 10 C and 30 C; the predictions correctly detected no phase transitions in this range.
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
8
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
Fig. 8. Simulated differential scanning calorimetry (eq-DSC) for the blend palm stearin (PS) - canola oil (CO) at different fractions.
Fig. 9. Predicted solid fat content (SFC) for all possible formulations using palm stearin (PS) - canola oil (CO) and fully hydrogenated soybean oil (FHSO) at 0 C (left) and 25 C (right).
- In both cases (experiments and predictions) a sharp difference between a low melting fat and a high melting fat was observed. Comparing the experimental DSC and the computationally predicted phase transitions for the blend PS-FHSO 1/1 (Figs. 4 and 7), one can observe that: - The final melting temperature was correctly predicted (60 C). - The larger peak was correctly identified by the models (around 55 C) - Excluding the exothermic peak, the experimental curve did not show peaks before 35 C; however, the model predicted two peaks in this range (at 3 C and 25 C). The presence of peaks not observed experimentally is due to the hypothesis of equilibrium used by the model: each temperature corresponds to a phase equilibrium state, which does not take into account kinetics factors or the presence of metastable states normally found in finite scan rates. Comparing the experimental DSC and the computationally predicted phase transitions for the blend PS-CO 1/1 (Figs. 4 and 8), one can observe that:
- The final melting temperature was predicted with relative agreement: 41 C by the model and 43 C experiments. - Several exothermic peaks were observed experimentally. However, the models do not take into account recrystallization and these peaks cannot be observed in simulations. - The larger peak was observed at 11 C experimentally and at 7 C by predictions. - There are several overlapped peaks in the range 8 Ce37 C (experimental). In the same range, the predictions showed several peaks. In Fig. 4, only one large peak is observed at high temperature (55 C); in the other hand, at this temperature, simulations showed two peaks (Figs. 6 and 7). Actually, they all correspond to the same high melting TAGs in the FHSO and at a finite experimental scan rate, the two peaks overlap. In order to evaluate the solid fat content (SFC) of all formulations that can be done with the three fats, the whole ternary diagram was evaluated (Fig. 9) at a low temperature (0 C) and at standard ambient conditions (25 C). Examples of evaluations in other temperatures for other mixtures can be found in another work of the authors (Teles dos Santos et al., 2014). Each diagram is achieved in about 31 min55 s in a PC Intel(R)
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
Core(TM) i7-3770 CPU 3.40 GHz, 8.00 GB RAM, Windows 7 64 bits. The model correctly predicted that the regions rich in FHSO is almost solid, in both temperatures. Also, as the temperature was increased, the SFC decreases. Although this is known a priori, a quantitative evaluation can be useful to orient experiments through the most promised formulations in terms of SFC. The region with lower SFC or at liquid state is that with higher amount of CO. As discussed earlier, this is a highly unsaturated vegetable oil, with 92.7% of low melting unsaturated fatty acids. Also, the region with higher SFC is that in which the CO fraction tends to zero, as expected. 5. Conclusions Thermodynamic modeling and computational tools (numerical optimization and data visualizing) are useful tools when coupled with experiments to access thermal properties for product design using fats and oils. The use of kinetic factors in the phase transitions modeling, improvements in the pure component properties and the use of non random models to generate triacylglycerol composition from fatty acids data are some improvements suggested. Computational approaches allow the experiments to be focused on the most promising formulations in terms of melting profile. Also, estimating the SFC value is useful to orient experiments, which can offer more accurate analysis in a smaller set of formulations. Acknowledgements Authors thank the financial support received from CAPES Coordenaç~ ao de Aperfeiçoamento de Pessoal de Nível Superior (Brazil) (23038.002924/2013-41), project CsF Jovens Talentos 057/ 2012. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jfoodeng.2016.03.029. References AOCS, 2009. Official Methods and Recommended Practices of the American Oil Chemists' Society. American Oil Chemists' Society, Champaign. Cc 3-25; Cd 1c85; Cd 3a-94; Cj 1e94. Bricknell, J., Hartel, R.W., 1998. Relation of fat bloom in chocolate to polymorphic transition of cocoa butter. J. Am. Oil Chem. Soc. 75, 1609e1615. Buchgraber, M., Ulberth, F., Emons, H., Anklam, E., 2004. Triacylglycerol profiling by using chromatographic techniques. Eur. J. Lipid Sci. Technol. 106, 621e648. Chemmangattuvalappil, N.G., Eljack, F.T., Solvason, C.C., Eden, M.R., 2009. A novel algorithm for molecular synthesis using enhanced property operators. Comput. Chem. Eng. 33, 636e643. Foubert, I., Vanrolleghem, P.A., Vanhoutte, B., Dewettinck, K., 2002. Dynamic mathematical model of the crystallization kinetics of fats. Food Res. Int. 35, 945e956. Gregersen, S.B., Miller, R.L., Hammershøj, M., Andersen, M.D., Wiking, L., 2015. Texture and microstructure of cocoa butter replacers: influence of composition and cooling rate. Food Struct. 4, 2e15. Heintz, J., Belaud, J.-P., Pandya, N., Teles dos Santos, M., Gerbaud, V., 2014. Computer aided product design tool for sustainable product development. Comput. Chem.
9
Eng. 71, 362e376. Hjorth, J.L., Miller, R., Woodley, J.M., Kill, S., 2014. Thermodynamic modeling of multi-phase solid-liquid equilibria in industrial-grade oils and fats. J. Am. Oil Chem. Soc. 92, 17e28. Jahurul, M.H.A., Zaidul, I.S.M., Norulaini, N.A.N., Sahena, F., Abedin, M.Z., Ghafoor, K., Mohd Omar, A.K., 2014. Characterization of crystallization and melting profiles of blends of mango seed fat and palm oil mid-fraction as cocoa butter replacers using differential scanning calorimetry and pulse nuclear magnetic resonance. Food Res. Int. 55, 103e109. ve rend, B.J.D., Fryer, P.J., Coles, S., Bakalis, S., 2010. A method to qualify and Le Re quantify the crystalline state of cocoa butter in industrial chocolate. J. Am. Oil Chem. Soc. 87, 239e246. Lipp, M., Anklam, E., 1997. Review of cocoa butter and alternative fats for use in chocolate - Part A. Compositional data. Food Chem. 62, 73e97. Loisel, C., Keller, G., Lecq, G., Bourgaux, C., Ollivon, M., 1998. Phase transitions and polymorphism of cocoa butter. J. Am. Oil Chem. Soc. 75, 425e439. Los, J.H., Van Enckevort, W.J.P., Vlieg, E., Floter, E., 2002. Metastable states in multicomponent liquid-solid systems I: a kinetic crystallization model. J. Phys. Chem. B 106, 7321e7330. Maximo, G.J., Costa, M.C., Coutinho, J.A.P., Meirelles, A.J.A., 2014. Trends and demands in the solid-liquid equilibrium of lipidic mixtures. RSC Adv. 4, 31840e31850. ~o Menezes, R.S., Leles, M.I.G., Soares, A.T., Franco, P.I.B.M., Filho, N.R.A., 2013. Avaliaça ria-prima graxa da potencialidade de microalgas dulcícolas como fonte de mate ~o de biodiesel. Quím. Nova 36 (1), 10e15. http://dx.doi.org/10. para a produça 1590/S0100-40422013000100003. Miyasaki, E.K., Santos, C.A., Vieira, L.R., Ming, C.C., Calligaris, G.A., Cardoso, L.P., Gonçalves, L.A.G., 2015. Acceleration of polymorphic transition of cocoa butter and cocoa butter equivalent by addition of D-limonene. Eur. J. Lipid Sci. Technol. http://dx.doi.org/10.1002/ejlt.201400557. Prausnitz, J.W., Lichtenthaler, R.N., de Azevedo, G.E., 1999. Molecular Thermodynamics of Fluid Phase Equilibria, third ed. Prentice-Hall, NewYork. Rosenthal, R.E., 2008. GAMS Release 23.2. A User's Guide. GAMS Development Corporation, Washington, DC. Sato, K., 2001. Crystallization behaviour of fats and lipids - a review. Chem. Eng. Sci. 56, 2255e2265. Satyanarayana, K.C., Abildskov, J., Gani, R., 2009. Computer-aided polymer design using group contribution plus property models. Comput. Chem. Eng. 33, 1004e1013. Standal, I.B., Axelson, D.E., Aursand, M., 2009. Differentiation of fish oils according to species by 13C-NMR regiospecific analyses of triacylglycerols. J. Am. Oil Chem. Soc. 86, 401e407. Teles dos Santos, M., Gerbaud, V., Le Roux, G.A.C., 2014. Ternary blends of vegetable oils: thermal profile predictions for product design. Comput. Aided Chem. Eng. 33, 1465e1470. Teles dos Santos, M., Gerbaud, V., Le Roux, G.A.C., 2013. Modeling and simulation fo melting curves and chemical interesterification of binary blends of vegetable oils. Chem. Eng. Sci. 87, 14e22. Teles dos Santos, M., Gerbaud, V., Le Roux, G.A.C., 2012. Comparison of predicted and experimental DSC curves for vegetable oils. Thermochim. Acta 545, 96e102. Torbica, A., Jovanovic, O., Pajin, B., 2006. The advantages of solid fat content determination in cocoa butter and cocoa butter equivalents by the Karlshamns method. Eur. Food Res. Technol. 222, 385e391. Vlahov, G., 2008. 13C NMR spectroscopy to determine structure and composition of waxes and glycerides. Curr. Pharm. Anal. v.4 3, 126e136. Wang, J., Suarez, E.R., Kralovec, J., Shahidi, F., 2010. Effect of chemical randomization on position distribution and stability of omega-3 oil triacylglycerols. J. Agric. Food Chem. 58 (15), 8842e8847. Wesdorp, L.H., et al., 2005. Liquid-multiple solid phase equilibria in fats: theory and experiments. In: Marangoni, A.G. (Ed.), Fat Crystal Networks. Marcel Dekker, New York, pp. 481e709. Wille, R.L., Lutton, E.S., 1966. Polymorphism of cocoa butter. J. Am. oil chemists' Soc. 43, 491e496. Yunus, N.A., Gernaey, K.V., Woodley, J.M., Gani, R., 2014. A systematic methodology for design of tailor-made blended products. Comput. Chem. Eng. 66, 201e213. ZebergeMikkelsen, C.K., Stendby, E.H., 1999. Predicting the melting points and the enthalpies of fusion of saturated triglycerides by a group contribution method. Fluid Phase Equilib. 162, 7e17.
́
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
Contents lists available at ScienceDirect
Journal of Food Engineering journal homepage: www.elsevier.com/locate/jfoodeng
Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling M. Teles dos Santos a, *, I.S. Viana b, J.N.R. Ract c, G.A.C. Le Roux a ~o Paulo, Av. Prof. Luciano Gualberto, 380, 05508 900, Sa ~o Paulo, Brazil Department of Chemical Engineering, University of Sa ~o Paulo, Av. Prof. Lineu Prestes, 748, 05508 900, Sa ~o Paulo, Brazil Chemistry Institut, University of Sa c ~o Paulo, Av. Prof. Lineu Prestes, 580, 05508 900, Sa ~o Paulo, Brazil Faculty of Pharmaceutical Sciences, University of Sa a
b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 November 2015 Received in revised form 28 January 2016 Accepted 30 March 2016 Available online xxx
In order to evaluate thermal properties of fats and oils for product design, palm stearin, canola oil and fully hydrogenated soybean oil blends were studied. The fatty acids composition (by gas chromatography), the regiospecific distribution of saturated, monounsaturated and polyunsaturated fatty acids (by Nuclear Magnetic Resonance), the softening point and thermal transitions (by differential scanning calorimetry) were measured experimentally. Iodine value and saponification value were calculated using experimental fatty acids composition. Thermodynamic modeling (Solid-liquid Equilibrium) and computational simulations of solid-liquid transitions were used to predict the Solid Fat Content (SFC) of blends with all possible mass fraction of each oil at 0 C and 25 C and to predict changes in heat capacity of the mixtures in the whole melting range. The computational predictions were able to identify the correlation between the amount of saturated fatty acids and melting profile, offering quantitative insights for the whole ternary diagram. The experimental DSC curves, in average, showed more peaks than the predicted curves, due to the use of equilibrium hypothesis by the model and the presence of kinetics factor in experimental DSC. © 2016 Elsevier Ltd. All rights reserved.
Keywords: Vegetable oil Fats Solid-liquid equilibrium DSC Fatty acids
1. Introduction The solid-liquid phase behavior of vegetable oils influences many characteristics of fatty foods, including organoleptic properties and flavor release. Chocolate is one of the products largely influenced by this phase behavior. The final melting point of cocoa butter (CB) occurs above typical room temperature but below mouth temperature, making it solid under ambient conditions but melting in the mouth, releasing desired sensory properties, such as cooling sensation (Torbica et al., 2006). However, there are many concerns about CB shortage of supply and increasing prices. Some efforts have been made to overcome that, seeking for CB alternatives (Lipp and Anklam, 1997; Jahurul et al., 2014). A large number of natural and modified fats and oils have been studied to check their suitability to reproduce physical properties close to that of cocoa butter. These studies deal mainly with polymorphism in crystallization (Wille and Lutton, 1966; Bricknell and Hartel, 1998; ve rend et al., 2010; Miyasaki et al., 2015), Loisel et al., 1998; Le Re
* Corresponding author. E-mail address: [email protected] (M. Teles dos Santos).
kinetics (Los et al., 2002; Foubert et al., 2002) and physical measurements, such as cloud point, solid fat content, brittleness and hardness (Gregersen et al., 2015). Choose molecules or mixtures components and their ratios in order to match desired properties can be also viewed as a Computational Product Design problem, faced by other areas of study, such as polymers, solvents, fuels and lubricants (Satyanarayana et al., 2009; Chemmangattuvalappil et al., 2009; Yunus et al., 2014; Heintz et al., 2014). However, in order to use these computational approaches to narrow the search space for new mixtures using vegetable oils, it is necessary to be able to compute physical properties in solid-liquid mixtures. The nature of the solideliquid equilibrium of fats and oils is complex and imposes several challenges to describe in details the multiple solid phases that can be formed (Hjorth et al., 2014; Maximo et al., 2014). In this work we aim to estimate the solid-liquid transitions of fats and oils treating vegetable oils as triacylglycerol mixtures and predicting the distribution of such compounds in solid and liquid phases according to the thermodynamic principle of minimum Gibbs free energy. We have showed how computational tools can be useful to predict SFC of binary blends of vegetable oils, with and without interesterification (Teles dos Santos et al., 2013). This work
http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029 0260-8774/© 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
2
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
aims to compare computational predictions with experimental data for a system composed by canola oil (CO), palm stearin (PS) and fully hydrogenated soybean oil (FHSO). 2. Materials and experimental methods 2.1. Materials Formulations were prepared using palm stearin (PS) from Agropalma (Brazil), fully hydrogenated soybean oil (FHSO) from Cargill (Brazil) and canola oil (CO) from Bunge (Brazil). All reagents used in this study were of analytical grade. 2.2. Experimental methods 2.2.1. Blends preparation Fat blends formulated with CO, PS, and FHSO were prepared at different ratios, according to Table 1. Blends 1e3 represent the original components, blends 4e6 are binary blends, and blends 7e10 are ternary mixtures. The mixtures were prepared after complete melting of the fats at 70 C, kept under stirring for 10 min to ensure complete homogenization and stored under refrigeration at 5 C. 2.2.2. Fatty acids composition Fatty acid (FA) composition was determined on a Varian GC gas chromatograph (430 GC, Varian Chromatograph Systems, USA), equipped with a CP 8412 auto injector. The Galaxie software was used for quantification and identification of peaks. Injections were performed on a 100 m fused silica capillary column (ID ¼ 0.25 mm) coated with 0.2 mm of polyethylene glycol (SP-2560, Supelco, USA) using helium as the carrier gas at an isobaric pressure of 37 psi; linear velocity of 20 cm/s; the makeup gas was helium at 29 mL/ min at a split ratio of 1:50; volume injected: 1.0 mL. The injector temperature was set at 250 C and the detector temperature at 280 C. The oven temperature was initially held at 140 C (5 min) and programmed to increase to 240 C at a rate of 4 C/min, and then held isothermally for 30 min. Fatty acids (FAs) in the triacylglycerols (TAG) of the blends were converted into fatty acid methyl esters (FAME) by saponification with 0.5 mol/L methanolic sodium hydroxide, followed by refluxing with a solution of ammonium chloride and sulphuric acid in methanol, according to the methodology proposed by Menezes et al. (2013). Qualitative FA composition of the samples was determined by comparing the retention times of the peaks with those of the respective standards of the fatty acids. Quantitative composition was accomplished by area normalization, expressed as mass percent. Samples were run in triplicate and values were averaged.
Table 1 Mass fraction of the prepared blends. PS: palm stearin, CO: canola oil, FHSO: fully hidrogenated soybean oil. Blend
1 2 3 4 5 6 7 8 9 10
Mass fraction PS
CO
FHSO
1 0 0 1/2 1/2 0 1/3 2/3 1/6 1/6
0 1 0 1/2 0 1/2 1/3 1/6 2/3 1/6
0 0 1 0 1/2 1/2 1/3 1/6 1/6 2/3
2.2.3. Regiospecific distribution of fatty acids The regiospecific distribution of fatty acids on the triacylglycerol backbones was determined by proton-decoupled 13C NMR (Vlahov, 2008; Standal et al., 2009; Wang et al., 2010). Lipid samples of 250 mg were dissolved in 0.5 mL of deuterated chloroform (CDCL3) using 5-mm NMR tubes. NMR spectra were recorded on a Bruker Avance DPX spectrometer operating at 300 MHz. The 13C spectra were acquired with a spectral width of 2332.090 Hz, pulse of 10.2 ms, and relaxation delay of 30 s. Determination of 13C was performed at a frequency of 75.8 MHz with a 5 mm multinuclear probe operating at 30 C. The results show the fatty acids composition in terms of saturated fatty acids, oleic acid (monounsaturated) and linoleic þ linolenic acids (polyunsaturated). This composition is further depicted in the two possible positions in the glycerol: the middle position (sn-2) and the terminal positions (sn1,3), identified by the stereo specific number (sn). 2.2.4. Iodine value and saponification value Iodine values and saponification values were calculated based on the fatty acid composition, according to the AOCS official methods Cd 1c-85 and AOCS Cd 3a-94, respectively (AOCS, 2009). 2.2.5. Softening point Softening point was determined using the standard open tube melting point method, according to the AOCS official method Cc 325 (AOCS, 2009). This analysis was performed in triplicate. 2.2.6. Differential scanning calorimetry (DSC) The thermal behavior curves were obtained using a differential scanning calorimetry (DSC) cell on a DSC 4000 Perkin Elmer (Perkin Elmer Corp., Norwalk, CT, USA), using nitrogen and sealed aluminum capsules, each containing a sample mass of 5e10 mg. The samples were initially kept at 80 C for 10 min, then cooled at 10 C/min up to 60 C, following an isothermal time of 30 min, and finally heated to 80 C at 5 C/min, according to the AOCS Official Method Cj 1e94 (AOCS, 2009). The calibration was performed with indium (temperature of fusion: 156.6 C and heat of fusion: 28.45 J/g). Curves were processed by Pyris software, and crystallization curves analyzed for the onset of crystallization (Tonset C), peak crystallization temperatures (Tpeak C) and crystallization enthalpies (DHc J/g). 3. Modeling and simulation 3.1. Solid-liquid equilibrium modeling Vegetable oils are formed mainly by triacylglycerols (TAGs), which are molecules formed by three fatty acids sterified to a glycerol structure. Due to their high molecular weight, such compounds tend to crystallize in 3 main polymorphic forms from liquid phase (Sato, 2001). (Fig. 1). The goal is to estimate, for a given temperature, the amount of TAGs present in the solid and liquid phases. This phase equilibrium problem was solved using the second order condition for phase equilibrium (minimum of Gibbs free energy). This leads to a nonlinear programming problem (NLP) searching for the minimization of the Gibbs free energy function (G), subject to linear material balance constraints. The problem can be stated as:
min GðnÞ ¼
np nc X X i¼1 j¼1
j j
ni mi ðnÞ ¼
np X
nj g j
(1)
j¼1
s.t:
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
3
0
msolidðjÞ i;0
¼
1 solidðjÞ T DHm;i @ solidðjÞ Tm;i solidðjÞ
1 1A T
(5)
solidðjÞ
Where DHm;i and Tm;i are, respectively, the melting enthalpy and melting temperature of TAG i in solid state j. 3.2. Excess Gibbs energy model The 2-sufixe Margules model was chosen for three main reasons: 1- it is suitable for mixtures formed by molecules with similar molar volume, shape and chemical nature (Prausnitz et al., 1999); 2- An experimental database in TAGs is available allowing to compute the model interaction parameters (Aij) (Wesdorp et al., 2005) and 3- it allows flexibility/simplicity required in the optimization step. The 2-sufixe Margules equation for multicomponent mixtures is given by:
Fig. 1. Solid-liquid transitions in triacylglycerols molecules.
gE ¼
nc X nc X
Aij xi xj
; where Aij ¼ 2qaij
(6)
i¼1 j¼iþ1
ni ¼
np X
nji
i ¼ 1…nc
(2)
j¼1
j
0 ni ni i ¼ 1…nc; j ¼ 1…np
(3)
Where nc and np are the number of different triacylglycerols j (TAGs) and the number of phases in the mixture, respectively; ni j and mi represent the number of mols and the chemical potential of TAG i in phase j, respectively and ni is the total number of mols of TAG i. The triacylglycerol molecules can crystallize in 3 main crystalline states: a, b0 and b. The decision variables are the number of mols of each component i in each phase j (nji ). Activity coefficients were used to deal with non-idealities in solid phases. The intensive Gibbs energy for a phase j (g j ) is then computed using an Excess Gibbs free energy model to compute the activity coefficients, according to Equation (4):
gj ¼
nc nc X X j j j j j j xi mi ¼ > g j ¼ xi mi;0 þ RT ln gi xi i¼1
(4)
i¼1
gji
xji
Where and are the activity coefficient and molar fraction of component i on phase j, respectively, and mji;0 is the chemical potential of pure component i at the same conditions (T,P) of the mixture. For the reference state of pure liquid phase (j ¼ liquid), chemical potential is zero. For solid phases (j ¼ solid), the chemical potential of a pure component i in the solid state j in the temperature of the mixture (T) is given by:
The parameter q is a measure of molecular size in the considered pair (i,j) and xi is the molar fraction of TAG i. The parameters aij are related to interactions between TAGs i and j (Prausnitz et al., 1999). The necessary binary interaction parameters (Aij) are calculated using correlations with the fatty acids chain and position in triacylglycerols i and j (Wesdorp et al., 2005). 3.3. Melting temperature and melting enthalpy For each TAG, 6 values are required: 3 values of melting enthalpies and 3 values of melting points (each one corresponding to a solid-liquid transition, see Fig. 1). A program, developed in FORTRAN 90, includes a set of available experimental data. Due to the high number of TAGs that can be formed from few fatty acids, it is common that experimental data is not always available. In such cases, group contribution models are used (Zeberg-Mikkelsen and Stenby, 1999; Wesdorp et al., 2005). 3.4. Solution approach The NLP optimization problem was implemented in GAMS (v.23) (Rosenthal, 2008), using a Generalized Reduced Gradient Method (CONOPT 3) solver. This optimization program was then coupled (using batch files) with the main program written in FORTRAN 90, which handles the calculation of interaction parameters, melting temperature, melting enthalpy and the generation of triacylglycerols from fatty acids data. This last one is done by random distribution of fatty acids in the glycerol, generating all possible
Table 2 Experimental fatty acids composition of the blends. Fatty acid composition (% mass) Fatty acid
PS Blend 1
CO Blend 2
FHSO Blend 3
PS:CO (1/2,1/2) Blend 4
PS:FHSO (1/2,1/2) Blend 5
CO:FHSO (1/2,1/2) Blend 6
PS:CO:FHSO (1/3,1/3,1/3) Blend 7
PS:CO:FHSO (2/3,1/6,1/6) Blend 8
PS:CO:FHSO (1/6,2/3,1/6) Blend 9
PS:CO:FHSO (1/6,1/6,2/3) Blend 10
C12:0 C14:0 C16:0 C18:0 C18:1 C18:2 C18:3 Total
0.2 1.1 55.7 5.4 31.5 6.0 0.0 100.0
0.0 0.0 4.7 2.6 67.5 18.0 7.1 100.0
0.0 0.0 12.2 87.8 0.0 0.0 0.0 100.0
0.1 0.6 30.5 4.0 49.3 11.9 3.5 100.0
0.1 0.6 34.2 45.9 15.9 3.3 0.0 100.0
0.0 0.0 8.2 45.5 33.7 9.2 3.4 100.0
0.1 0.4 24.4 32.1 32.8 8.0 2.2 100.0
0.1 0.5 26.0 45.4 21.8 5.2 1.0 100.0
0.0 0.3 14.8 17.0 50.3 13.1 4.6 100.0
0.2 0.7 39.9 18.8 32.4 6.9 1.1 100.0
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
4
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
TAGs by combinatorial analysis, as described in details in our previous work (Teles dos Santos et al., 2013). When a particular polymorphic form is evaluated in the solid state, the number of mols of all TAGs in the other two polymorphic forms is set to zero in the optimization problem. All results of this work refers to the b0 solid state. The results from the optimization step are the number of mols of each TAG in each phase (solid and liquid). Thus, once the number of mols of each TAG in each phase is determined, one can compute the SFC for a given the temperature (SFC(T)) and estimate the changes in heat capacity (Cp), according to Eqs. (7) and (8), respectively, where Mi is the molar mass of TAG i and GE the extensive Excess Gibbs energy of the mixture.
Pnp1 Pnc j i¼1 ni Mi j¼1 SFCðTÞ ¼ P np Pnc j i¼1 ni Mi j¼1
app
Cp
¼ Cp þ
np nc vni vGE X X þ DHmji j vT vT j¼1 i¼1
(7)
(8)
For further details, a previous work of the authors can be consulted (Teles dos Santos et al., 2012).
4. Results 4.1. Experimental results 4.1.1. Chemical composition (fatty acids) Table 2 shows the fatty acids composition of the blends. As expected, the fat with higher degree of saturation is FHSO (100%), followed by PS (62.5%) and canola oil (7.3%). As discussed in the following steps, the degree of unsaturation (fatty acids with double bonds) will direct influence the physical properties of the blends, due to the lower melting points of triacylglycerols formed by unsaturated fatty acids, which has bended chains and lower crystalline packing. The amount of saturated fatty acids must be viewed with caution. Firstly, the triacylglycerols are the chemical compounds
responsible for the bulk physical properties, such as melting points, not individual fatty acids. Thus, the position of fatty acids in the triacylglycerols plays an important role (isomerism). TAGs with the same fatty acids can show different properties. For example, the TAG C16:0-C18:1-C16:0(palmitic-oleic-palmitic acid) has higher melting point than the TAG C16:0- C16:0-C18:1(palmitic-palmiticoleic acid). Also, blends with similar bulk fractions of saturated/ unsaturated FAs can have different fatty acids profile and, as consequence, different properties. For example, blend 7 and 10 have similar amount of saturated fatty acids (57.0% and 59.6%, respectively). However, as one can see from Table 2, the major saturated fatty acid in blend 7 is stearic acid - C18:0 (32.1%), while for blend 10 palmitic acid is the most important FA (39.9%). For these reasons, the position of fatty acids in the glycerol structure and the chemical composition expressed in triacylglycerols must be further analyzed. 4.1.2. Regiospecific distribution of fatty acids Fig. 2 shows the regiospecific distribution of fatty acids in glycerol for the blend 6, 9 and 4. One can note that, for blends 6 and 9, the partition of saturated and unsaturated fatty acids is similar regardless the glycerol position. In the other hand, this partition is quite different for blend 4. For the latter, amount of saturated FA in the middle position is about 19%, while in the terminal positions they account for more than 45%. The nonrandom distribution of fatty acids in the glycerol leads to deviations in the computational predictions, as the models use random distribution of fatty acids to predict TAG composition of the starting materials (PS, CO, FHSO). According to Buchgraber et al. (2004), the non random distribution found in vegetable oils is due to the biosynthetic pathways that plant cells use to synthesize TAGs. 4.1.3. Physical properties Table 3 shows the physical properties for each blend and for the pure oils. We can note from Table 3 that the higher the amount of unsaturations, the higher the iodine value and the lower the softening point.
Table 3 Softening point, iodine value and saponification value of the blends.
Fig. 2. Mass fraction (%) of saturated, monounsaturated and polyunsaturated fatty acids in the two positions in the glycerol for blend 6, 9 and 4.
Blend
Softening point ( C)
Iodine value
Saponification value
PS (blend 1) CO (blend 2) FHSO (blend 3) PS:CO (1/2,1/2) (blend 4) PS:FHSO (1/2,1/2) (blend 5) CO:FHSO (1/2,1/2) (blend 6) PS:CO:FHSO (1/3,1/3,1/3) (blend 7) PS:CO:FHSO (2/3,1/6,1/6) (blend 8) PS:CO:FHSO (1/6,2/3,1/6) (blend 9) PS:CO:FHSO (1/6,1/6,2/3) (blend 10)
51.10 NM 67.93 46.47
37.51 107.97 0.00 72.09
2.10 2.00 2.00 2.05
62.57
19.35
2.05
63.57
53.79
2.00
60.97
47.68
2.03
63.27
42.65
2.06
55.07
78.01
2.01
52.37
30.41
2.03
NM: not measured.
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
5
Fig. 3. Differential scanning calorimetry (DSC) for fully hydrogenated soybean oil (FHSO), palm stearin (PS) and canola oil (CO).
4.1.4. DSC The temperatures where the main phase transitions occur are observed by differential scanning calorimetry (DSC). Fig. 3 shows the experimental DSC for the pure samples. The final solid-liquid transitions is about 10 C for canola oil, 50 C for palm stearin and above 60 C for fully hydrogenated soybean oil. This is in agreement with the amount of saturated fatty acids in each sample. Exothermic peaks are observed in the PS and FHSO DSC, indicating recrystallization. One can also note that sharp peaks are observed for CO and FHSO, indicating that in these samples almost all TAGs melt in a narrow temperature range, while for PS, different peaks are observed in a temperature range from 20 C to 60 C. From the experimental DSC, one can observe that the melting
occur mainly in lower temperatures for CO (30 C to - 10 C), higher temperatures for FHSO (60 C) and over a broad temperature range for PS (10 Ce50 C), following the relative amount of saturated/unsaturated fatty acids indicated by Table 2. This potentially allows a large diversity of melting behaviors' of formulations using these fats. Fig. 4 shows binary blends and Fig. 5 a ternary blend, all with equal amounts of the starting oils. It can be observed that the addition of CO or PS does not significantly changes the final melting point of FHSO if they are in the same proportion. However, the addition of CO decreases the melting point of PS (Fig. 4). In the ternary blend showed in Fig. 5, two temperature ranges of
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
6
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
Fig. 4. Differential scanning calorimetry (DSC) for the 1:1 blends canola oil (CO) - fully hydrogenated soybean oil (FHSO), palm stearin (PS) - canola oil (CO) and palm stearin (PS) fully hydrogenated soybean oil (FHSO).
phase transitions can be clearly distinguished: one ranging from 20 C to 5 C and other from 40 C to 60 C, the latter showing two sharply separated peaks (at 49 C and 56 C). From 0 C to 20 C, an exothermic peak indicating recrystallization is observed. As observed for binary blends, the final melting point of FHSO is not significantly affected by the presence of the other two oils (PS and CO), indicating a weak interaction among triacylglycerols from FHSO and PS/CO. Indeed, the predicted TAG composition for FHSO is very different from that of PS/CO (see Supplementary Data). FHSO is composed mainly by only two saturated fatty acids (see Table 2), while the system PS/CO contains a wide variety of unsaturated TAGs.
4.2. Computational predictions Computational predictions for solid-liquid transitions for binary blends are shown in Figs. 6e8. The simulations used also blend compositions not studied experimentally (20/80 and 80/20), in order to evaluate computationally how the fraction of each oil in the blend affects the phase transitions. The model detected the influence of the more saturated fat in the final melting point. As the amount of the more saturated fat is increased (FHSO in Figs. 6 and 7 and PS in Fig. 8), the final melting point of the blend increases as well. Comparing the experimental DSC and the computationally
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
7
Fig. 5. Differential scanning calorimetry (DSC) for the blend palm stearin (PS) - canola oil (CO) - fully hydrogenated soybean oil (FHSO) 1:1:1.
Fig. 6. Simulated differential scanning calorimetry (eq-DSC) for the blend fully hydrogenated soybean oil (FHSO) - canola oil (CO) at different fractions.
Fig. 7. Simulated differential scanning calorimetry (eq-DSC) for the blend fully hydrogenated soybean oil (FHSO) - palm stearin (PS) at different fractions.
predicted phase transitions for the blend CO - FHSO 1/1 (Figs. 4 and 6), one can observe that: - The final melting temperature was correctly predicted (60 C). - The larger peak was correctly identified by the models (around 56 C).
- In the simulated curve (Fig. 6), several peaks were predicted in the range 30 C to 10 C, which correspond to the low temperatures experimental phase transitions observed in Fig. 4. - Excluding the exothermic peak, the experimental curve did not show peaks between 10 C and 30 C; the predictions correctly detected no phase transitions in this range.
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
8
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
Fig. 8. Simulated differential scanning calorimetry (eq-DSC) for the blend palm stearin (PS) - canola oil (CO) at different fractions.
Fig. 9. Predicted solid fat content (SFC) for all possible formulations using palm stearin (PS) - canola oil (CO) and fully hydrogenated soybean oil (FHSO) at 0 C (left) and 25 C (right).
- In both cases (experiments and predictions) a sharp difference between a low melting fat and a high melting fat was observed. Comparing the experimental DSC and the computationally predicted phase transitions for the blend PS-FHSO 1/1 (Figs. 4 and 7), one can observe that: - The final melting temperature was correctly predicted (60 C). - The larger peak was correctly identified by the models (around 55 C) - Excluding the exothermic peak, the experimental curve did not show peaks before 35 C; however, the model predicted two peaks in this range (at 3 C and 25 C). The presence of peaks not observed experimentally is due to the hypothesis of equilibrium used by the model: each temperature corresponds to a phase equilibrium state, which does not take into account kinetics factors or the presence of metastable states normally found in finite scan rates. Comparing the experimental DSC and the computationally predicted phase transitions for the blend PS-CO 1/1 (Figs. 4 and 8), one can observe that:
- The final melting temperature was predicted with relative agreement: 41 C by the model and 43 C experiments. - Several exothermic peaks were observed experimentally. However, the models do not take into account recrystallization and these peaks cannot be observed in simulations. - The larger peak was observed at 11 C experimentally and at 7 C by predictions. - There are several overlapped peaks in the range 8 Ce37 C (experimental). In the same range, the predictions showed several peaks. In Fig. 4, only one large peak is observed at high temperature (55 C); in the other hand, at this temperature, simulations showed two peaks (Figs. 6 and 7). Actually, they all correspond to the same high melting TAGs in the FHSO and at a finite experimental scan rate, the two peaks overlap. In order to evaluate the solid fat content (SFC) of all formulations that can be done with the three fats, the whole ternary diagram was evaluated (Fig. 9) at a low temperature (0 C) and at standard ambient conditions (25 C). Examples of evaluations in other temperatures for other mixtures can be found in another work of the authors (Teles dos Santos et al., 2014). Each diagram is achieved in about 31 min55 s in a PC Intel(R)
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029
M. Teles dos Santos et al. / Journal of Food Engineering xxx (2016) 1e9
Core(TM) i7-3770 CPU 3.40 GHz, 8.00 GB RAM, Windows 7 64 bits. The model correctly predicted that the regions rich in FHSO is almost solid, in both temperatures. Also, as the temperature was increased, the SFC decreases. Although this is known a priori, a quantitative evaluation can be useful to orient experiments through the most promised formulations in terms of SFC. The region with lower SFC or at liquid state is that with higher amount of CO. As discussed earlier, this is a highly unsaturated vegetable oil, with 92.7% of low melting unsaturated fatty acids. Also, the region with higher SFC is that in which the CO fraction tends to zero, as expected. 5. Conclusions Thermodynamic modeling and computational tools (numerical optimization and data visualizing) are useful tools when coupled with experiments to access thermal properties for product design using fats and oils. The use of kinetic factors in the phase transitions modeling, improvements in the pure component properties and the use of non random models to generate triacylglycerol composition from fatty acids data are some improvements suggested. Computational approaches allow the experiments to be focused on the most promising formulations in terms of melting profile. Also, estimating the SFC value is useful to orient experiments, which can offer more accurate analysis in a smaller set of formulations. Acknowledgements Authors thank the financial support received from CAPES Coordenaç~ ao de Aperfeiçoamento de Pessoal de Nível Superior (Brazil) (23038.002924/2013-41), project CsF Jovens Talentos 057/ 2012. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jfoodeng.2016.03.029. References AOCS, 2009. Official Methods and Recommended Practices of the American Oil Chemists' Society. American Oil Chemists' Society, Champaign. Cc 3-25; Cd 1c85; Cd 3a-94; Cj 1e94. Bricknell, J., Hartel, R.W., 1998. Relation of fat bloom in chocolate to polymorphic transition of cocoa butter. J. Am. Oil Chem. Soc. 75, 1609e1615. Buchgraber, M., Ulberth, F., Emons, H., Anklam, E., 2004. Triacylglycerol profiling by using chromatographic techniques. Eur. J. Lipid Sci. Technol. 106, 621e648. Chemmangattuvalappil, N.G., Eljack, F.T., Solvason, C.C., Eden, M.R., 2009. A novel algorithm for molecular synthesis using enhanced property operators. Comput. Chem. Eng. 33, 636e643. Foubert, I., Vanrolleghem, P.A., Vanhoutte, B., Dewettinck, K., 2002. Dynamic mathematical model of the crystallization kinetics of fats. Food Res. Int. 35, 945e956. Gregersen, S.B., Miller, R.L., Hammershøj, M., Andersen, M.D., Wiking, L., 2015. Texture and microstructure of cocoa butter replacers: influence of composition and cooling rate. Food Struct. 4, 2e15. Heintz, J., Belaud, J.-P., Pandya, N., Teles dos Santos, M., Gerbaud, V., 2014. Computer aided product design tool for sustainable product development. Comput. Chem.
9
Eng. 71, 362e376. Hjorth, J.L., Miller, R., Woodley, J.M., Kill, S., 2014. Thermodynamic modeling of multi-phase solid-liquid equilibria in industrial-grade oils and fats. J. Am. Oil Chem. Soc. 92, 17e28. Jahurul, M.H.A., Zaidul, I.S.M., Norulaini, N.A.N., Sahena, F., Abedin, M.Z., Ghafoor, K., Mohd Omar, A.K., 2014. Characterization of crystallization and melting profiles of blends of mango seed fat and palm oil mid-fraction as cocoa butter replacers using differential scanning calorimetry and pulse nuclear magnetic resonance. Food Res. Int. 55, 103e109. ve rend, B.J.D., Fryer, P.J., Coles, S., Bakalis, S., 2010. A method to qualify and Le Re quantify the crystalline state of cocoa butter in industrial chocolate. J. Am. Oil Chem. Soc. 87, 239e246. Lipp, M., Anklam, E., 1997. Review of cocoa butter and alternative fats for use in chocolate - Part A. Compositional data. Food Chem. 62, 73e97. Loisel, C., Keller, G., Lecq, G., Bourgaux, C., Ollivon, M., 1998. Phase transitions and polymorphism of cocoa butter. J. Am. Oil Chem. Soc. 75, 425e439. Los, J.H., Van Enckevort, W.J.P., Vlieg, E., Floter, E., 2002. Metastable states in multicomponent liquid-solid systems I: a kinetic crystallization model. J. Phys. Chem. B 106, 7321e7330. Maximo, G.J., Costa, M.C., Coutinho, J.A.P., Meirelles, A.J.A., 2014. Trends and demands in the solid-liquid equilibrium of lipidic mixtures. RSC Adv. 4, 31840e31850. ~o Menezes, R.S., Leles, M.I.G., Soares, A.T., Franco, P.I.B.M., Filho, N.R.A., 2013. Avaliaça ria-prima graxa da potencialidade de microalgas dulcícolas como fonte de mate ~o de biodiesel. Quím. Nova 36 (1), 10e15. http://dx.doi.org/10. para a produça 1590/S0100-40422013000100003. Miyasaki, E.K., Santos, C.A., Vieira, L.R., Ming, C.C., Calligaris, G.A., Cardoso, L.P., Gonçalves, L.A.G., 2015. Acceleration of polymorphic transition of cocoa butter and cocoa butter equivalent by addition of D-limonene. Eur. J. Lipid Sci. Technol. http://dx.doi.org/10.1002/ejlt.201400557. Prausnitz, J.W., Lichtenthaler, R.N., de Azevedo, G.E., 1999. Molecular Thermodynamics of Fluid Phase Equilibria, third ed. Prentice-Hall, NewYork. Rosenthal, R.E., 2008. GAMS Release 23.2. A User's Guide. GAMS Development Corporation, Washington, DC. Sato, K., 2001. Crystallization behaviour of fats and lipids - a review. Chem. Eng. Sci. 56, 2255e2265. Satyanarayana, K.C., Abildskov, J., Gani, R., 2009. Computer-aided polymer design using group contribution plus property models. Comput. Chem. Eng. 33, 1004e1013. Standal, I.B., Axelson, D.E., Aursand, M., 2009. Differentiation of fish oils according to species by 13C-NMR regiospecific analyses of triacylglycerols. J. Am. Oil Chem. Soc. 86, 401e407. Teles dos Santos, M., Gerbaud, V., Le Roux, G.A.C., 2014. Ternary blends of vegetable oils: thermal profile predictions for product design. Comput. Aided Chem. Eng. 33, 1465e1470. Teles dos Santos, M., Gerbaud, V., Le Roux, G.A.C., 2013. Modeling and simulation fo melting curves and chemical interesterification of binary blends of vegetable oils. Chem. Eng. Sci. 87, 14e22. Teles dos Santos, M., Gerbaud, V., Le Roux, G.A.C., 2012. Comparison of predicted and experimental DSC curves for vegetable oils. Thermochim. Acta 545, 96e102. Torbica, A., Jovanovic, O., Pajin, B., 2006. The advantages of solid fat content determination in cocoa butter and cocoa butter equivalents by the Karlshamns method. Eur. Food Res. Technol. 222, 385e391. Vlahov, G., 2008. 13C NMR spectroscopy to determine structure and composition of waxes and glycerides. Curr. Pharm. Anal. v.4 3, 126e136. Wang, J., Suarez, E.R., Kralovec, J., Shahidi, F., 2010. Effect of chemical randomization on position distribution and stability of omega-3 oil triacylglycerols. J. Agric. Food Chem. 58 (15), 8842e8847. Wesdorp, L.H., et al., 2005. Liquid-multiple solid phase equilibria in fats: theory and experiments. In: Marangoni, A.G. (Ed.), Fat Crystal Networks. Marcel Dekker, New York, pp. 481e709. Wille, R.L., Lutton, E.S., 1966. Polymorphism of cocoa butter. J. Am. oil chemists' Soc. 43, 491e496. Yunus, N.A., Gernaey, K.V., Woodley, J.M., Gani, R., 2014. A systematic methodology for design of tailor-made blended products. Comput. Chem. Eng. 66, 201e213. ZebergeMikkelsen, C.K., Stendby, E.H., 1999. Predicting the melting points and the enthalpies of fusion of saturated triglycerides by a group contribution method. Fluid Phase Equilib. 162, 7e17.
́
Please cite this article in press as: Teles dos Santos, M., et al., Thermal properties of palm stearin, canola oil and fully hydrogenated soybean oil blends: Coupling experiments and modeling, Journal of Food Engineering (2016), http://dx.doi.org/10.1016/j.jfoodeng.2016.03.029