Crystallization of cocoa butter with and without polar lipids evaluated by rheometry, calorimetry and polarized light microscopy

Jorge F. Toro-Vazqueza

Esmeralda Rangel-Vargasb

Elena Dibildox-Alvaradoa

Miriam A. Char�-Alonsoa

a Facultad de CienciasQu�micas-CIEP,Universidad Aut�noma de SanLuis Potos�,San Luis Potos�, Mexico

b Facultad de Qu�mica-DIPA(PROPAC),Universidad Aut�noma deQuer�taro,Quer�taro, Mexico

Crystallization of cocoa butter with and withoutpolar lipids evaluated by rheometry, calorimetry andpolarized light microscopy

In this investigation, we evaluated the crystallization process of triacylglycerols (TAG) inunrefined cocoa butter (CB) and in CB without polar lipids (CB-WPL), under non-iso-thermal and isothermal conditions. These conditions were obtained by cooling from80 7C at a cooling rate of 1 7C/min until achieving particular crystallization temperatures(i.e., 18.5 7C, 19.0 7C, 19.5 7C, and 20 7C), at which isothermal crystallization studieswere done. Phase shift angle rheograms (d) showed that the onset of crystallizationduring the non-isothermal stage, independently of the crystallization temperature (TCr)used, was 59.8 6 3.7 7C in CB and 39.5 6 1.5 7C in CB-WPL. These results pointed outthe nucleating role of phosphatidylcholine (26.53 6 0.04%) and phosphatidylethano-lamine (57.14 6 0.07%), the main phospholipids present in the CB used in this inves-tigation. Under similar crystallization conditions, differential scanning calorimetry(DSC) provided an onset of crystallization during the non-isothermal stage of15.0 6 0.0 7C in CB-WPL and of 15.5 6 0.1 7C in CB. Then, d rheograms were moresensitive to detect the nucleating role of phospholipids than DSC. Under isothermalconditions, both DSC and d rheograms showed that shorter times were needed in CB-WPL to complete crystallization than in CB. In the same way, at all TCr investigated, ahigher crystallization rate was achieved in CB-WPL than in CB, this as measured by thecrystallization rate constant (z) of the Avrami equation. These results showed thatcrystallization of a and b’ polymorphs took longer in CB than in CB-WPL, pointing outthat polar lipids delayed the a-to-b’ polymorphic transition. Additional results con-firmed that polar lipids affect the kinetics of TAG crystallization in CB. However, polarlipids do not affect the thermodynamic properties of the crystals [i.e., heat of fusion(DHM), temperature of fusion (TM’)], the storage modulus at the end of the crystallizationprocess (i.e., G’ of pseudo-equilibrium), or the mechanism of crystal growth (i.e., theAvrami index, n).

Keywords: Cocoa butter, rheology, crystallization, phospholipids, calorimetry.

1 Introduction

Three different events are involved during isothermalcrystallization of triacylglycerols (TAG) in vegetable oils orfats: the induction of crystallization (i.e., nucleation),crystal growth, and crystal perfection or ripening. How-ever, appropriate thermodynamic conditions must exist tostart nucleation. Thus, in a vegetable oil, TAG below theirmelting temperature (i.e., under supercooling conditions)decrease their free energy by undergoing nucleation.Nucleation is considered as the result of the addition ofmonomers (i.e., TAG) to aggregates (i.e., lamellas of TAG),the sizes of which must achieve a critical size before a

stable nucleus is attained. Once a stable nucleus isformed, crystal growth occurs. However, all these eventsoccur simultaneously at different rates, as affected by thecrystallization conditions such as supercooling, coolingrate, viscosity of the melt, and stirring. Additionally, it hasbecome evident that minor components of fats and oils,e.g., phospholipids, di-, and mono-acylglycerols, have aneffect on the crystallization process of TAG and, subse-quently, on the crystallization of vegetable oils and fats [1,2]. Thus, the effect of di-acylglycerols on the crystal-lization of milk fat TAG, particularly in the induction time ofcrystallization, has been studied by polarizedmicroscopy,turbidity, light scattering, and pulsed nuclear magneticresonance [3]. In the same way, the effect of di-acylgly-cerols on the crystal growth mechanism and on the fractaldimension of milk fat TAG has been studied by solid fatcontent and rheology measurements [4]. In the particularcase of cocoa butter (CB), Arruda and Dimick [5] showed

Correspondence: Jorge F. Toro-Vazquez, Facultad de CienciasQu�micas-CIEP, Av. Dr. Manuel Nava 6, Zona Universitaria, SanLuis Potos�, SLP 78210, Mexico. Phone: 152 444 8262450, Fax:152 444 8262371/72, e-mail: [email protected]

Eur. J. Lipid Sci. Technol. 107 (2005) 641–655 DOI 10.1002/ejlt.200501163 641

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

Res

earchPap

er

642 J. F. Toro-Vazquez et al. Eur. J. Lipid Sci. Technol. 107 (2005) 641–655

that lipid seed crystals developed at 26.5 7C have 12-foldmore phospholipids (3.99%) compared to the concentra-tion originally present in CB (0.34%). This study sug-gested that phospholipids, particularly phosphatidylcho-line and phosphatidylethanolamine, develop liquid crys-tals providing the initial nucleus for further TAGcrystallization. The same authors associated the nuclea-tion rate of CB with the polarity of phospholipids [5]. Thus,high concentrations of polar phospholipids such as lyso-phosphatidylcholine and phosphatidylinositol were asso-ciated with slow-nucleating CB, whereas the rapid-nucleating CB had lower concentrations of these phos-pholipids and high concentrations of phosphatidylcholine[6]. Additional results showed that removal of phospholi-pids by refining slowed the CB crystallization rate andincreased the induction time of crystallization in compar-ison with native CB, suggesting that phospholipids areneeded to provide the nucleus for CB crystallization [7].However, these crystallization studies used mainly ab-sorbance measurements at 500 nm, a technique of lim-ited resolution to follow TAG nucleation [3]. The effect ofminor constituents (i.e., polar lipids) on crystallization ofCB must be investigated using methodologies that allowa better characterization of both the nucleation and crys-tallization kinetics of TAG under different supercoolingconditions.

CB is a vegetable oil of functional and technologicalimportance since it constitutes the main lipid phase inchocolate. CB is also used in combination with othervegetable oils, such as hydrogenated soybean oil andpalm oil fractions, in the elaboration of confectioneryproducts. However, the relatively simple TAG compositionof CB with its diverse TAG conformations results in acomplex polymorphism, with at least six polymorphicforms [8, 9]. Crystallization of CB in the appropriate poly-morphic state (i.e., type V or b2) provides the character-istic texture, mouthfeel and flavor release of chocolate.

Within this framework, the aim of this research was toevaluate the crystallization process of TAG in native CBand in CB free of polar lipids, under non-isothermal andisothermal conditions. For this investigation, we con-sidered polar lipids as mono- and di-acylglycerols, andphospholipids. The analytical methodologies used toevaluate TAG crystallization in CB were differential scan-ning calorimetry (DSC), polarized light microscopy, andoscillatory rheological measurements. Previous researchwith CB [10] and blends of palm stearin in sesame oil [11]showed that parameters obtained through oscillatoryrheometry (i.e., the phase shift angle, d) closely followTAG crystallization, particularly at the stages of the onsetof nucleation and crystal growth.

2 Materials and methods

2.1 Preparation of CB and separation of polarlipids

One block (1000g) of unrefined CB was obtained from alocal distributor (PEALPAN, San Luis Potos�, M�xico).The block was melted (80 7C), vacuum-filtered throughWhatman paper No. 5, homogenized by constant stirringat 80 7C for 20min and stored at 4 7C under nitrogen inthe dark in amber flasks (250mL). The separation ofpolar lipids from unrefined CB was achieved by solid-phase extraction following the methodology of Pan et al.[12] with slight modifications. Thus, 10-g cartridges of adiol phase (50 mm particle diameter and 60 � averageporosity; Alltech Associates Inc., IL, USA) were condi-tioned first with methanol (40mL), followed by chloro-form (40mL), and then hexane (80mL). The CB samples(4 g), previously dissolved in 5mL chloroform, wereeluted with 80mL chloroform. This fraction was con-sidered as the TAG from CB free of polar lipids. The polarlipids retained by the diol phase were washed with anadditional 50mL of chloroform, then 20mL of acetone,followed by 140mL methanol/ammonium hydroxide(0.5mL of a 25% ammonium hydroxide solution/100mLmethanol). The acetone fraction contained the mono-and di-acylglycerols, and the methanolic fraction thephospholipids. All fractions were collected, evaporatedto dryness in a rotary evaporator using high-purity nitro-gen, and analyzed by HPLC for mono- and di-acylgly-cerols, TAG, and phospholipid concentration. The elutionwas carried out in a 12-unit vacuum chamber connectedto a bench top vacuum station (Alltech Associates Inc.,IL, USA).

2.2 HPLC analysis

The mono-, di-acylglycerols and TAG profile was deter-mined by HPLC utilizing a Waters 1525 equipment (WaterMillipore Co., Milford, MA, USA) with a light scatteringdetector (Eurosep Instruments, Cedex, France) and twoNova Pack C18 columns (3.9 6 150mm; Water MilliporeCo.) connected in series. Sample elution was carried outat room temperature using a solvent gradient of acetoni-trile/dichloromethane from 70:30 to 30:70. The gradientprogram was applied in 1 h, using two mixtures of aceto-nitrile/dichloromethane (90:10 and 10:90). The assign-ment of the TAG peaks and their concentrations weredetermined based on the retention times of TAG stand-ards (Supelco, Bellefonte, PA, USA; Sigma ChemicalCo., St. Louis, MO, USA) and the development of cali-bration curves, respectively. The total concentration ofmono- and di-acylglycerols was calculated adding the

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

Eur. J. Lipid Sci. Technol. 107 (2005) 641–655 Crystallization of cocoa butter evaluated by rheometry 643

areas under the respective peaks in the chromatogram. Inall cases, at least two independent determinations weredone.

Phospholipid determination was done following themethod of Singleton and Stikeleather [13] using a diolcolumn (5 mm particle size, 4.6 6 250mm; YMC Co.Ltd., Japan). Sample elution was carried out at roomtemperature with a solvent flow of 0.5mL/min, using a10-min linear gradient developed from 100:0 to 0:100 ofsolvent A/solvent B. Solvent A consisted of a 4:3 (vol/vol)mixture of 2-propanol/hexane and solvent B of a 4:3:0.5(vol/vol/vol) mixture of 2-propanol/hexane/water. Thisgradient was followed by a 20-min isocratic elution withsolvent B. The sample was injected as 15% (wt/vol) so-lution in solvent A. Total concentration of phospholipidswas calculated adding the areas under the respectivepeaks in the chromatogram. The assignment of phos-pholipid peaks was established based on the retentiontimes of standards (Supelco, Bellefonte, PA, USA; SigmaChemical Co., St. Louis, MO, USA). As in TAG determi-nation, at least two independent determinations weredone.

2.3 Non-isothermal DSC analysis and experi-ment design

Thermograms were obtained in a DSC equipment (Mod-el 2920; TA Instruments, Inc., New Castle, DE, USA) cali-brated as previously indicated [14]. A sample (8mg) ofnative CB or CBwithout polar lipids (CB-WPL) was sealedin an aluminum pan and held for 20min at 80 7C. Non-isothermal crystallization was carried out from 80 7C to250 7C using a cooling rate of 10 7C/min. After 1min at250 7C, the melting thermogram was obtained by heatingat a rate of 5 7C/min. The difference in heat capacity be-tween the reference and sample cells was partly com-pensated by using in the reference cell a sealed pan withthe same weight of inert material (i.e., aluminum) as theCB or CB-WPL in the sample pan (8mg). With the DSCsoftware, the onset of crystallization (To) was calculatedfrom the corresponding non-isothermal crystallizationthermogram. The end of melting (Te) was determinedusing the first derivative of the heat capacity of the sampleand by comparison with the baseline. To was defined asthe temperature where the baseline intersected with thetangent of the upward curve of the first exotherm. Te wasdefined as the temperature where the first derivative ofthe heat capacity of the last endotherm returned to thebaseline.

From the corresponding To and Te interval, four crystal-lization temperatures were selected (TCr: 18.5 7C, 19.0 7C,19.5 7C, 20.0 7C). These four TCr were used for further

isothermal crystallization studies with CB and CB-WPLusing, in each case, a completely randomized experimentdesign with two replicates.

2.4 Isothermal DSC analysis, determination ofthe Avrami index, and the crystallization rateconstant

Isothermal crystallization thermograms at the TCr of18.5 7C, 19.0 7C, 19.5 7C, and 20.0 7C were obtained byDSC for CB or CB-WPL. Thus, after erasing the crystal-lization memory of the sample (80 7C for 20min), the sys-tem was cooled at 1 7C/min until attaining isothermalconditions at each particular TCr. After completion of thecrystallization exotherm (i.e., heat capacity returned tothe baseline) at each TCr, the system was left for an addi-tional 30min. After this time, the melting thermogramswere determined at a heating rate of 5 7C/min. As for thenon-isothermal DSC analysis, the difference in heat ca-pacity between the reference and sample cell was com-pensated by using, in the reference cell, a sealed pan withthe same weight of aluminum as the CB or CB-WPL in thesample pan (8mg). For each TCr, the heat of transition foreach of the crystallization exotherms (DHCr) was calcu-lated as the area under the corresponding exotherm. Thelimits to calculate DHCr were established using the firstderivative of the heat flow from the time at which the heatflow of the corresponding exotherm initially departed fromthe baseline, to the time at which it returned to the base-line. On the other hand, the heat of melting (DHM) and thetemperature at the peak (TM’) of the corresponding endo-therm were determined using the first derivative of theheat capacity (i.e., TM’ was the temperature where the firstderivative of the heat capacity crossed the baseline). TM’was considered the apparent melting temperature of thepolymorph crystallized from the melt of CB or CB-WPL ata particular TCr.

On the other hand, using as crystallization model theAvrami equation (Eq. 1), the DHCr was used to calculatethe fractional crystallization, F, as a function of the iso-thermal crystallization time, t, as described previously[14]. With the F and t values, both the Avrami index (n) andthe crystallization rate constant (z) were determined byfitting Eq. 1 through the nonlinear estimation procedureavailable in STATISTICA (v. 6.0, StatSoft, Inc., Tulsa, OK,USA).

1 2 F = exp(2ztn) (1)

According to Avrami’s theory, the n value is a function ofthe time dependence of nucleation (i.e., sporadic orinstantaneous) and the dimensionality of the crystalgrowth process (i.e., one, two, or three dimensions), while

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

644 J. F. Toro-Vazquez et al. Eur. J. Lipid Sci. Technol. 107 (2005) 641–655

z is a complex rate constant of crystallization that con-sideers both the nucleation and the crystal growth rateand depends on n. This holds true as long as conditionsassumed by the Avrami model prevail, in particular freecrystal growth (i.e., no impingement between growingcrystals) [15]. The fulfillment of this particular condition tocalculate n and z values at the TCr investigated was eval-uated by getting microphotographs as a function of thecrystallization time with a polarized light microscope (seebelow).

2.4 Rheometry measurements

A mechanical spectrometer with 50-mm-diameter-paral-lel-plates geometry with a gap of 1mm was used for allmeasurements. Temperature control was achieved by aPeltier system located in the base of the measurementgeometry. Melted CB or CB-WPL (80 7C) was applied onthe base of the plate, avoiding bubble formation, and thesuperior plate was positioned on the sample surfaceusing the auto-gap function available in the rheometersoftware. A temperature program was applied as follows:first 80 7C for 20min and then cooling at 1 7C/min untilachieving the corresponding TCr. During cooling and iso-thermal conditions at each TCr, the linear viscoelasticregion (LVR) of the system was established for the differ-ent crystallization stages of both CB and CB-WPL. Aconstant angular frequency of 1 rad/s was used in theestablishment of the LVR. With this information, oscilla-tory stress programs were developed with the rheometersoftware to apply a given strain to the system, alwayswithin the LVR for the different crystallization stages. Thestrain interval applied, considering all TCr investigated andthe different crystallization stages, varied from 0.0075%to up to 2.5% strain (0.003–1mrads). At each TCr, at leasttwo independent determinations of the phase shiftangle (d) and the storage modulus (G’) as a function ofcrystallization time were obtained for both CB and CB-WPL. The rheograms were obtained by plotting the meanvalues of d and G’ as a function of crystallization time.

2.5 Polarized light microscopy

Microphotographs of the crystallizing CB or CB-WPLwere obtained with a polarized light microscope (Olym-pus BX51; Olympus Optical Co., Ltd., Tokyo, Japan)equipped with a color video camera (KP-D50; HitachiDigital, Tokyo, Japan) and a platina (TP94; Linkam Scien-tific Instruments, Ltd., Surrey, UK) connected to a tem-perature control station (LTS 350; Linkam ScientificInstruments, Ltd.) and a liquid-nitrogen tank. To guaran-tee a uniform sample thickness during crystallization, twocover slips were glued on a glass microscope slide leav-

ing a 2.2-cm distance between them. The sample (80 7C)was placed within the gap of a preheated (80 7C) glassslide, and a glass cover slip was placed over, resting onthe glued cover slips. This provided a uniform samplethickness of 0.16mm (160 mm). A program was devel-oped with the equipment software (Lynksys32 Ver. 1.6.1;Linkam Scientific Instruments, Ltd.) to heat the sample(80 7C for 20min) and then cooling it at 1 7C/min untilattaining a particular TCr. Microphotographs were taken at30-s intervals, including both the cooling and the iso-thermal stages.

2.6 Statistical analysis

Statistical analysis for the different crystallization param-eters evaluated (i.e., DHM, TM’, n, and z) was done byorthogonal comparisons after ANOVA of the correspond-ing completely randomized experiment design usingSTATISTICA (v. 6.0; StatSoft, Inc., Tulsa, OK, USA).

3 Results and discussion

The distribution of major classes of lipids in native CBwas97.94 6 0.80% of TAG, 1.71 6 0.64% of mono- and di-acylglycerols, and 0.35 6 0.15% of phospholipids. Theseresults agree with others previously reported for unrefinedCB [5, 16, 17]. The CB-WPL obtained after solid-phaseextraction did not show the presence of mono-, di-acyl-glycerols, or phospholipids by HPLC. In the same way,phosphorous was not detected in CB-WPL by plasmaemission (results not shown). On the other hand, no sig-nificant differences were observed in TAG profiles be-tween CB and CB-WPL (Tab. 1). The TAG profile for CBand CB-WPL showed that the symmetrical SUS-type TAG(i.e., StOSt, POSt, and POP) added up to 91% and90.3%, respectively, and tri-saturated TAG (i.e., StStSt,PStSt, and PPSt) were 0.81% and 0.92%, respectively. Incontrast, di-unsaturated asymmetrical TAG (i.e., POO andStOO) were 4.3% in both CB and CB-WPL. Regarding thephospholipids profile, from the total concentration innative CB, 57.14 6 0.07% was phosphatidylethanola-mine, 26.53 6 0.04% was phosphatidylcholine, and16.33 6 0.06% lysophosphatidylcholine. In this context,it is important to point out that phosphatidylethanolamineand phosphatidylcholine were the major phospholipidsfound in seed crystals developed in CB at 26.5 7C understatic conditions [5]. Using the concentrations of StOSt,POSt, and POP, Chaiseri and Dimick [17, 18] reporteddifferences among CB of six different origins. Such differ-ences supported the classification of CB as rapid or slownucleating. In the same way, rapid- or slow-nucleating CBhad been associated with the concentration and polarity

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

Eur. J. Lipid Sci. Technol. 107 (2005) 641–655 Crystallization of cocoa butter evaluated by rheometry 645

Tab. 1. TAG profile for native CB and CB-WPL

TAG§ CB [%] CB-WPL [%]

PLiO 0.13 6 0.01* 0.14 6 0.01*PLiP 0.53 6 0.06 0.56 6 0.04OOO 0.09 6 0.01 0.11 6 0.01POO 1.40 6 0.09 1.42 6 0.01PLiSt 1.56 6 0.03 1.55 6 0.01POP 14.69 6 0.20 14.83 6 0.57StOO 2.88 6 0.21 2.88 6 0.00StLiSt 0.86 6 0.13 0.80 6 0.02POSt 46.41 6 0.45 45.64 6 0.04PPSt 0.39 6 0.03 0.45 6 0.01StOSt 29.87 6 0.24 29.86 6 0.78PStSt 0.28 6 0.02 0.32 6 0.04StOA 0.87 6 0.09 0.96 6 0.01StStSt 0.14 6 0.01 0.15 6 0.02Unknown 0.15 6 0.05 0.15 6 0.01

§ P, palmitic; Li, linoleic; O, oleic; St, stearic; A, arachi-donic.* Mean and standard deviation of at least two inde-pendent determinations.

of the phospholipid species present [6, 19]. Following theabove criteria, both CB and CB-WPL used in this investi-gation might be considered as rapid nucleating.

The non-isothermal crystallization and melting thermo-grams of CB and CB-WPL did not show any significantdifferences (results not shown). Similar results wereobtained for anhydrous milk fat, milk fat TAG, and milk fatTAG added with the original diacylglycerols [1]. From thenon-isothermal thermograms, To was 15.5 7C and Te was27.0 7C. As previously indicated, the TCr used in the iso-thermal studies (i.e., 18.5 7C, 19.0 7C, 19.5 7C and 20 7C)were selected from the interval between To and Te. Theisothermal crystallization thermograms for CB and CB-WPL at the corresponding TCr are shown in Fig. 1. At theTCr investigated, the presence of two exothermal peakswas evident in both CB and CB-WPL (Fig. 1). Theseobservations agreed with previous DSC results obtainedwith CB at similar TCr [10]. Giving the crystallization con-ditions investigated and the phase diagram for CB report-ed by van Malssen et al. [9], the first exotherm wasassigned to the a polymorph, while the second wasassigned to the b’ polymorph. The initial development ofb’ crystals was quite probably developed from a crystalsthrough a polymorphic transition [9]. Previous studies [10]indicated that once the initial b’ crystals are developed viaa polymorphic transition from a crystals, additionalb’ crystallization might occur directly from the meltthrough a facilitated nucleation mechanism. This mech-anism occurred giving the b’ template previously devel-

oped in the system. The crystallization thermograms(Fig. 1) showed that as the crystallization temperatureincreased, the magnitude of the a polymorph peakdecreased (Fig. 1) in both CB and CB-WPL. Conse-quently, as TCr increased, less b’ crystals were developedvia the polymorphic transition from a, and b’ crystal-lization took longer (Fig. 1) since its main crystallizationmechanism was direct nucleation from the CB melt (i.e.,TCr = 20 7C). This crystallization behavior agrees withresults previously reported [10] and with the CB phasediagram developed by van Malssen et al. [9].

On the other hand, at all TCr investigated, shorter timeswere needed to complete crystallization with CB-WPL(Fig. 1B) than with native CB (Fig. 1A). Bunjes et al. [20]showed that the presence of polar lipids, such as phos-pholipids, slows the time course of the a-to-b’ transitionof tripalmitin in comparison with the one observed in thepresence of non-ionic surfactants. These authors asso-ciated this effect as the result of an interaction betweenthe polar lipids and TAG during the cooling process at lowsupercooling, which results in the formation of an unusu-al, less ordered a polymorph than the normal a poly-morph [20]. As previously discussed, TAG in CB initiallycrystallize in the a polymorph, which transforms into b’through a polymorphic transition. After the b’ template isdeveloped through this mechanism, additional b’ crystal-lization occurs directly from the melt through a facilitatednucleation mechanism [10]. Our results showed that thiswhole process took longer in CB (Fig. 1A) than in CB-WPL(Fig. 1B). Then, our results are in line with the delayingeffect of polar lipids in the time course of polymorphictransitions [20].

The behavior of the heat of melting (DHM) and the tem-perature at the peak (TM’) of the corresponding endo-therm for CB and CB-WPL as a function of TCr is shown inFig. 2. The melting thermograms from where DHM and TM’were calculated for both CB and CB-WPL showed thepresence of a single and quite symmetrical exotherm(results not shown). In general, in CB and CB-WPL therewas an inverse relationship of DHM (Fig. 2A). This effectwas associated with the lower amount of solid phasedeveloped in the melt through TAG crystallization assupercooling decreased. No significant difference(p = 0.05) was observed between the DHM of CB and CB-WPL at any of the TCr investigated. On the other hand, theTM’ of the corresponding endotherm observed a signifi-cant direct linear relationship with TCr (p ,0.005) in bothCB and CB-WPL (Fig. 2B). However, the TM’ of CB andCB-WPL were statistically the same (p = 0.05) at any ofthe TCr studied. From the above we concluded that polarlipids have an effect on the kinetics of TAG crystallizationin CB (Fig. 1), but do not affect the thermodynamic prop-

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

646 J. F. Toro-Vazquez et al. Eur. J. Lipid Sci. Technol. 107 (2005) 641–655

Fig. 1. Isothermal crystallization at different crystallization temperatures for CB (A) and CB-WPL (B).

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

Eur. J. Lipid Sci. Technol. 107 (2005) 641–655 Crystallization of cocoa butter evaluated by rheometry 647

Fig. 2. Relationship between the isothermal crystallization temperature (TCr), the heat of melting(DHmelting) (A), and the melting temperature (TM’) (B) for CB and CB-WPL. The solid line shows thegeneralized least squared fitting of TCr on TM’ for CB and CB-WPL.

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

648 J. F. Toro-Vazquez et al. Eur. J. Lipid Sci. Technol. 107 (2005) 641–655

erties of the crystals developed (Fig. 2). To further evalu-ate this aspect, phase shift angle (d) rheograms wereobtained during the non-isothermal and isothermalstages of CB and CB-WPL crystallization. Figs. 3 and 4show the d rheograms for CB and CB-WPL crystallizationduring the cooling and the isothermal stages at differentTCr. To understand the use of d rheograms in evaluatingTAG crystallization, some concepts have to be clarifiedfirst. The ratio between G’ and G” establishes changes inthe viscoelastic properties of dispersed systems, like theone studied here. The d parameter is defined asd = tan21(G”/G’), evaluating in a quite sensitive way theviscoelastic changes of complex systems. Thus, in a pureviscous system (i.e., a vegetable oil or completely meltedCB) d = 907. Once TAG achieve the lamellar organizationsthat develop the nucleus, a drastic decrease in d followsthe phase change from liquid to solid. Eventually, whenthe system becomes fully crystallized, d becomes zero.Additionally, the heat released during the crystal growthprocess is shown in d rheograms as a peak [11, 14]. In thed rheograms (Figs. 3, 4), the association of the d peakswith the development of particular crystallization pro-cesses in CB was done based on X-ray studies by vanMalssen et al. [9] and previous crystallization studiesdone with CB [10]. Then, in Figs. 3 and 4, it is evident thatin CB and CB-WPL some solid phase formation alreadyoccurred even before attaining isothermal conditions.This was evaluated by the drastic decrease in d observedbefore attaining TCr (Figs. 3, 4). Using the same method-ology and cooling rate, similar results have been obtainedwith blends of palm stearin in sesame oil [11] and withnative cocoa butter [10]. Previous research involving themeasurement of the induction time of crystallization (Ti) invegetable oil blends by rheometry, scanning diffusivelight-scattering (SDLS) and DSC [11] showed that Ti

measured by SDLS (TiSDLS) is associated with the earlyformation of a solid phase. Therefore, a high correlationwas found between TiSDLS and Ti as measured by the de-crease in the d rheograms [11]. In contrast, Ti measuredby DSC (TiDSC) was associated more with the crystalgrowth process than with the nucleation process [11, 14].Thus, in the present research, the onset of crystallization(i.e., nucleation) in CB and CB-WPL at TCr between18.5 7C and 20 7C occurred under non-isothermal condi-tions (Figs. 3, 4). As a result, crystal growth of the a poly-morph began even before attaining TCr (first d peak;Fig. 3). Afterwards, under isothermal conditions, a crys-tals transformed into b’ through a polymorphic transition.These b0 crystals grew (second d peak; Fig. 3) andworked as a template that facilitated crystallization ofadditional b0 crystals directly from the CB melt (thirdd peak; Fig. 3). As with the crystallization exotherms(Fig. 1) at all TCr investigated, shorter times were neededto complete crystallization (i.e., d tended to zero, achiev-

ing a plateau) with CB-WPL than with native CB(Figs. 3, 4). Then, the whole crystallization process in-volving the development of a crystals, their polymorphictransformation into b’, and the nucleation of additionalb’ crystals from the melt occurred at a faster rate andconcluded in shorter time in CB-WPL than in CB(Figs. 3, 4). Apparently, the d rheograms show the delay-ing effect of phospholipids in the time course of poly-morphic transitions (i.e., a to b’) [20]. However, theseresults disagree with the ones of Lawler [7] who reported adecrease in the CB crystallization rate and an increase inthe induction time of crystallization when phospholipidswere removed from native CB of different origins. Thesestudies used absorbance measurements (500 nm) to fol-low crystallization, a technique of limited resolution tofollow TAG nucleation [3] and the crystallization process ingeneral. An indirect evaluation of the faster crystallizationrate obtained in CB-WPL, in comparison with the oneachieved in CB, might be observed in the G’ rheogramsobtained under isothermal conditions (Fig. 5). Assumingthat the increase in G’ as a function of crystallization timeunder isothermal conditions is associated with the crystalmass developed, a faster crystallization rate wasobtained in CB-WPL than in CB (Fig. 5). This effect wasmore evident as TCr increased (Fig. 5B). On the otherhand, it is important to point out that in crystallized TAGsystems, G’ is associated with the three-dimensionalorganization of the TAG crystal network [10, 14] (i.e.,fractal dimension). Within this context, for the same TCr,similar G’ values at the end of the crystallization process(i.e., G’ of pseudo-equilibrium) were obtained in both CBand CB-WPL (Fig. 5), despite the fact that G’ in CB or CB-WPL followed a different profile as a function of crystal-lization time at the same TCr (Fig. 5). These results supportour previous observation indicating that polar lipids havean effect on the kinetics of TAG crystallization in CB, butdo not affect the properties of the crystallized systemunder equilibrium conditions. However, we did not meas-ure the fractal dimension developed by the CB systemsinvestigated.

On the other hand, Figs. 3 and 4 also show the tempera-ture at the onset of crystallization (i.e., nucleation) duringthe non-isothermal stage. As previously mentioned, atthis temperature, TAG achieved the lamellar organizationsthat developed the crystal nucleus, as established by thedramatic decrease in d from its original 907 value. Thecorresponding mean temperature for the onset of crys-tallization during the non-isothermal stage, independentof TCr, was 59.8 6 3.7 7C in CB and 39.5 6 1.5 7C for CB-WPL. The difference in the onset of crystallization duringthe non-isothermal stage between CB-WPL and CBpointed out the nucleating role of phosphatidylcholineand phosphatidylethanolamine proposed by Dimick

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

Eur. J. Lipid Sci. Technol. 107 (2005) 641–655 Crystallization of cocoa butter evaluated by rheometry 649

Fig. 3. Phase shift angle rheograms for the crystallization of CB (A) and CB-WPL (B) at a crystal-lization temperature (TCr) of 19.0 7C. Data represent the means of two independent determinations.The dotted line shows the beginning of the isothermal conditions. The onset of crystallization undernon-isothermal conditions (i.e., before the dotted line) is also shown. The polymorphic transitionsinvolved during crystallization, according to the discussion in the text, are also indicated.

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

650 J. F. Toro-Vazquez et al. Eur. J. Lipid Sci. Technol. 107 (2005) 641–655

Fig. 4. Phase shift angle rheograms for the crystallization of CB and CB-WPL at crystallization tem-peratures (TCr) of 18.5 7C (A) and 20.0 7C (B). Data represent the means of two independent determi-nations. The onset of crystallization under non-isothermal conditions (i.e., before the dotted line) isalso shown.

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

Eur. J. Lipid Sci. Technol. 107 (2005) 641–655 Crystallization of cocoa butter evaluated by rheometry 651

Fig. 5. Storage modulus (G’) rheograms for the crystallization of CB and CB-WPL at crystallizationtemperatures of 18.5 7C (A) and 20.0 7C (B). Data represent the means of two independent determi-nations.

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

652 J. F. Toro-Vazquez et al. Eur. J. Lipid Sci. Technol. 107 (2005) 641–655

[6, 21]. Accordingly, the less polar phospholipids (i.e.,phosphatidylcholine and phosphatidylethanolamine)might develop an inverse hexagonal mesophase (HII), withtrace amounts of water acting as the core. This structurewould be the initial nucleus on which, initially, the high-melting TAG of CB (i.e., StStSt, PStSt, PPSt) and then thesymmetrical SUS-type TAG (i.e., StOSt, POSt, and POP)crystallize under supercooling conditions. Without thepresence of this type of phospholipids, no inverse hex-agonal mesophase would be developed in the CB melt,resulting in a requirement for a higher thermodynamicdrive (i.e., higher supercooling, lower temperature) toachieve TAG nucleation. When a non-isothermal crystal-lization DSC thermogram was obtained using the samecooling rate as in the d rheograms (1 7C/min), a significantbut small difference (p ,0.05) was obtained between theonset of crystallization during the non-isothermal stage ofCB-WPL (15.0 6 0.0 7C) and the one of CB(15.5 6 0.1 7C). Although these results agree with thenucleating role of phosphatidylcholine and phosphatidyl-ethanolamine in CB [6, 21], d rheograms were evidentlymore sensitive to detect this effect than DSC. As pre-viously pointed out, DSC exotherms obtained under iso-thermal conditions are associated more with crystalgrowth than with the nucleation process [11, 14]. Thesame observation must apply also for DSC measure-ments under non-isothermal conditions. However, sev-eral factors must be kept in mind when comparing crys-tallization data obtained from different equipments. Thus,differences in weight or volume of samples used, equip-ment design and its impact on the thermodynamics of the

system, and heat and mass transfer conditions existing ineach measurement device affect crystallization to a dif-ferent extent [11].

On the other hand, the Avrami model was used to evalu-ate the effect of polar lipids on the crystallization rate andthe dimensionality of the crystal growth process. In theAvrami model, the n value describes the dimensionality ofthe crystal growth mechanism (i.e., one, two, or threedimensions), while z is a complex rate constant of crys-tallization that considers both the nucleation and thecrystal growth rate and depends on n. This holds true aslong as conditions assumed by the Avrami model prevail,particularly free crystal growth (i.e., no impingement be-tween growing crystals) [15]. The DSC exothermsobtained under isothermal conditions (Fig. 1) were used inthis analysis, keeping in mind that these measurementsare associated more with crystal growth than with thenucleation process [11, 14]. The n and z values for each ofthe two exotherms developed during the isothermalcrystallization of CB and CB-WPL (Fig. 1) at the TCr

investigated are shown in Tab. 2. Microphotographs ofthe crystals developed at equivalent crystallization timesfor the first and second exotherm in CB and CB-WPL areshown in Figs. 6 and 7. In both CB and CB-WPL, crystalimpingement was evident during the development of thesecond exotherm (Figs. 6B, D and 7B, D). In contrast,crystals developed during the first exotherm grew freely,with limited space restrictions (Figs. 6A, C and 7A, C). Asa result, the application of the Avrami model to calculate n

Tab. 2. Crystallization rate constant (z) and Avrami index (n) for the first and second exotherm ofnative CB and CB-WPL crystallization.

TCr [7C]First exotherm (Æ polymorph) Second exotherm (�’ polymorph)

z 6 1024 (h2n) n z 6 10210 (h2n) n

CB-WPL 18.5 253.84 6 1.21a 1.88 6 0.11e 4.32 6 2.52 5.39 6 0.1819.0 683.65 6 6.53b 1.67 6 0.25e 0.11 6 3.43 5.16 6 0.1119.5 415.75 6 60.26c 2.90 6 0.25f 8.50 6 1.92 4.96 6 0.0420.0 46.08 6 0.83d 2.43 6 0.001g 0.10 6 0.12 4.69 6 0.50

TCr [7C]First exotherm (Æ polymorph) Second exotherm (�’ polymorph)

z 6 1024 (h2n) n z 6 10210 (h2n) n

CB 18.5 4.99 6 0.08a 2.68 6 0.05e 38.37 6 41.70 4.23 6 0.4919.0 16.68 6 2.17b 2.58 6 0.58e,f 471.49 6 567.02 3.47 6 0.1519.5 0.66 6 0.49c 2.99 6 0.23f 104.24 6 128.61 3.70 6 0.2820.0 0.81 6 1.12c 2.99 6 0.73e,f 59.92 6 83.05 3.81 6 0.73

a,b,c,d Within each system studied, a different letter indicates a significant difference (p ,0.05).e,f,g Within each system studied, a different letter indicates a significant difference (p ,0.05).

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

Eur. J. Lipid Sci. Technol. 107 (2005) 641–655 Crystallization of cocoa butter evaluated by rheometry 653

Fig. 6. Polarized light microphotographs of CB at a crystallization temperature of 19.0 7C after 5min (A) and 70min (B), andat a crystallization temperature of 20.0 7C after 5min (C) and 120min (D).

and z values was more reliable with crystallization datafrom the first crystallization exotherm than with data fromthe second exotherm. Thus, in both CB and CB-WPL, ahigher coefficient of variation was observed in the z va-lues obtained for the second exotherm at all TCr than inthe ones obtained from the first exotherm (Tab. 2). In thesame way, the n values from CB-WPL at all TCr investi-gated were outside the ones described by the Avramitheory (Tab. 2). Nevertheless, in CB, the n values for thesecond exotherm were within the interval described bythe Avrami theory, with a magnitude similar to valuespreviously reported for the second exotherm in CB crys-tallization at similar TCr [10]. From the above, the followingdiscussion will include just the n and z values from the firstexotherm. Thus, at all TCr investigated, a higher crystal-

lization rate was achieved in CB-WPL than in CB(p ,0.05). Then, for a given TCr, shorter times were nee-ded to complete crystallization of the a polymorph (i.e.,first exotherm) in CB-WPL (Fig. 1B) than in native CB(Fig. 1A). As discussed previously, Bunjes et al. [20] pro-posed that an interaction between polar lipids (i.e., phos-pholipids) and TAG in the melt might be occurring duringthe cooling process at low supercooling, resulting in theformation of a less ordered a polymorph than the normala polymorph [20]. The molecular arrangement of TAG todevelop stable nuclei might be more difficult under theseconditions. In fact, once the onset of crystallization undernon-isothermal conditions was achieved, the decrease ind was faster in CB-WPL than in CB (Figs. 3, 4), pointingout the slower development of a stable solid phase when

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

654 J. F. Toro-Vazquez et al. Eur. J. Lipid Sci. Technol. 107 (2005) 641–655

Fig. 7. Polarized light microphotographs of CB-WPL at a crystallization temperature of 19.0 7C after 2min (A) and20min (B), and at a crystallization temperature of 20.0 7C after 2min (C) and 30min (D).

polar lipids were present in the melt. The overall effectwould be a lower nucleation rate and a slower crystalgrowth process for a polymorph development in CB thanin CB-WPL. Consequently, at similar TCr, lower z valueswere obtained in CB than in CB-WPL (Tab. 2). Again,these results disagree with previous reports [7] that indi-cate a decrease in the crystallization rate when polarlipids (i.e., phospholipids) were removed from native CBof different origins. On the other hand, in both systemsinvestigated, a maximum in the behavior of the z valuewas observed at TCr = 19.0 7C (p ,0.05), which was fol-lowed by a concomitant decrease as TCr increased(Tab. 2). A similar behavior has been observed with thepre-exponential term g for CB crystallized at similar TCr

[10]. The maximum in the g parameter was followed by a

change in morphology and size of the TAG crystals [10]. Inthis investigation, the parameter gwas calculated with theweak-link regime equation for colloidal dispersions [10].As described by Narine and Marangoni [22], the dimen-sionality of the TAG crystal network might be describedfollowing the weak-link regime equation for colloidal dis-persions, where g is a constant independent of the vol-ume fraction of solid fat (i.e., TAG crystals), but dependenton the size of the primary particles and on the interactionsbetween them (i.e., polymorphic state) [22]. In the sameway, the statistical analysis of the Avrami index (n) (Tab. 2)determined in CB and CB-WPL indicates a change in thedimensionality for crystal growth when TCr change from19.0 7C (n = 2.0–2.5) to 19.5 7C (n = 3.0) (Tab. 2). Thus, ingeneral and independent of the presence of polar lipids,

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de

Eur. J. Lipid Sci. Technol. 107 (2005) 641–655 Crystallization of cocoa butter evaluated by rheometry 655

at TCr of 18.5 7C and 19.0 7C, n = 2.0 for CB-WPL andn = 2.5 for CB, indicating that at these TCr a crystallizationfollowed a one-dimension growth mechanism (i.e., cylin-der-like). In contrast, at a TCr of 19.5 7C, a bi-dimensional(i.e., disk-like) crystal growth mechanism was followed.The fractional values obtained in the determination of n inCB and CB-WPL might be associated with secondarycrystallization occurring during crystal growth. Theseresults support our previous observation indicating thatthe rheological parameter, g, is quite sensitive to changesin the shape (i.e., cylinder-like vs. disk-like crystal growthmechanisms) or size of the crystals and their spatial dis-tribution in the crystal network of CB [10]. However, themicrophotographs obtained at the TCr investigated didnot show any morphological differences in crystal shape,and no measurement of the fractal dimension was per-formed in this investigation.

Acknowledgments

The present research was supported by CONACYTthrough the grant 2002-C01-39897.

References

[1] A. J. Wright, R. W. Hartel, S. S. Narine, A. G. Marangoni: Theeffect of minor components on milk crystallization. J Am OilChem Soc. 2000, 77, 463–475.

[2] W. L. Siew, W. L. Ng: Effect of diglycerides on the crystal-lization of palm oleins. J Sci Food Agric. 1996, 71, 496–500.

[3] A. J. Wright, S. S. Narine, A. G. Marangoni: Comparison ofexperimental techniques used in lipid crystallization studies.J Am Oil Chem Soc. 2000, 77, 1239–1242.

[4] A. J. Wright, A. G. Marangoni: The effect of minor compo-nents on milk crystallization, microstructures and rheologicalproperties. In: Physical Properties of Lipids. Eds. A. G. Mar-angoni, S. S. Narine, Marcel Dekker, New York (USA) 2002, pp125–161.

[5] D. H. Arruda, P. S. Dimick: Phospholipid composition of lipidseed crystal isolates from Ivory Coast cocoa butter. J Am OilChem Soc. 1991, 68, 385–390.

[6] P. S. Dimick: Compositional effect on crystallization of cocoabutter. In: Physical Properties of Fats, Oils and Emulsifiers. Ed.N. Widlak, AOCS Press, Champaign, IL (USA) 1999, pp 140–163.

[7] P. J. Lawler: Ph. D. Thesis. The Pennsylvania State University,Pennsylvania (USA) 1997.

[8] R. L. Wille, E. S. Lutton: Polymorphism of cocoa butter. J AmOil Chem Soc. 1966, 43, 491–496.

[9] K. van Malssen, A. van Langevelde, R. Peschar, H. Schenk:Phase behavior and extended phase scheme of static cocoabutter investigated with real-time X-ray powder diffraction. JAm Oil Chem Soc. 1999, 76, 669–676.

[10] J. F. Toro-Vazquez, D. P�rez-Mart�nez, E. Dibildox-Alvarado,M. Char�-Alonso, J. Reyes-Hern�ndez: Rheometry ofcocoa butter crystallization under static and stirring condi-tions. J Am Oil Chem Soc. 2004, 81, 195–202.

[11] J. F. Toro-Vazquez, V. Herrera-Coronado, E. Dibildox-Alvar-ado, M. Char�-Alonso, C. A. G�mez-Aldapa: Induction timeof crystallization in vegetable oils. Comparative measure-ments by differential scanning calorimetry and diffusive lightscattering. J Food Sci. 2001, 67, 1057–1065.

[12] L. G. Pan, A. Campana, M. C. Tom�s, M. C. Az�n: A kineticstudy of phospholipid extraction by degumming process insunflower seed oil. J Am Oil Chem Soc. 2000, 77, 1273–1276.

[13] J. A. Singleton, L. F. Stikeleather: High performance liquidchromatography analysis of peanut phospholipids. II. Effectof postharvest stress on phospholipid composition. J AmOilChem Soc. 1995, 72, 485–488.

[14] J. F. Toro-Vazquez, E. Dibildox-Alvarado, M. A. Char�-Alonso, V. Herrera-Coronado, C. A. G�mez-Aldapa: TheAvrami index and the fractal dimension in vegetable oilcrystallization. J Am Oil Chem Soc. 2002, 79, 855–866.

[15] J. F. Toro-Vazquez, E. Dibildox-Alvarado, V. Herrera-Cor-onado, M. A. Char�-Alonso: Triacylglyceride crystallizationin vegetable oils: Application of models, measurements, andlimitations. In: Crystallization and Solidification Properties ofLipids. Eds. N. Widlak, R. Hartel, S. Narine, AOCS Press,Champaign, IL (USA) 2001, pp. 53–78.

[16] T. R. Davis, P. S. Dimick: Lipid composition of high-meltingseed crystals formed during cocoa butter solidification. JAm Oil Chem Soc. 1989, 66, 1494–1498.

[17] S. Chaiseri, P. S. Dimick: Dynamic crystallization of cocoabutter. I. Characterization of simple lipids in rapid- and slow-nucleating cocoa butters and their seed crystals. J Am OilChem Soc. 1995, 72, 1491–1496.

[18] S. Chaiseri, P. S. Dimick: Lipid and hardness characteristicsof cocoa butter from different geographic regions. J Am OilChem Soc. 1989, 66, 1771–1776.

[19] G. M. Savage, P. S. Dimick: Influence of phospholipids dur-ing crystallization of hard and soft cocoa butter. Proc PennMannuf Confect Assoc. 1994, 49, 54–59.

[20] H. Bunjes, M. H. J. Koch, K. Westesen: Influence of emulsi-fiers on the crystallization of solid lipid nanoparticles. JPharm Sci. 2003, 92, 1509–1520.

[21] P. S. Dimick: Influence de la composition sur la cristallisationdu beurre de cacao. La Cristallization, 3i�me Colluge, Paris(France) 1994, pp. 33–46.

[22] S. S. Narine, A. G. Marangoni: Fractal nature of fat crystalnetworks. Phys Rev E 1999, 59, 1908–1920.

[Received: March 24, 2005; accepted: June 13, 2005]

ª 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ejlst.de