On the fractional crystallization of palm olein: Solid solutions and eutectic solidification

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

On the fractional crystallization of palm olein: Solid solutionsand eutectic solidification

Gijs Calliauw a,b,*, Eveline Fredrick b,*, Véronique Gibon a, Wim De Greyt a, Johan Wouters c,Imogen Foubert b, Koen Dewettinck b

a Desmet Ballestra Group, Zaventem, Belgiumb Laboratory of Food Technology and Engineering, Department of Food Safety and Food Quality, Faculty of Bioscience-Engineering, Ghent University, Belgiumc Laboratoire de Chimie Biologique Structurale, Facultés Universitaires de Notre Dame de la Paix, Namur, Belgium

a r t i c l e i n f o

Article history:Received 18 September 2009Accepted 19 January 2010

Keywords:Dry fractionationFractional crystallizationPalm oleinEutectic solidificationPolymorphism

a b s t r a c t

In this contribution, the nature of particular crystal growth behavior observed in industrial dry fraction-ation of palm olein is further elucidated. Refined palm olein was prepared with three different concen-trations of tripalmitoyl-glycerol (0.6%, 0.8% and 1%) and crystallized under stirring at 13, 15 and 17 �Cto study the effect of internal seeding with tripalmitoyl-glycerol (PPP) on bulk crystallization properties.Notably at 15 �C, the increase in the solid fat content of the crystal suspension as function of time washeavily promoted by higher PPP-contents. The resulting crystals displayed a broad melting peak between28 and 40 �C in DSC-measurements, which suggested the presence of a solid solution in a metastableform, mainly consisting of PPP and dipalmitoyl-oleoylglycerol (POP). SAXS-measurements supported thisthesis by revealing long spacings with varying length, while short spacings obtained in WAXD consis-tently indicated the orthorhombic subcell of a b0-polymorph. The data suggest that the internal seedingeffect of PPP in palm olein, as applied in industrial crystallization, in fact relies on substantial miscibilitywith POP, rather than on a heterogeneous nucleation effect resulting in two separate crystalline phases.The miscibility of PPP and POP then would result in a facilitated alignment of POP-molecules in the direc-tion of the methyl end (or lamellar) plane and higher growth rates. Secondly, a peculiar change in crystalmorphology could be consistently linked with the occurrence of a sharp melting peak at 24 �C in DSC, aswell as with a substantial increase of the viscosity of the crystal suspension. This morphological shift con-curred with an increased incorporation of palmitoyl-oleoyl-stearoylglycerol in the lattice, possibly givingrise to eutectic crystallization in lamellar crystal structures instead of the commonly observed dendriticcrystals. The clarification of both phenomena provides a valuable, profound insight in the crystal growthstep in the processing of edible oils.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

One of the most important properties of an edible fat is its phys-ical behavior, since this determines its functionality in food, butalso affects the food processing conditions. Probably the mostimportant application of fat crystallization in the edible oil pro-cessing is the fractionation, or more fashionably ‘fractional crystal-lization’ (Kellens, Gibon, Hendrix, & De Greyt, 2007; Timms, 2005).This process commonly consists of a controlled fractional crystalli-

zation in bulk crystallizers followed by a physical separation of theliquid olein fraction from the crystalline fraction, nowadays mostlycarried out with membrane press filters. The motive for this pro-cess of edible oils is evident: for nature endows each fat with a par-ticular composition and distribution of the fatty acids on theglycerol, their use in specific applications is limited. Modificationtechniques or combination of these make it possible to go beyondthose limitations leading to a wide range of new products. Typi-cally, the obtained liquid fractions will exhibit a higher cold stabil-ity while the solid stearin fraction can often find use in foodapplications where a specific melting behavior is desired, for exam-ple in margarines, cocoa butter equivalents, vanaspati, etc. (Will-ner, 1994).

The backbone of any fractionation process is the fractionalcrystallization of the oil. Because natural fats are complex mixturesof numerous triacylglycerols (TAG), binary or ternary systems of

0963-9969/$ - see front matter � 2010 Elsevier Ltd. All rights reserved.doi:10.1016/j.foodres.2010.01.002

* Corresponding authors. Address: Laboratory of Food Technology and Engineer-ing, Department of Food Safety and Quality, Faculty of Bioscience-Engineering,Ghent University, Coupure Links 653, B-9000 Gent, Belgium. Tel.: +32 9 264 61 98;fax: +32 9 264 62 18 (G. Calliauw).

E-mail address: [email protected] (G. Calliauw).

Food Research International 43 (2010) 972–981

Contents lists available at ScienceDirect

Food Research International

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

Author's personal copy

pure components are often used to study and understand themain crystallization effects on a molecular level first. Results ofsuch studies have been extensively reviewed over the last dec-ades (Hagemann, 1988; Himawan, Starov, & Stapley, 2006; Sato,Ueno, & Yano, 1999) and have dramatically increased the insightin lipid crystallization. The extent of applicability of these resultsin real systems such as dry fractionation however is unsure, gi-ven that the specific processing conditions and complexity of thematrix involved are both quite distinct from model systems.Adding to the complexity of studying multicomponent systems,the liquid phase may not always be assumed an arbitrary ran-dom collection of molecules and intersolubility effects are likelyto take place, even in fairly simple blends of vegetable oils (Zhou& Hartel, 2006). For these reasons, there exists an added value instudying the more fundamental crystallization behavior of natu-ral oil systems in the actual industrial processing conditions.

In this contribution, two principal crystal growth phenomena ofthe fractional crystallization of refined palm olein for edible pur-poses are further examined:

� Crystal seeding practices are used to facilitate secondarynucleation, and/or improve crystal growth. There exists littleexperimental evidence of how the interactions on a molecularlevel can render the macroscopic properties at industrially-scale fractionation. In multi-stage palm oil fractionation, pro-cess improvements based on seeding (e.g. blending of palm oilinto its derived olein fraction) are often attributed to the pres-ence of more high melting tripalmitoyl-glycerol in the melt,facilitating bulk crystallization, somewhat reminiscent of theaddition of trisaturated triacylglycerols in lipid shorteningsand margarines (Humphrey & Narine, 2004). The nature ofthis presumed interaction acting in palm olein crystallizationis further examined and discussed in this work.� Possibly the most crucial technical issue in industrial oil frac-

tionation is maintaining a moderate viscosity of the crystalsuspension while the crystallization degree further increases.It is well documented that rheological properties of fat crys-tal networks are determined by many factors, and not neces-sarily proportionally related to its crystallization degree(Braipson-Danthine & Deroanne, 2004; Marangoni & Narine,2002). The only partial relevance of solid fat to the rheologyof the system has also been demonstrated in a fat crystalsuspension upon fractional crystallization of palm olein (Cal-liauw et al., 2007), in fact suggesting a specific fat crystalnetwork build-up, or possibly a gel-phase formation restrict-ing free-flow of the particles in the crystallizer. Additionally,it was shown that this effect is associated with a rather sub-tle but consistent shift in crystal composition. The currentpaper investigates the driving force for the formation andthe nature of these peculiar crystals.

2. Materials

Refined, bleached and deodorized (RBD) palm olein and a PPP-rich palm stearin fraction was obtained from an industrial refinery.This stearin fraction was blended in minor amounts with the origi-nal palm olein in order to increase the PPP-content without signif-icantly altering the levels of other TAG. In total, three blends wereprepared with a PPP-content of 0.6%, 0.8% and 1%. The TAG distri-butions of the palm stearin and the respective oleins are presentedas the average of two independent samples in Table 1.

Chemicals like standard POP and PPP (Sigma–Aldrich, Bornem,Belgium), both of more than 99% purity, reagents and solvents usedwere all analytical or HPLC-grade.

3. Experimental section

3.1. Crystallizer

Approximately 400 ml of palm olein was crystallized in a stain-less steel cylindrical vessel with a diameter of 8 cm and 10 cmheight, put in a temperature-controlled Julabo F12 waterbath(Seelbach, Germany), and stirred continuously by a 6 cm wide ro-tor, at 100 rpm, driven by an RW 20.n mixer (Ika Labortechnik,Staufen, Germany). The oil temperature was continuouslymonitored with a digital thermometer Testo 950, calibrated be-tween �20 �C and 60 �C by a standard mercury-in-glass thermom-eter. Isothermal crystallization was studied at 17, 15 and 13 �C. Allcrystallization experiments were repeated two (17 �C) or threetimes (13 �C and 15 �C).

3.2. Viscosity

The dynamic viscosity of the crystal suspension was measuredafter 2 min of stabilization at a shear rate of 10 s�1, with a BohlinCVO 50 digital rheometer with a KTB 50 waterbath (MalvernInstruments, Worcestershire, UK), using a C14 DIN 53019 plate-plate geometry.

3.3. Solid fat content (SFC)

The solid fat content of the crystal suspension was analyzedusing low field pulsed Nuclear Magnetic Resonance (p-NMR) bymeans of a Bruker Minispec mq 20 (Bruker, Germany) calibratedwith three supplied standards (0%, 31.4%, 73.8%). Samples were ta-ken directly from the crystallizer in thermo stabilized glass tubesand immediately measured using the direct method.

3.4. Polarized light microscopy (PLM)

A few drops of the crystal suspension were put on an acclima-tized carrier glass, and images were taken with a Canon 4500 Cool-pix digital camera, mounted on top of a Leitz Diaplan PolarisationMicroscope (Leica, Germany) with a PE 94 Linkam temperature-controlled carrier plate.

3.5. Differential scanning calorimetry (DSC)

The DSC-experiments were performed with a 2010 CE DSC (TAInstruments, New Castle, USA) with a Refrigerated Cooling System(TA Instruments, New Castle, USA). The DSC was calibrated withindium (TA Instruments, New Castle, USA), azobenzene (Sigma–Al-drich, Bornem, Belgium) and undecane (Acros Organics, Geel, Bel-gium) prior to the analyses. Nitrogen was used to purge thesystem.

Oil samples were taken from the crystallizing oil with an accli-matized glass Pasteur pipette and were sealed in alodined alumi-num pans in a thermostatic cabinet, pre-set at the correspondingexperimental cooling water temperature. Weighing of the filledpan was done after the DSC-experiment, in order to avoid possiblemelting through handling as much as possible. An empty pan wasused as a reference. Melting experiments were conducted at aspeed of 5 �C/min to 60 �C. The initial temperature was set at theisothermal temperature of the crystallization experiments. Themelting curves were integrated with a horizontal baseline. Thestart and end points of integration were determined as the pointsat which the slope of the heat flow curve started and respectivelystopped differing from the reference slope of the liquid sample inthat same run. Due to baseline instability, this procedure was notalways possible for the start points of the melting curves of crystals

G. Calliauw et al. / Food Research International 43 (2010) 972–981 973

Author's personal copy

obtained at 15 �C, therefore alternatively, these start points weredetermined as the intersection of the curve with the (extrapolated)liquid reference slope. The onset temperature was determined asthe intersection of the initial and the maximal absolute tangentto the curve, and is reported together with the integrated peakarea. Peak maximum is determined as the temperature at whichmaximal deviation between curve and baseline occurs.

3.6. Small angle X-ray (powder) scattering (SAXS)

SAXS-patterns were obtained using an X’pert Pro diffractometer(PANalytical, Almelo, The Netherlands). This diffractometer wasequipped with a sealed Cu-X-ray tube, 0.04 rad primary and sec-ondary Soller slits and a Ni-filter that produces the Cu Ka1 X-raybeam (k = 1.5419 Å). The X’celerator strip detector was used at0.52�2h, increasing with step size of 0.0170–4.98�2h. Samples ofthe crystallizing suspension were isolated and put on an absorbingtissue to reduce the liquid content and improve the resolution ofthe measurement. However, because of different amounts andsizes of crystals and variable liquid contents with every discretesampling, the exact peak intensities and widths cannot be usedas representative parameters. 3D-mesh plots were constructedby sampling and immediate measuring every 20–30 min, and lin-ear interpolation in between the obtained diffraction patterns.

3.7. Wide angle X-ray (powder) diffraction (WAXD)

WAXD-patterns were obtained using a PW1710 Philips diffrac-tometer (k Cu = 1.5419 Å, power 1200 W) (Philips, Eindhoven, TheNetherlands) equipped with a temperature control system (Pt-probes). Short-spacing runs ran from 15�2h to 27.7�2h, increasingwith a step size of 0.027�/s.

4. Results and discussion

4.1. NMR and DSC-experiments

The oleins with varying PPP-content were dynamically crystal-lized at 13 �C, 15 �C and 17 �C in the crystallizer.

At 15 �C, the crystallization rate increases exponentially withmore PPP present in the original melt (Fig. 1), while for all oleincompositions, ca. 20% SFC and comparable (high) viscosities(1300–1500 mPa s) were observed after 2 h at 13 �C. At 17 �Chardly any crystalline matter was formed in the considered timeframe. These observations demonstrate that a slightly increasedsupersaturation of PPP has a significant effect on the bulk crystal-lization kinetics, but also that this specific ‘internal seeding’ effect

of PPP, for industrial fractionation purposes, would be only practi-cally applicable in a rather narrow temperature range. The crystal-lization reaction is not finished after 2 h, and thermodynamiccalculations or classic crystallization models were not applied tothese limited data.

DSC was subsequently used to investigate the melting behaviorof the early crystalline phases observed with NMR. The results arediscussed only for 13 and 15 �C since at 17 �C insufficient crystal-line matter was formed to permit consistent analysis. Sampleswere taken after 30, 60, 90 and 120 min of isothermal crystalliza-tion for every olein, and immediately melted at 5 �C/min.

In Fig. 2, the thermograms overlay shows that little variation inthe total amount of crystallized fat existed between the differentoleins after 120 min crystallization at 13 �C, as the integratedendothermal peaks show very similar values (20–22 J/g). It is likelythat any possible seeding effect of PPP that would contribute to thequantity of crystallization, is subdued by the deep supersaturationof the bulk SUS-triglycerides such as POP and POS. However, thepositions and proportions of the various melting peaks do revealthe following: the melting area registered above 30 �C is substan-tially bigger with increasing initial PPP-content, which illustratesthat PPP will largely form independently from the bulk crystals(main central melting at 24 �C) at the imposed supercooling condi-tions. A similar shift, albeit a lot smaller, can be observed for themain central melting peak. This result suggests that the presenceof PPP does matter for the crystallization properties of the bulkpalm olein triglycerides as well, in the sense that PPP forms a solidsolution with a limited amount of POP crystals, and thus willslightly increase the melting point. So, at isothermal crystallizationof palm olein at 13 �C, PPP will largely contribute to formation ofcrystals that melt above 30 �C, but can display a certain degree ofintersolubility in solid form with the lower melting triglycerides.Hence, imagining a simple binary phase diagram of PPP and SUS-triglyceride, according to classic thermodynamics, the intersolu-bility between the solid PPP and SUS-triglycerides shall increasewith rising temperature.

Effectively at 15 �C, the effect of PPP in palm olein crystalliza-tion is reflected both in size and temperature range of the meltingpeaks (Fig. 3). Most clearly after 90 min of isothermal crystalliza-tion, the high melting peak, in this case ranging from 25 to 40 �C,is considerably bigger when the PPP-content of the olein is higher:0.44 J/g for 0.6%-PPP olein versus 6.2 J/g for 1%-PPP olein. Accord-ingly, the peak covers a slightly broader melting range than ob-served at 13 �C. This melting peak is likely the result of themelting of some more abundant PPP-rich mixed crystals. In theconsidered temperature range and for the amounts crystallized,about 60–70% of the palm olein crystals would be constituted by

Table 1Diacylglycerol (DAG) content and triacylglycerol (TAG) distribution of the palm stearin and oleins.

PPP stearin RBD palm olein 0.8% PPP olein 1% PPP olein

DAG content (%) 5.82 ± 0.07 6.27 ± 0.05 6.28 ± 0.03 6.28 ± 0.04

TAG distribution (w/w%)OLL 0.13 ± 0.03 0.51 ± 0.02 0.51 ± 0.02 0.51 ± 0.02PLL 0.49 ± 0.02 2.85 ± 0.02 2.83 ± 0.03 2.81 ± 0.02MLP 0.84 ± 0.04 1.04 ± 0.03 1.04 ± 0.04 1.02 ± 0.04OOL 2.87 ± 0.03 2.04 ± 0.02 2.05 ± 0.02 2.06 ± 0.01POL 6.14 ± 0.07 10.87 ± 0.06 10.83 ± 0.07 10.80 ± 0.04PLP 2.71 ± 0.03 9.81 ± 0.04 9.76 ± 0.05 9.71 ± 0.05OOO 1.07 ± 0.02 3.61 ± 0.03 3.59 ± 0.03 3.56 ± 0.04POO 7.06 ± 0.11 24.56 ± 0.09 24.40 ± 0.09 24.24 ± 0.11POP 37.58 ± 0.21 29.39 ± 0.15 29.46 ± 0.13 29.53 ± 0.16PPP 23.20 ± 0.13 0.60 ± 0.02 0.80 ± 0.02 1.01 ± 0.04SOO 0.54 ± 0.02 2.36 ± 0.03 2.34 ± 0.04 2.33 ± 0.03POS 6.50 ± 0.08 5.22 ± 0.04 5.23 ± 0.05 5.23 ± 0.04PPS 5.05 ± 0.09 0.09 ± 0.03 0.12 ± 0.01 0.15 ± 0.02

974 G. Calliauw et al. / Food Research International 43 (2010) 972–981

Author's personal copy

POP (Calliauw et al., 2007). If POP would only interact with the PPPseed crystals without or very limited miscibility of the molecularspecies, two separate melting peaks of two different crystallinephases would be expected: one well above 30 �C with an areaaccounting for ca. 1% PPP and one around 25 �C, representingPOP in b0-conformation (for the experimental conditions and con-centration of POP in natural palm olein crystals probably rule outformation of the c-polymorph of POP (Sato et al., 1989)), steadilygrowing in time. Instead in the experiment, one wide temperaturepeak is observed in the considered temperature range, suggestingthe presence of one solid solution.

4.2. SAXS and WAXD-measurements

Miscibility of PPP-molecules and POP-molecules in palm oleinfractionation conditions was further examined with SAXS andWAXD.

For comparison, the long spacing values obtained for pure POPand PPP, both in the 2L-packing (2L = double layer) induced byappropriate tempering, were measured. The respective peak max-ima were found at 2h = 2.13�, corresponding to a long spacing of41.6 Å (for k = 1.5419 Å) (POP); and at 2h = 2.22�, correspondingto a long spacing of 39.9 Å (PPP), i.e. two distinct values.

Fig. 1. Increase of SFC during crystallization at 15 �C for palm olein with different PPP-content.

Fig. 2. DSC-melting profiles of crystals formed after 120 min of isothermal crystallization of palm olein with different PPP-content at 13 �C. Onset temperature and peak areasare indicated; peak maximum temperature is shown below the actual peak apex; total integrated area is given between brackets.

G. Calliauw et al. / Food Research International 43 (2010) 972–981 975

Author's personal copy

Fig. 4 illustrates the evolution of the diffraction peak accountingfor the long spacing as function of time, for the 1% PPP olein crys-tallized at 15 �C at 100 rpm in the lab-scale crystallizer. The follow-ing tendency was observed: initially the strongest diffraction wasobserved for angles corresponding to long spacings slightly below40 Å (2h > 2.20�). Upon time, the maximal diffraction shifted toslightly smaller diffraction angles (2h < 2.16�) and so the corre-sponding long spacings expanded for about 1.5 Å. After 2 h of crys-tallization, only long spacings between 41.5 and 42 Å wereobserved. Albeit valid for all considered compositions, this shiftwas most notable in the 1% PPP olein.

Notably, the gradual evolution of the long spacings, in fact froma value close to the pure PPP b0-phase towards the measured value

for pure POP in b0, in combination with the observation of only onediffraction peak is in compliance with the general image deliveredby the DSC-measurements: more unsaturated entities get incorpo-rated into the growing crystallites, and would imply an outward‘gradient’ of increasing concentration of POP. This seems to confirmthe hypothesis of the existence of a PPP–POP 2L-solid solutionformed as a distinct physical entity in palm olein. The slight exten-sion of the observed long spacing would be explained by interfer-ing POP-molecules in the crystal lattice that inhibit the furtherformation of the original dense packing of PPP-molecules, howeverstill sufficiently compatible to be (to a certain extent) soluble inPPP-crystals. Specifically, the rigid character of the cis-double bondin the oleic chain of POP could cause substantial distortion of the

Fig. 3. DSC-melting profiles of crystals formed after 90 min of isothermal crystallization of palm olein with indicated PPP-content at 15 �C. Onset temperature and peak areasare indicated above the curve; peak maximum temperature is shown below the actual peak apex.

Fig. 4. Evolution of long spacings during isothermal crystallization of 1% PPP-palm olein at 15 �C, obtained by SAXS.

976 G. Calliauw et al. / Food Research International 43 (2010) 972–981

Author's personal copy

lattice. The broad melting range of these crystals as observed in thecrystals can then also be explained by concentration gradients ofPPP and POP. It also has to be mentioned that (saturated) diacylgly-cerols are likely to interfere in the early stage of crystallization too,and could influence the observed long spacings.

To verify the temperature dependency of the miscibility of PPPand POP in fractionation conditions, in a separate experiment,palm olein was very slowly crystallized at 24 �C for 72 h in orderto induce a fractional crystallization, allowing mainly trisaturatedcomponents and only limited amounts of POP to crystallize be-cause of the decreased supersaturation (Fig. 5). Again most obviousfor 1% PPP-palm olein, it was shown that the filtered crystals, evenat low resolutions and broad peak widths, exhibit two peaks orshoulders in the area of the long spacings of TAG in a 2L-arrange-ment, conform to the values of pure POP and pure PPP. So, as theused SAXS measurement is able to differentiate between the twocrystalline entities formed at lower supersaturation, the singlepeak pattern observed for the long spacings at higher supersatura-tion (Fig. 4) likely indicates a solid solution, in which the constitut-ing components can be hardly distinguished from one another.

WAXD-patterns of all crystals discussed so far revealed a con-sistent b0 character, as could be deduced from by the orthorhombicsubcells indicated by strong diffraction at angles corresponding to4.2 and 3.8 Å. Because of insufficiently resolved peaks, further de-tailed evaluation of these measurements (e.g. subtle peak shifts)was not possible.

In an additional experiment to examine how the PPP–POP solidsolutions are constituted in fractionation conditions, palm oleincrystals were isolated from the melt in the early stages of the crys-tallization under the said conditions (15 �C, 100 rpm) (SFC: 3–4%)and stored at 24 �C for 3 days. The crystals did not ‘revert’ intotwo separate peaks like shown in Fig. 5, but on the contrary a partof the crystalline mass developed a 3L-packing (Fig. 6). The diffrac-tion angles are entirely consistent with the independently deter-mined values of the pure POP 3L-packing (b-polymorph).

4.2.1. DiscussionFrom the sum of experiments discussed above, it was concluded

that PPP and POP would have sufficient structural resemblance toassociate into a relatively stable solid solution under the conditionsimposed during fractional crystallization of palm olein. Fundamen-tal molecular studies on this subject have indicated that indeedthis miscibility can be substantial, but would be quite dependenton processing conditions:

In binary mixtures, Gibon (1984) suggested miscibility of the b-forms of PPP and POP. These phases however were formed afterample tempering. A powder (synchrotron radiation) X-ray diffrac-tion study on the phase behavior of very pure PPP–POP mixturespublished by Minato et al. (1996) showed that upon cooling at15 �C/min, only a-phases of PPP and POP formed and largely sepa-rately (only up to 10% POP could dissolve in the PPP a-phase).Upon heating, at POP-levels above 50%, the transition from a to b

Fig. 5. SAXS-patterns of palm olein crystals, crystallized for 72 h at 24 �C.

Fig. 6. SAXS-patterns obtained for isolated initial palm olein crystals, after 72 h at 24 �C. Reference peaks of pure POP in a 3L-packing (b) are also shown for comparison.

G. Calliauw et al. / Food Research International 43 (2010) 972–981 977

Author's personal copy

occurred after a stable occurrence in the b0-polymorph, which wasattributed to the ‘molecular contacts’ between the fraction of sta-ble POP b0-polymorph and the crystal fraction of PPP.

Applying quite low cooling rates, Smith, Cain, and Talbot (2005)then demonstrated that in an acetone solution, PPP added to purePOP (in a w:w ratio of 1:50) relevantly alters the morphology of theresulting crystals. The authors provided little explanation on themolecular crystal formation involved, but concluded that in thestudied binary mixture, PPP had little influence on the crystalliza-tion kinetics of POP. However, the effect of PPP could be significantin real systems since diffusion coefficients, heat transfer and there-fore crystallization kinetics vary substantially between a melt andan organic solvent.

In their study on very slowly crystallizing crude palm oil, Chonget al. (2007) observed the formation of a stable solid solution in a2L(b0)-phase. Based on the TAG-compositions, it was concludedthat one chain out of six should be mono-unsaturated in this par-ticular phase. It can be remarked that a PPP/POP solution would fitin this scheme. Additionally the observation of a specific subcellarrangement pointed to a possible peculiar role of molecular com-pounds (as formed by POP/PPO (Minato, Ueno, Smith, Amemiya,and Sato (1997)). Indeed, due to increased solubility in PPP and ab0-tending property, even in very limited amounts (as in palmolein) PPO could play an important part as mediating factor to-wards the stabilization of a PPP/POP phase.

In summary, the collection of results, included those describedin this paper, leads to believe that in very pure mixtures of POPand PPP, the significance of interaction between the two speciesis limited, but that with the more complex compositions of edibleoils, and under appropriate conditions, a stable b0-phase solid solu-tion with PPP and POP as major constituents can occur.

Concerning the 3L(0 0 1)- and 3L(0 0 2)- diffraction peaksshown in Fig. 6, interestingly, it has been postulated by Uenoet al. (2008) that the presence of PPP in the center of granularcrystals in palm oil-based margarine fats could speed up a pre-sumably diffusion-regulated polymorphic transition of POP froma 2L(b0) to a 3L(b) conformation upon storage at room tempera-ture. Indeed, based on the phase diagram by Minato et al.(1996) that shows no indication of a truly stable b0-phase ofPPP in any mixing ratio with POP, it can be assumed that thePPP in the granular crystals can only be in a b-phase after thistempering, and therefore ‘trigger’ the re-crystallization of thePOP in its proximity first.

However, an alternative interpretation could be that the ob-served 3L(b) phase of POP is a distinct phase, formed indepen-dently from the original 2L(b0) solid solution:

(1) Whereas the prominent polymorph in palm olein and softPMF (containing resp. 28–30% and 43–47% POP) is the b0-polymorph (deMan, deMan, & Blackman, 1989), Braipson-Danthine and Gibon (2007) have attributed a limited b-tend-ing property to the hard PMF fraction (containing 60–70%POP) based on DSC and XRD-analysis. The effect of PPP inthis b-tending behavior cannot be ruled out, but it shouldbe remarked that PPP is present in relatively higher amountsin soft PMF than in hard PMF, and in the first fraction doesnot lead to b-formation. Taking into account that 100% POPis an unequivocal b-stable fat, the sum of results does pointtowards stronger b-stability with increasing POP, and notnecessarily with PPP. Hence, it is possible that the observedpolymorphic transition in the current research occurred as aresult of a fairly high concentration of POP.

(2) There is no separate peak or shoulder observed around2.23�2h in SAXS. This might indicate that the PPP presentin the crystals is packed in a b0-lattice that does not permitany conversion to a more stable form of PPP (b). So, if PPP

itself is not undergoing a polymorphic transition, the ques-tion remains how it instigates the transition from b0 to b ofthe other molecular species POP.

Therefore, it is our postulation that the supposed concentrationgradient of the solid solution of PPP and POP (and PPO) formedupon fractional crystallization could explain, next to a growth-enhancing effect, the only partial polymorphic transformation ofPOP in the isolated crystals. And also, albeit working as templatefor b0-crystallization of POP, the PPP in the core of crystals formedupon fractionation would not enhance polymorphic transitions ofPOP to 3L(b), but rather restrict them, as the solid solution in b0

is already sufficiently stable.

4.3. Eutectic lamellar growth in palm olein

Upon further crystal growth in the palm olein, the main crystalmorphological aspects were studied by polarized light microscopy.Based on crystal morphology, the studied crystallization process,for all studies compositions and temperatures could be dividedinto three stages (Fig. 7):

� the initiation stage or primary crystallization phase (after thenucleus formation), featuring small, spherulitic crystals formedabundantly in the melt. All observed crystalline structuresshowed a dendritic, spherulitic crystal morphology from start.

� aggregate formation through collision. This clustering indicatescrystal surface interactions, typically attributed to Van derWaals forces, which become more prominent although separatecrystals can still be observed; PLM did not allow distinguishingsignificantly different crystal morphologies or crystal centernumbers as a function of PPP-content in the initial melt.

� formation of a distinctly different crystal surface layer: a smoothcrystalline layer is formed completely around the original den-ser (or more intensively birefringent) and roughly-surfacedcrystals. This leads to even more pronounced agglomeration ofthe individual crystal centers. The development of this phasenotably coincides with earlier described sharply increased vis-cosities of the crystal slurry (Calliauw et al., 2007), hence thename ‘sticky layer’.

The changes in crystal morphology could be related to the evo-lution of the thermal properties of the developing crystals. Anoverlay is shown of the four melting curves of the actual crystalsformed in the 1% PPP olein during crystallization at 15 �C (Fig. 8).After 30 min, the melting profile reveals a low and a high meltingpeak. Both peaks persist in the melting curves after 60 and 90 min,but at this latter point the high melting peak expands towardshigher values (a little broadening to a higher temperature in thelow melting peak is also observed). After 120 min, a mid meltingpeak is observed in the range of 23–26 �C (peak maximum at24.1 �C), seemingly superposed on the two original melting peaks.This mid melting peak has an apex at a similar temperature as thebig central melting peak observed after 120 min at 13 �C (Fig. 2). Inboth cases, the occurrence of this very peak impeccably coincidedwith the observation of the secondary ‘viscous’ layer around theinitially formed crystals.

4.3.1. DiscussionIn a previous publication (Calliauw et al., 2007), it has been

shown that the increased viscosities in palm olein are associatedwith increased depletion of palmitoyl-oleoyl-stearoylglycerol fromthe melt. This particular observation could be linked with the sec-ondary crystal fraction with a relatively high melting point of 24–25 �C. It confirms the finding that secondary crystallization is notthe result of uncontrolled crystallization of lower melting TAG,

978 G. Calliauw et al. / Food Research International 43 (2010) 972–981

Author's personal copy

but is most likely the result of marginally formed crystal matterprincipally containing POP, POS and PLP in a very distinct ratio(Calliauw et al., 2007). Again, also the interference of positional iso-mers of these very mono-unsaturated TAG may not be excluded.

Combining the conclusions with the here presented results ofmicroscopy, DSC and XRD, it can be concluded that the change ofsurface appearance around the initial crystals coincides with theincreased incorporation of POS in a predominantly POP-containingcrystalline matrix (from ratio 1:6 to 1:4.5 (Calliauw et al., 2007)).

The thus formed secondary crystal layer is somewhat reminis-cent to what has been shown and described by Shi, Liang, and Har-

tel (2005) as a ‘gel-like’ material, and observed in model mono-unsaturated/tri-unsaturated TAG-mixtures. The possibility of amelt-mediated formation of viscous crystal-gels cannot be re-jected, but the experimental near-isothermal conditions are fairlydifferent from what has been described in literature (Higaki, Koy-ano, Hachiya, Sato, & Suzuki, 2004).

In fact, the clear separation between the two solid phases ratherbrings to mind a likeness with eutectic metallic alloy structures. In-deed, the possibility of eutectic formation in palm olein underthese particular conditions is real. It has been already establishedfor decades that in stabilized binary systems of palmitic and stearic

Fig. 7. (above) Consecutive stages of crystal morphologies of palm olein containing 0.8% PPP under isothermal conditions; field of view of pictures is 400 lm � 400 lm;(below) Magnifications of crystal structures observed in palm olein containing 1% PPP after (left) 30 min and (mid) 40 min of isothermal crystallization at 13 �C and (right)after 100 min at 15 �C. Field of view of pictures is 100 lm � 100 lm.

Fig. 8. DSC-melting profiles of crystals of palm olein with 1% PPP at 15 �C after indicated times, peak maxima are indicated.

G. Calliauw et al. / Food Research International 43 (2010) 972–981 979

Author's personal copy

acid, a melting depression (i.e. a eutectic point) exists at a ratio 25/75 (Bailey, 1950). For TAG-systems, the appearance of a similardepression in a ternary mixture of POP, POS and SOS (all in b-form),more precisely at low concentrations of SOS and 4:1 POP:POS ratiohave been reported (Wesdorp, 1990). Although also depending oncooling conditions and presence of impurities, a eutectic composi-tion is often associated with lamellar crystal structures, whereasthe non-eutectic crystallization would give rise to dendritic crys-tals (Kurz, 2006). In the studied case, the main crystal morphologyof the first formed, well-filterable crystals is of clear spheruliticnature. Possibly, the secondary layer however reflects a morelamellar crystal structure formation as a result of the uncontrolledsurface-activated crystallization of another (POS-rich) phase.Lamellar growth can also display high growth rates, which can ex-plain the virtually instant appearance of the secondary crystalstructure.

It cannot be excluded that at least one of the two phases haspreferred crystallographic growth directions which can lead todeviations from the formation of regular lamellar crystalline platesparallel with the growth direction. And possibly even more impor-tant in palm olein, the presence of other surface-active moleculessuch as diglycerides could substantially deviate the crystal growthdirection, inducing bridge formation and rapidly entrainingremaining melt. The result would be a speedily formed network,forming compartments in which liquid remain immobilized. Therelatively easy bridging would then also account for the ‘sticki-ness’, i.e. the viscous behavior of the total palm olein system.

Therefore, it is postulated that excessive supersaturation of themelt could induce a shift towards eutectic solidification: rapid sur-face-activated nucleation of POS, at the expense of its slower melt-controlled nucleation and steady crystal growth. Because of thehigh growth rates, it can be imagined that these structures wouldbe capable of entraining high amounts of liquid olein in betweenthe lamellae or rods, and this would explain the maze-appearanceof the structure as well as why the observed crystal size increasesheavily with only a moderate increase in solid fat content.

The shift in melt composition and the altered surface appear-ance of the crystals thus indicate that the crystallization kineticsand even the mechanisms are not necessarily constant during frac-tional crystallization, as a result of the specific ratios of the in-volved triacylglycerols (and possibly positional isomers of these,presence of impurities, etc.) and the applied cooling temperature.

5. Conclusion

With regard to kinetic properties of fractionally crystallizingpalm olein, a disproportional increase in the solid fat content ofpalm olein at 15 �C is shown as function of PPP-content. The molec-ular explanation for this industrially applied internal seeding effectis that PPP and POP can form a relatively high melting solid solu-tion in b0, possibly stabilized by presence of PPO, under the specificconditions acting during palm olein fractionation. The initial crys-tals primarily consist of PPP that serve as template for enhanced b0-crystallization of POP. The PPP/POP ratio in the solid solution willdecrease from the center to the surface of the seeds. Indicated bythe decreasing melting temperature in DSC and extending longspacings in XRD, the incorporated POP exhibits a destabilizing ef-fect on the PPP crystal lattice, and also during tempering, PPPwould hinder polymorphic transitions of POP. It is therefore the la-ter precipitated POP, not entirely integrated in the solid solution,which will likely develop into a 3L-b phase.

A distinct change in crystal morphology in subsequent crystalgrowth stages could be assigned to a shift in the proportional crys-tallization of the main mono-unsaturated TAG as a result of toostringent supersaturation conditions. The morphology change from

dendritic to ‘‘gel-layer” coincides with sharply increasing viscosi-ties that impede proper mass and heat exchange and phase separa-tion. These rapidly forming, supposedly lamellar crystal structuresdevelop via surface-activated nucleation around the originalspherulites. The particular crystalline phase would consist ofmainly mono-unsaturated triacylglycerols, in a eutectic ratio. Thephase is maintaining a b0-subcell arrangement yet is considerablylower melting than the POP–PPP solid solution.

Acknowledgements

The authors wish to acknowledge Laurence Plees and Berna-dette Norberg for their countless efforts and practical assistance.Also, Professor Kyotaka Sato is sincerely acknowledged for the con-tinued interest and encouragement to prepare this paper. FWOVlaanderen is also acknowledged for their financial support.

References

Bailey, A. E. (1950). Melting and solidification of fats. New York: IntersciencePublishers, Inc.. pp. 238.

Braipson-Danthine, S., & Deroanne, C. (2004). Influence of SFC, microstructure andpolymorphism on texture (hardness) of binary blends of fats involved in thepreparation of industrial shortenings. Food Research International, 37(10),941–948.

Braipson-Danthine, S., & Gibon, V. (2007). Comparative analysis of triacylglycerolcomposition, melting properties and polymorphic behavior of palm oil andfractions. European Journal of Lipid Science and Technology, 109(4), 359–372.

Calliauw, G., Gibon, V., De Greyt, W., Plees, L., Foubert, I., & Dewettinck, K. (2007).Phase composition during palm olein fractionation and its effect on soft PMFand superolein quality. Journal of the American Oil Chemists’ Society, 84(9),885–891.

Chong, C. L., Kamarudin, Z., Lesieur, P., Marangoni, A., Bourgaux, C., & Ollivon, M.(2007). Thermal and structural behaviour of crude palm oil: Crystallisation atvery slow cooling rate. European Journal of Lipid Science and Technology, 109(4),410–421.

deMan, L., deMan, J. M., & Blackman, B. (1989). Physical and textural evaluation ofsome shortenings and margarines. Journal of the American Oil Chemists’ Society,66(1), 128–132.

Gibon, V. (1984). Etude de polymorphisme et l’intersolubilité des triglycéridesconstitutifs des matières grasses végétales. PhD Thesis, Université de NotreDame de la Paix Namur, Belgium.

Hagemann, J. W. (1988). Thermal behaviour and polymorphism of acylglycerides. InN. Garti & K. Sato (Eds.), Crystallization and polymorphism of fats and fatty acids(pp. 9–95). New York: Marcel Dekker Inc.

Higaki, K., Koyano, T., Hachiya, I., Sato, K., & Suzuki, K. (2004). Rheological propertiesof b-fat gel made of binary mixtures of high-melting and low-melting fats. FoodResearch International, 37(8), 799–804.

Himawan, C., Starov, V. M., & Stapley, A. G. F. (2006). Thermodynamic and kineticaspects of fat crystallization. Advances in Colloid and Interface Science, 122(1–3),3–33.

Humphrey, K. L., & Narine, S. S. (2004). A comparison of lipid shorteningfunctionality as a function of molecular ensemble and shear: Crystallizationand melting. Food Research International, 37(1), 11–27.

Kellens, M., Gibon, V., Hendrix, M., & De Greyt, W. (2007). Palm oil fractionation.European Journal of Lipid Science & Technology, 109, 336–349.

Kurz, W. (2006). A brief history of the science of dendritic and eutectic growth until1980. Material Science Forum, 508, 313–324.

Marangoni, A. G., & Narine, S. S. (2002). Structure and mechanical properties of fatcrystal networks. In A. G. Marangoni & S. S. Narine (Eds.), Physical properties oflipids. New York: Marcel Dekker Inc.

Minato, A., Ueno, S., Smith, K., Amemiya, Y., & Sato, K. (1997). Thermodynamic andkinetic study on phase behavior of binary mixtures of POP and PPO formingmolecular compound systems. Journal of Physical Chemistry, 101, 3498–3505.

Minato, A., Ueno, S., Yano, J., Wang, Z. H., Seto, H., Amemiya, Y., et al. (1996).Synchrotron radiation X-ray diffraction study on phase behavior of PPP–POPbinary mixtures. Journal of the American Oil Chemists’ Society, 73, 1567–1572.

Sato, K., Arishima, T., Wang, Z. H., Ojima, K., Sagi, N., & Mori, H. (1989).Polymorphism of POP and SOS. I. Occurrence and polymorphictransformation. Journal of the American Oil Chemists’ Society, 66(5), 664–674.

Sato, K., Ueno, S., & Yano, J. (1999). Molecular interactions and kinetic properties offats. Progress in Lipid Research, 38, 91–116.

Shi, Y. P., Liang, B. M., & Hartel, R. W. (2005). Crystal morphology, microstructure,and textural properties of model lipid systems. Journal of the American OilChemists’ Society, 82, 399–408.

Smith, K. W., Cain, F. W., & Talbot, G. (2005). Crystallisation of 1,3-dipalmitoyl-2-oleoylglycerol and tripalmitoylglycerol and their mixtures from acetone.European Journal of Lipid Science & Technology, 107, 583–593.

Timms, R. E. (2005). Fractional crystallization – The fat modification process for the21st century. European Journal Lipid Science & Technology, 105, 48–57.

980 G. Calliauw et al. / Food Research International 43 (2010) 972–981

Author's personal copy

Ueno, S., Yamato, S., Tanaka, M., & Sato, K. (2008). Microbeam X-ray diffractionstudy for structure of granular crystal of margarine. Abstract of presentationgiven at 6th EuroFedLipid conference, Athens, Greece.

Wesdorp, L. (1990). Liquid–multiple solid phase equilibria in fats – Theory andexperiments. PhD Thesis, Technical University of Delft, The Netherlands.

Willner, T. (1994). High pressure dry fractionation of fats. Lecture paper: Oils & fatsgroup symposium: Fractional crystallization of fats, London, UK.

Zhou, Y., & Hartel, R. W. (2006). Phase behavior of model lipid systems: Solubility ofhigh-melting fats in low-melting fats. Journal of the American Oil Chemists’Society, 83(6), 505–511.

G. Calliauw et al. / Food Research International 43 (2010) 972–981 981