The Effect of Polyglycerol Esters of Fatty Acids on the Crystallization of Palm Olein

: This research investigated the effect of polyglycerol ester of fatty acids (PGE) on the crystallization of palm olein (POL). Three PGEs were studied: two solid-state PGEs (PGE1105 and PGE1117) and one liquid-state PGE (PGE1155). The addition of 0.5-5% wt. PGEs influenced the crystallization kinetics of POL and this depended on the type and concentration of the emulsifiers. During cooling down with a cooling rate of 5 ℃ /min, the samples containing PGE1105 and PGE1117 started to crystallize at higher temperatures when compared with POL but the crystallization began at lower temperatures for the samples containing PGE1155. All samples with added PGEs exhibited lower solid fat content than that of POL after 12 h of crystallization time. The number of crystals decreased with an increase in the crystal size with PGE addition but there was no effect on polymorphism. Overall, the results suggested that PGE1105 and PGE1117 enhanced the early stages of POL crystallization possibly via the template effects but suppressed the later stages, whereas PGE1155 delayed the whole process of POL crystallization. The application of POL is often limited by its tendency to get cloudy at low temperatures during long-term storage. Based on the results, 1-5% wt. PGE1155 could be used to delay or prevent the crystallization of POL at low temperatures.

tion of fat systems is of great interest for improving the quality of high fat food products such as chocolate, margarines, spreads, confectionery, bakery and dairy products 5 . It is known that emulsifiers can modify the crystallization behavior of fats by either promoting or delaying the crystallization 2, 6 8 . The emulsifiers can interact with the nucleation, crystal growth, and/or polymorphic transitions of fats depending on the miscibility of the emulsifiers and the fats 9 . The influence of emulsifiers on fat crystallization could also be related to the chemical structure of triacylglycerols TAGs in the fat and the emulsifiers where the similarity between their structures leads to co-crystallization and the structural dissimilarity may delay the nucleation and possibly inhibit the crystal growth 1, 10,11 . In addition, emulsifiers can act as heteronuclei, which accelerated the crystallization by direct catalytic action as impurities or seeding effect. The rate of crystal growth can be altered as a consequence of adsorption of the emulsifiers on the fat crystal surface or inclusion in the fat crystals 12 .
Polyglycerol ester of fatty acids PGE are a broad class of food emulsifiers which are produced by polymerization of glycerol in the presence of an alkaline catalyst followed by esterification with fatty acids 13 . The advantageous properties of PGEs are derived from simple modification of their hydrophobicity and hydrophilicity by varying the degree of polymerization of glycerol and esterification with fatty acid moieties 14 . This gives access to a wide range of HLB values from 3 to 14, popularly used in many different food and nonfood applications 15 . PGE can be used as a crystallization modifier as they can influence nucleation and crystal growth and modify polymorphic transformation. Petruccelli and Añón 16 reported that a PGE rich in C 16 , C 18 and C 18:1 74 in total increased the number of sunflower wax crystals in sunflower oil but decreased the crystal size. Sakamoto et al. 17 revealed that polyglycerol behenic acid ester significantly affected the crystallization properties of PO. With the addition of 1 wt., the size of PO crystals decreased with an increase in the crystal number, indicating that polyglycerol behenic ester promoted nucleation and inhibited crystal growth of PO. In addition, the XRD patterns showed that, without the PGE, PO crystallized in the α form after rapid quenching to 10 and transformed to the β and β forms during heating at around 15 and 40 , respectively. In contrast, with the PGE, PO crystallized in β form at 10 and did not transform to β-form during heating. Hamada et al. 18 reported that a PGE rich in long chain fatty acids FA C 16 -C 22 , 60 in total and medium chain FA C 8 -C 14 , 40 in total decreased the crystallization temperature and suppressed the crystal growth of PO-based biodiesel fuels. A more recent study by Saitou et al. 19 showed that the FA composition of PGE is important in controlling the initial step of crystallization. The addition of 0.2 wt. PGE containing C 16 and C 18:1 effectively retarded the crystallization in the diacylglycerol-rich oil whereas the same amount of PGE containing only C 18:1 did not. The retardation of crystallization may be caused by molecular interactions between fats and additives, which occur during the cluster formation processes so that cluster formation is prohibited 11 . This research investigated the effect of different types of PGEs PGE1105, PGE1117 and PGE1155 on the crystallization properties crystallization kinetics, morphology and polymorphism of POL. The main objective of this work was to find PGE that could delay the crystallization of POL for industrial use.

Materials
POL IV 54.6 g I 2 /100 g was kindly supplied by P.P. Oil and Foods Co., Ltd. Thailand , and was composed mainly of C 18 . PGE1105 and PGE1117 were defined as solid-state surfactants given their paste-like appearance at room temperature, whereas PGE1155 was considered a liquid-state surfactant given that it is a viscous liquid at room temperature. Blends were prepared by adding 0.5-5 wt. PGEs to POL and stirred with a magnetic stirrer at 80 until a homogenous sample was obtained.

Crystallization and thermal pro les
The crystallization and melting profiles of the samples were analyzed with the Perkin-Elmer differential scanning calorimeter DSC Model D8000, Perkin-Elmer Co., Shelton, CT, USA following AOCS method Cj 1-94 20 . The instrument was calibrated with indium mp 156. 6 as a reference standard. A sample of 3-5 mg was placed in an aluminum pan 20 µL capacity and hermetically sealed with a sample press. An empty pan served as reference. The samples were heated from 25 to 80 and held for 10 min to remove any memory effect. Then they were cooled at 5 /min to 60 and held for 30 min followed by heating at 5 /min to 80 . The crystallization and melting profiles were generated during the cooling and heating, respectively. The profiles were analyzed by the software provided with the DSC.

Isothermal crystallization
The isothermal crystallization at 10 of the samples was studied by using a pulsed-nuclear magnetic resonance p-NMR spectrometer Minispec-mq20, Bruker, Karlsruhe, Germany to follow changes in solid fat content SFC as a function of time. The melted samples were poured into 10 mm O.D. p-NMR tubes to a height of 4 cm, heated at 80 for 10 min and then transferred to a cooling bath set at 8 . Once the sample temperature reached 10.5 , the tube was removed from the cooling bath wiped dry and rapidly put into the p-NMR sample port set at 10 0.5 . The cooling rate was 20 /min. The SFC was then continuously recorded for 12 h.
The SFC data were fitted to Avrami equation 21 : where, t is time, k is the Avrami rate constant, X is the fractional extent of crystallization at a given time t, and n is the Avrami exponent, which represents the dimensionality of growth and type of nucleation. X was taken as the SFC at any given time normalized by the equilibrium SFC SFC eq obtained at the end of the crystallization time 22 .
Curve fitting was performed after subtracting the induction time for onset of crystallization 23 . In addition, the SFC data were also fitted to the Gompertz equation 24 : where F t fractional extent of crystallization at a given time t, A is the maximum fractional extent of crystallization when t approaches infinity, µ is the maximum crystallization rate and τ is the crystallization induction time. Estimation of the Gompertz parameters was performed on the basis of experimental data by non-linear regression and the parameters were used to describe the crystallization kinetics of the samples 25 .

Crystal Microstructure
The microstructure of the samples was observed by using polarized light microscopy PLM Olympus BX51, Olympus Optical Co., Ltd., Tokyo, Japan equipped with a digital camera Olympus C-7070, Olympus Optical Co., Ltd., Tokyo, Japan . All samples were melted at 80 for 10 min and 20 µL of each molten sample was placed on the preheated slide and covered by the cover slip. Then, the samples were stored at 10 0.5 for 24 h in an incubator. The crystal morphology was imaged by a 10 lens. The sample temperature was maintained at 10 using Peltier temperature control stage for microscopes T95-PE Peltier System, Linkam Scientific Instruments Ltd, UK while the images of the samples were being taken. The Image J software was used to determine the crystal size and the crystallized area of the samples.

Polymorphism
The polymorphic structure of the samples was determined by an x-ray diffractometer XRD MiniFlexII, Rigaku, Japan . The samples were melted at 80 for 10 min and poured into rectangular plastic moulds 20 mm 25 mm 3 mm and then crystallized at 10 0.5 for 24 h in an incubator before XRD analysis. Scans were made in wide-angel x-ray scattering WAXS from 15 2θ to 30 2θ with a scan rate and a step width of 2 2θ /min and 0.02 2θ , respectively.

Statistical analysis
All experiments were performed in triplicate. The results were analyzed by analysis of variance with least significant difference ANOVA/LSD at 95 confidence interval.

Crystallization and melting thermograms
The most important aspect of the physical properties of fats and oils is related to their crystallization and melting behavior. If the crystallization conditions changed, crystal habit, crystal size and the number of crystals could be affected and these changes will eventually be reflected in the product performance 26 . The crystallization and melting thermograms of the bulk emulsifiers used in this work are shown in Fig. 1. PGE1105 showed a crystallization peak at 32 and a broad melting peak at the highest peak temperature of 38 due to its highest content of saturated FA. PGE1117 exhibited one broad crystallization peak at 24 and one broad melting peak at 36 . PGE1155 exhibited one distinct crystallization peak at 31.9 and one small peak at 1.4 . Its melting profile had multiple peaks with the main one locating at 20.6 . The crystallization and melting thermograms of all POL PGE samples are presented in Fig. 2. POL began to crystallize with a crystallization onset temperature T CO of 5 and showed one crystallization peak at 3.9 flanked by a broad shoulder on the lower-temperature side of the peak Fig. 2a and Table 1 . The result was in line with the DSC crystallization thermogram of POL IV 57.7 g I 2 /100 g reported by Siew and Faridah 27 . With the addition of 0.5-5 of PGE1105, the crystallization thermograms exhibited one main crystallization peak but with a broader characteristic than that of POL. T CO of all PO PGE1105 blends increased from that of POL p 0.05 and increased with the content of the emulsifier Table 1 , suggesting that PGE1105 induced the crystallization of the higher-melting TAGs in the fat system 28 due to the changes in molecular interactions between TAGs and between TAGs and the emulsifier 19 . The increase in T CO by the addition of the solid emulsifier was in agreement with previous studies. Basso et al. 29 reported an increase in T CO of PO with the addition of tripalmitin and monoacylglycerols and Domingues et al. 1 mentioned that T CO of PO increased with the addition of sorbitan tristearate and sucrose ester S370 . In addition,   Note: For crystallization, T C is the crystallization peak temperature and T CO is the crystallization onset temperature. For melting, T M1 is the melting temperature for the melting peak at lowest temperature and T M3 is the melting temperature for the melting peak at highest temperature. T MC is the melting completion temperature. ΔH C and ΔH M are the enthalpies of crystallization and melting, respectively. Different letters in the same column are significantly different (p < 0.05).
the sample with 5 PGE1105 exhibited a small exothermic peak at 21 arrow a which could represent the crystallization of the emulsifier 11,30 or a segregation of the highmelting TAGs of POL due to the interactions among TAGs and PGE1105 28  The melting thermograms of POL exhibited multiple endothermic peaks Fig. 2b relating to the melting of different groups of TAGs, indicating that there were several types of homogeneous polycrystalline of POL 12 . The two main melting peaks of POL were at 7 and 12 , representing the melting of low-and high-melting fractions. The addition of all PGEs at 0.5-5 altered the melting thermogram slightly from that of POL with an additional melting peak at 5 arrow b . The peak was more prominent with the addition of PGE1105 and PGE1117 compared to PGE1155. In general, the melting completion temperatures T MC were not significantly affected by the addition of the three surfactants p 0.05 Table 1 . The melting enthalpies ΔH M of POL PGE1105 blends increased as the content of the emulsifier in the blends increased p 0.05 , indicating the more ordered arrangement of POL TAGs. In contrast, ΔH M of POL PGE1117 blends and POL PGE1155 blends decreased as their content in the blends increased p 0.05 .

Isothermal crystallization
The isothermal crystallization curves of POL and POL PGE blends at 10 are given in Fig. 3. The shape of the crystallization curve, which is an SFC versus time curve, can give important clues for the crystallization mechanism of a particular fat system. POL started to crystallize at 0.5 h and showed two-step crystallization with an initial SFC rise to 2 within 1.5 h step 1 followed by a rapid SFC rise starting at 3 h before reaching an equilibrium SFC SFC eq of 38 after 12 h step 2 Table 2 . This was consistent with what was observed by Zhang et al. 26 during isothermal crystallization of POL at 10 .
The crystallization curves of POL PGEs were different from that of POL, an indication of different crystallization mechanisms. PGE1105 did not alter the two-step crystallization pathway of POL Fig. 3a . However, the samples began to crystallize before the temperature reached 10 during fast cooling with no induction time, suggesting that PGE1105 accelerated the early stages of POL crystalliza- tion. The effect was stronger with higher content of PGE1105. The observed behavior was similar to the effect of solid-state sucrose esters S170 and P170 on the crystallization of POL 31 and a PO-PS blend 32 . The same effect was also found in the crystallization of cocoa butter with the addition of solid-state sorbitan esters 33 . After the first step of crystallization, the SFC climbed rapidly to reach a plateau and SFC eq . At the beginning of the SFC increase, there was a crossover in SFC at a time point arrow c after which the sample with lower content of PGE1105 showed higher SFC and vice versa. At the end of crystallization time, the sample with higher PGE1105 content showed lower SFC eq p 0.05 Table 2 . This result was in contrast to the effect of S170 on the SFC eq of POL which increased with higher content of S170 31 . It is unclear why the concentration responses were different between the emulsifiers, but this suggests that they have different mechanisms in controlling POL crystallization, probably depending on the solubility of the emulsifiers in the oil 19 .
The crystallization curves of POL PGE1117 blends demonstrated a more complex behavior than POL PGE1105 blends Fig. 3b . The blends containing ≤ 3 of PGE1117 started to crystallize at approximately the same time as POL 0.5 h . However, the second step of crystalli-

The Effect of Polyglycerol Esters of Fatty Acids on the Crystallization of Palm Olein
zation of the blends with 0.5-2 occurred sooner than POL at 1.5 h for 0.5 and at 2 h for 1-2 PGE1117 . At the end of the experiments, these samples reached SFC eq of 36 . The blend with 3 PGE1117 showed a small effect on the early stage of POL crystallization but exhibited a stronger effect on the later stage with a significantly lower SFC eq of 33 p 0.05 . In contrast, the blends with 5 of PGE1117, although began the first step crystallization before the temperature reached 10 , drastically delayed the second step crystallization until 4 h with a significant drop in SFC eq 30 at the end of the experiment compared to POL p 0.05 .
The addition of PGE1155 at all concentrations did not affect the first step of POL crystallization Fig. 3c . However, the second step of POL crystallization was significantly delayed by all concentrations of PGE1155, except 0.5 . At the end of the experiments, all samples exhibited the SFC eq significantly lower than that of POL p 0.05 . All POL PGE blends showed lower SFC eq than that of POL, suggesting that all PGEs suppressed the later stages of POL crystallization with varying degree. The interaction of the fatty acids of the TAGs and the acyl groups of the emulsifiers was selective and led to different rates of retar-dation and acceleration of crystallization 34 .
The SFC crystallization curves were fitted to Avrami model via linear regression. The Avrami rate constant k and Avrami exponent n for the crystallization of POL at 10 were 0.04 and 2.47, respectively Table 3 . With the addition of PGE1105 and PGE1117, the k value significantly increased from that of POL p 0.05 , indicating the increase in the driving force for crystallization. The k values for POL PGE1105 blends decreased from 0.435 to 0.206 p 0.05 and the k values for POL PGE1117 decreased from 0.334 to 0.034 p 0.05 when the concentration of both emulsifiers increased from 0.5 to 5 . In contrast, the k values for all POL PGE1155 blends, except the blend with 0.5 PGE1155, were lower than that of POL p 0.05 and the k value decreased as the content of PGE1155 in the blend increased, suggesting that the crystallization of POL was suppressed by PGE1155. The n values of all POL PGE blends were between 1.8 and 2.4. The n values can be associated with the mechanisms of crystal growth and fractional numbers for n values indicate the contribution of more than one crystallization mechanism 23 .
The SFC crystallization curves fitted to the Gompertz equation revealed that the crystallization kinetics of POL were significantly influenced by the addition of PGEs Table 4 . The equation presents an induction time and crystallization rate in crystallization process that depend on the number of nuclei available for crystal growth 24 . Adding 0.5 and 1 PGE1105 to POL led to a reduction in the induction time with an increase in the crystallization rate p 0.05 . The addition of 0.5-2 PGE1117 also resulted in a decrease in the induction time and an increase in the crystallization rate from that of POL p 0.05 but 3-5 PGE1117 decreased the crystallization rate p 0.05 whereas the addition of all concentrations of PGE1155 cause the crystallization rate to decrease significantly p 0.05 . The results were in line with the data fitting with Avrami model presented above which showed that the crystallization rate generally increased with the addition of PGE1105 and PGE1117 but decreased when PGE1155 was added. Figure 4 presents the microstructure of POL crystallized at 10 for 24 h with and without PGE addition. POL exhibited a large number of small spherulites measuring 13.0 3.3 µm in diameter with some degree of aggregation and a crystallized area of 60.7 0.4 . The result confirmed that POL can readily form a number of small crystals at low temperature 10 , leading to the clouding problem during storage. The addition of PGEs did not affect the crystal shape of POL, with all blends showing spherulitic crystals. However, there were some changes in the crystal number and size compared to POL without additives. 0.5 PGE1105 did not have any significant effect on the microstructure of POL. As the content of PGE1105 increased further, the crystal size increased significantly from 9.8 3.3 µm for 0.5 PGE1105 to 46.7 6.7 µm for 5 PGE1105 p 0.05 with a decrease in the crystallized area 60.6 1.9 for 0.5 PGE1105 to 43.1 1.3 for 5 PGE1105, p 0.05 and a reduction in the number of crystal aggregates. The addition of PGE1117 resulted in a similar trend as that with PGE1105 but with larger crystal sizes measuring 14.4 4.8 and 56.5 11.6 µm in diameter for 0.5 and 5 PGE1117, respectively. The crystallized area also reduced from 55.2 0.8 for 0.5 PGE1117 to 39.1 0.3 for 5 PGE1117 p 0.05 . The results were in agreement with a previous report by Saitou et al. 2014 which showed a decrease in the crystal number of diacylglycerol-rich oil with an increase in the crystal size due to an addition of PGEs, suggesting that the rate of nucleation was reduced by the addition of PGEs 19 , but contradicted what was reported by Sakamoto et al. that the addition of polyglycerol behenic acid esters to PO at 20 led to an increase in crystal number and a decrease in crystal size 17 .

Crystal microstructure
The PLM results with PGE1105 and PGE1117 correlated well with the SFC results, which showed lower SFC eq with increasing PGE concentration Table 2 . The presence of PGE1155 resulted in a smaller increase in the crystal size from 13.0 3.5 µm for 0.5 PGE1155 to 28.0 5.1 µm for 5 PGE1155, p 0.05 and the formation of noticeably fewer crystal aggregates. The crystallized area 32-33 did not changed significantly with an increase in the content of PGE1155 p 0.05 . Figure 5 presents the x-ray diffraction patterns of POL with and without PGEs crystallized at 10 . The XRD  This study has shown that the addition of PGEs influenced the crystallization kinetics of POL but did not affect the polymorphism. Overall, the results suggested that the solid-state PGEs PGE1105 and PGE1117 enhanced the early stages of POL crystallization but suppressed the later stages, whereas the liquid-state PGE PGE1155 delayed the whole process of POL crystallization. When emulsifiers begin to crystallize before the fat during cooling down, the emulsifiers could promote the crystallization of the fat via template effects 11 which refer to the phenomenon where a higher melting additive with significant structural and chemical similarities to the lipid serves as templates seeding nuclei for heterogeneous crystallisation of lipids 36 . PGE1105 and PGE1117 have higher melting points than POL, hence could crystallize before POL upon cooling and increase nucleation of POL via template effects 11 . The FA composition of PGEs is important in controlling the initial step of crystallization 19 and the concentration 0.1 of the emulsifiers may be enough to have a strong impact on nucleation 37 . PGE1105 exhibited stronger effect during early stages of POL crystallization than PGE1117 probably due to higher degree of similarity between its fatty acid moieties C 16 and C 18:1 and those of POL 34,37 . During the later stages of crystallization, both emulsifiers could be absorbed at the kink sites of crystallizing POL and then prohibit the incorporation of the POL TAGs through large polyglycerol groups 19 . In contrast, PGE1155, an emulsifier which contained the highest content of C 18:1 in this group, suppressed POL crystallization all the way. It was liquid at 10 hence it did not crystallize before POL to give the template effects. The nucleation retardation may be related to molecular interactions between fat and additives, which occur during the cluster formation process so that the cluster formation is prohibited. The large polar polyglycerol groups of PGE1155 could prevent the POL TAGs from incorporating into clusters and nuclei due to steric hindrance caused by the difference in size of the polar groups 19 , resulting in the delay in POL crystallization. It was also possible that PGE1155 delayed POL crystallization by preventing the clusters of the fat from forming crystal nuclei through attractive molecular interactions between the emulsifier and the fat molecules de-clustering effect 11 .

Conclusion
PGE1105, PGE1117 and PGE1155 affected the crystallization kinetics of POL and the effect depended on the type and concentration of the emulsifiers. In general, the crystallization induction time decreased and the crystallization rate increased with the addition of PGE1105 and PGE1117. The crystal number and the crystallized area decreased with an increase in the crystal size as the concentration of PGE1105 and PGE1117 in the blends increased. The addition of PGE1117 resulted in an increase in the crystallization induction time and a decrease in the crystallization rate. Overall, PGE1105 and PGE1117 accelerated the early stage possibly via the template effects but suppressed the later stages of POL crystallization. In contrast, PGE1155 suppressed POL crystallization at all time points. The was no effect on the polymorphism of POL by any of the PGEs. Based on the results, PGE1155 could be used to delay or prevent the crystallization of POL at low temperatures.