lable at ScienceDirect

LWT - Food Science and Technology 64 (2015) 483e489

Contents lists avai

LWT - Food Science and Technology

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

Fractal dimension in palm oil crystal networks during storage byimage analysis and rheological measurements

Zaliha Omar a, *, Norizzah Abd. Rashid b, Siti Hazirah Mohamad Fauzi a,Zaizuhana Shahrim a, Alejandro G. Marangoni c

a Product Development and Advisory Services Division, Malaysian Palm Oil Board, No. 6 Persiaran Institusi, Bandar Baru Bangi, 43000 Kajang, Selangor,Malaysiab Faculty of Applied Sciences, Universiti Teknologi MARA, 40450 Shah Alam, Selangor, Malaysiac Food, Health and Aging, Department of Food Science, University of Guelph, Ontario, ON NIG2WI, Canada

a r t i c l e i n f o

Article history:Received 30 October 2014Received in revised form22 April 2015Accepted 28 April 2015Available online 8 May 2015

Keywords:Crystallisation propertiesPolymorphic formFractal dimensionRecrystallisationCrystal network

* Corresponding author. Tel.: þ60 3 89254464; faxE-mail address: zaliha@mpob.gov.my (Z. Omar).

http://dx.doi.org/10.1016/j.lwt.2015.04.0590023-6438/© 2015 The Authors. Published by Elsevier

a b s t r a c t

The aims of this study are to determine the physicochemical properties and to quantify the fractaldimension of refined, bleached and deodourised palm oil (RBDPO) crystal network using image analysisand rheological methods during four weeks of storage at 15, 20 and 25 �C. There was no progressiveincrease in solid fat content (SFC) after one week of storage. X-ray diffraction indicated RBDPO stabilisedin the b0 polymorphic form without transformation to b polymorph upon storage. However, photomi-crographs of the crystal networks showed new small crystals formation in between the crystal clustersand on the surface of the existing crystals. The fractal dimensions of RBDPO crystal networks by imageanalysis (Db) and rheological measurements (Df) increased from day 1 to week 4 of storage. The resultsalso showed good agreement between (Db) and (Df) in quantifying the hardness, storage modulus (G0)and mass distribution of RBDPO crystal networks during storage.© 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND

license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Oils and fats play a crucial part in the human diet and haveimportant role in improving the palatability of foods. Because ofthat, palm oil (PO) is becoming important raw material and choicefor food manufacturers in producing plastic fats products such asmargarine and shortening (De Clercq, Danthine, Nguyen, Gibon, &Dewettinck, 2012; Fauzi, Norizzah, & Zaliha, 2013). The increasein demand of PO has encouraged researchers and food manufac-tures to carry outmany studies on nutritional, food formulation andits crystallisation behaviour. Palm oil crystallisation is known to bea complex process due to the slow crystallisation behaviour thatleads to post hardening phenomenon. The product tends to hardenslowlywith time and gives an unacceptable grainy or sandymouth-feel texture a few weeks after manufacturing (Duns, 1985; Ward,1998; Zaliha, Chong, Cheow, & Norizzah, 2005; Zaliha, Chong,Cheow, Norizzah, & Kellens, 2004). According to Timms (1985)the slow crystallisation behaviour of PO is due to the high levels

: þ60 3 89259658.

Ltd. This is an open access article u

of diacylglycerols present in PO which affect the crystallisationrates and polymorphic transformation. Later, Timms (1990) re-ported that post-crystallisation is caused by an increased inter-locking of the crystals in addition to an increase in the solid phase.He also reported that post-crystallisation process was due to thesmallness of the crystals. Smaller crystals lead to firmer fats withsmooth texture or mouthfeel, whereas larger crystals aggregatesproduce softer fats and may impart grainy texture or mouthfeel tothe final products (Pande & Akoh, 2013).

When PO fractions are used in products, a slow crystallisationcan take place after manufactured. Recrystallisation can occur toform a new solid phase as thermodynamic equilibrium isapproached (Smith, Finley, & Leveille, 1994). Crystal growth couldalso lead to the formation of solid bridges (sintering) between thenarrow gaps of the crystal networks upon storage (Johansson,Bergenst�ahl, & Lundgren, 1995) which could also contribute tothe recrystallisation phenomenon. However, the post-hardening inpalm oil/sunflower oil soft margarine formulation could be mini-mised by optimising the processing conditions (Mat Sahri & Omar,2014). To date, not much published information is available on thepost-hardening mechanism of fat crystal network in PO products

nder the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Fig. 1. Determination of the fractal dimension by Polarised Light Microscopy using the particle box counting method. A grayscale image (A) of the fat crystal network is thresholdedand inverted (B). The number of the crystal particles within the boxes of increasing sizes size placed over the thresholded image are counted (C). The fractal dimension is the slope ofthe logelog plot (D) of the number of particles (Np) as a function of the box size (R), cited from (Marangoni, 2005). Note: Df ¼ 1.69 (0.01), R2 ¼ 0.99.

Fig. 2. Effects of crystallisation temperatures on Solid Fat Content (g/100 g) usingPulsed Nuclear Magnetic Resonance (pNMR) during storage. Note: abccomparisons ofRBDPO at 15, 20 and 25 �C. Note: ¼ RBDPO 15 �C, ¼ RBDPO 20 �C, ¼ RBDPO25 �C.

Z. Omar et al. / LWT - Food Science and Technology 64 (2015) 483e489484

because no suitable technique is available in the past to investigatethis phenomenon except for the use of penetrometry.

The first model to describe fat crystal network was proposed byvan den Tempel (1961; 1979) and then extended by other re-searchers (Litwinenko, Rojas, Gerschenson, & Marangoni, 2002;Litwinenko, Singh, & Marangoni, 2004; Marangoni, 2005; Narine& Marangoni, 1999a; Vreeker, Hoekstra, den Boer, & Agterof,1992). Fat crystal network is arranged in a hierarchical fashion,with combinations of different levels being implicated in varyingdegrees depending on the nature of the rheological measurement(Heertje, van Eendenburg, Cornelissen, & Juriaanse, 1988). Aninterpretation of rheological data for aggregate fat networks in theframework of fractal theories was presented. The elastic modulus ofthe network (G0) varied with particle concentration of solid fat (F),or solid fat content, according to a power law, similar to models forthe elasticity of colloidal gels. It was reported that the value offractal dimension of tristearin in olive oil is about 1.7e1.8. Thisvalue is increased upon aging of the fat to D ¼ 2. Low fractal di-mensions are indicative of low-density, open structures (Vreekeret al., 1992).

Fractal geometry was first proposed to quantify natural objectswith a complex geometrical structure that defined quantificationby regular Euclidean geometry. Fractal systems are self-similar at alllevels of magnification and have fractional dimensions (D)(Mandelbrot, 1982). The fractal theory was adapted and used byresearchers in characterising the structure of fat crystal networks ofcocoa butter, milk fat, milk fat-canola oil blends, and shortening onthe basis of an original study (Heertje, 1993; Heertje et al., 1988).The scaling theory was applied by researchers to calculate thefractal dimensions of fat crystal network at the microstructurallevel from rheological and macroscopic fractal analysis. Polarisedlight microscopy was used as a method of imaging samples of fatcrystal networks at the microstructural level, and the images wereanalysed to yield fractal dimensions. The fractal dimensions werecompared to the values determined rheologically for similar fat

systems. A good agreement between the fractal dimensionsdetermined rheologically and by image analysis of the networkswas reported by Marangoni and Rousseau (1996), Narine andMarangoni (1999a, b).

In this study, fractal analyses via image analysis and rheologicalproperties were applied to the analysis of fat crystal structure in POproducts to shed light on the post-crystallisation phenomenonupon storage. The theory was used to quantify the fractality of thefat crystal networks of PO and its products during storage at varioustemperatures. This will help to give a better understanding on thecomplex crystallisation behaviour of PO and palm-based products.

Fig. 3. a: XRD pattern of b0 polymorphic form of palm oil (RBDPO) after 1 day ofstorage at 15 �C. Note: The short spacings of the b0 form are at 4.2 and 3.8 Å and that ofthe b form is at 4.6 Å (Souza et al., 1990). b: XRD pattern of b0 polymorphic form ofpalm oil (RBDPO) after 28 days of storage at 15 �C.

Fig. 4. Fractal dimension (Db) of two dimensional of RBDPO crystal networks duringstorage at 15, 20 and 25 �C. Note: ¼ 15 �C, ¼ 20 �C, ¼ 25 �C.

Fig. 5. Fractal dimension (Df) of three dimensional of RBDPO crystal networks duringstorage at 15, 20 and 25 �C. Note: ¼ 15 �C, ¼ 20 �C, ¼ 25 �C.

Z. Omar et al. / LWT - Food Science and Technology 64 (2015) 483e489 485

2. Materials and methods

2.1. Materials

Refined, bleached and deodourised palm oil (RBDPO) was ob-tained from Jomalina Sdn. Bhd. Klang, Selangor, Malaysia.

2.2. Solid fat content

Sample was melted at 70 �C for 30 min to erase all crystalmemory. Approximately 3.0 g of melted samples were pipetted intopulsed nuclear magnetic (pNMR) tubes (10.0 mm diameter, 1.0 mmthickness and 180mmheight). The samples weremelted for 30min

in the drying oven (Memmert, Germany) before being stored incontrolled temperature incubators (Binder, Germany) at 15, 20 and25 �C for 4 weeks. The measurements were carried out on a weeklybasis using pNMR spectrometer (Karlsuhe, Germany). After eachmeasurement, the samples were returned to the same storagetemperatures for the next measurement (the same tubes were usedfor the storage test). The measurement was carried out induplicates.

2.3. Polymorphic forms

The polymorphic forms of the samples was determined using anEnraf Nonius model FR 592 (Enraf Nonius, Delft, The Netherlands)and X-ray diffractometer (Rigaku, Japan). The instrument was fittedwith a fine focus copper X-ray tube. The samples were heated to70 �C for 10 min to destroy all crystal memory before being storedin controlled temperature incubators prior to measurements for 4weeks (Zaliha et al., 2005). On the measurement day, the X-raysample holder was pre-set at the desired crystallisation tempera-tures prior to measurements. The short spacings of the b0 form areat 4.2 and 3.8 Å and that of the b form is at 4.6 Å (Souza, deMan, &deMan, 1990). Levels of b0 and b crystals in the mixtures were

Plate 1a. Microstructures of RBDPO during storage at 20 �C [A] Day 1 [B] Day 3 [C] Week 1 [D] Week 2 [E] Week 3 [F] Week 4 SFC (g/100 g): [A] 24.8 [B] 31.5 [C] 32.0 [D] 31.8 [E] 31.9[F] 32.1.

Z. Omar et al. / LWT - Food Science and Technology 64 (2015) 483e489486

estimated by the relative intensities of the short spacings at 4.2 and4.6 Å.

2.4. Microstructure observation

Samples were heated at 70 �C for 30min, one droplet was placedon a pre-heated glass microscope slide, and a preheated cover slipwas gently dropped on top. Samples were then stored at 15 �C,20 �C and 25 �C for 4 weeks. The images were captured on aweeklybasis using an LTS 350 cold stage operated by a TPP 93 temperatureprogrammer (Linkam Scientific Instruments Ltd. Surrey, England)linked to an Olympus microscope. A black and white Sony XCe75CCD Video Camera (Sonny Corporation, Tokyo, Japan) recordedimages and Scion Image Software (Sion Corporation, Fredrick, MD,USA) was used to digitise the images. Using the Sion Software andan LGe3 PCI frame grabber, image quality was enhanced by takingan average of 16 frames.

2.5. Mass fractal dimension by image analysis (particle counting)

The fractal dimension was determined from in situ polarisedlight microscopy (PLM) images. The fractal dimension by box

countingmethodwas developed by Narine andMarangoni (1999b).The particleecounting procedure is based on counting the numberof crystal reflection within boxes of increasing sizes placed over aproperly threshold and inverted images. Threshold converts whitecrystals to black with black background. Various boxes of differentlength perimeter (R) are laid over the threshold image. Then thenumber of microstructural elements, N in each box is counted.Graph of log N vs log R is plotted, the slope of the regression linecorresponds to the fractal dimension (Db) of the crystal networkand c is a proportionality constant as in Fig. 1. The developedmethod is based on a polarised light microscopy (PLM) technique.In this case, crystals crystallised are restricted to a 2-dimensionalgrowth (crystals are between lines and plane) and the fractaldimension is between 1 and 2 (1 � D � 2).

N ¼ cRDb (1)

2.6. Viscoelasticity properties

The viscoelasticity properties of RBDPO crystal network wasdetermined using a Rheostress RS 600 Rheometer (Haake, Ger-many) equipped with a water circulator at the base of a 35 cm

Plate 1b. Microstructures of RBDPO during storage at 15 �C [A] Day 1 [B] Day 3 [C] Week 1 [D] Week 2 [E] Week 3 [F] Week 4 SFC (g/100 g): [A] 39.2 [B] 42.0 [C] 41.1 [D] 41.4 [E] 41.8[F] 41.8.

Z. Omar et al. / LWT - Food Science and Technology 64 (2015) 483e489 487

serrated parallel measuring plate. Sample was melted at 70 �C in anoven for 30 min before being pipetted into round molds. Themoulds were immediately transferred into controlled temperatureincubators (Memmert, Germany) and kept for 4 weeks duringmeasurements. An oscillatory stress sweep of 50e3000 Pa wasperformed to determine the storage modulus G0 within a linearviscoelastic region (LVR) at a frequency of 1 Hz. The fractaldimension (Df) of the PO crystal networkwas evaluated by applyingthe fractal model of:

G0 ¼ F1=d�Df (2)

whereF is the solid fat fraction, d is the Euclidean dimension (d¼ 3for a 3 e dimensional crystal network) and Df is the fractaldimension of the crystal network.

2.7. Statistical analysis

One-way ANOVA was performed using GraphPad Prism version4 for Windows (Prism GraphPad, 2003). Differences between

means were considered significant at p < 0.05. All data reported aremeans of at least two replicates.

3. Results and discussion

3.1. Solid fat content

The palm oil crystal networks showed rapid increase in solidlevels from day 1 to week 1 of storage at all storage temperatures.The crystal network of RBDPO at 15 and 20 �C showed higher in-crease in solid levels from 33 to 4 g/100 g and 27e32 g/100 g fromday 1 toweek 1, respectively as shown in Fig. 2. This could be due tothe slow crystallisation behaviour of PO to achieve maximum solidlevels. At 25 �C a slight increase in the solid levels, from 18.0 to18.7 g/100 g was observed. At this temperature (25 �C), some of thelowmelting triacylglycerols are still dissolved in the olein matrix ofRBDPO. This will reduce the compactness of the crystals present asin the microscopy observations shown in Plate 1c. From week 1 toweek 4 of storage, there was no significant changes in their solidlevels (p > 0.05). However, there was a significant difference

Plate 1c. Microstructures of RBDPO during storage at 25 �C [A] Day 1 [B] Day 3 [C] Week 1 [D] Week 2 [E] Week 3 [F] Week 4 SFC (g/100 g): [A] 12.4 [B] 13.0 [C] 13.3 [D] 13.0 [E] 13.2[F] 13.4.

Z. Omar et al. / LWT - Food Science and Technology 64 (2015) 483e489488

(p < 0.05) in solid levels at the different storage temperatures of 15,20 and 25 �C. No progressive increment of the solid fat content wasobserved up to 4 weeks of storage. This implies that the crystalnetwork of RBDPO has achieved its maximum solid levels within 1week.

Table 1Polymorphic forms using X-ray diffractometry at 15, 20 and 25 �C.

Storage time (days) Polymorphic form

15 (�C) 20 (�C) 25 (�C)

1 b þ b0 (b0 >>> b) b0 b0

3 b0 b0 b0

7 b0 b0 b0

14 b0 b0 b0

21 b0 b0 b0

28 b0 b0 b0

3.2. Polymorphic forms

Table 1 shows the polymorphic form by X-ray diffractometrycarried on the samples stored at 15, 20 and 25 �C. All the RBDPOcrystal networks stabillised in the b0 polymorphic form fromweek 1to week 4, except for a very weak b polymorph observed in theRBDPO crystal network at 15 �C at the earlier stage of the storagetime. As more b0 crystals are formed during the subsequent storageperiod, this trace amount of b crystals could be swamped by theamounts of b0 crystals as was also indicated in Fig. 3a and b. Thisobservation is consistent with the previous work carried out byZaliha et al. (2005), Zaliha, Siti Hazirah, Zaizuhana, and Norizzah(2014).

3.3. Microscopy observation and mass fractal dimension by imageanalysis (particle counting method)

Plate 1a shows the photomicrograph of microstructures ofRBDPO crystal networks stored at 20 �C. During the crystallisationprocess, the RBDPO crystallised and aggregated to form a combi-nation of larger clusters and small crystals. It was observed thatthose clusters formed at 20 �C were larger than the clusters formed

Z. Omar et al. / LWT - Food Science and Technology 64 (2015) 483e489 489

at 15 �C as shown in Plate 1b. After 1 week of storage it was notedthat more small crystals filled up the spaces in between the clustersuntil week 4 of storage. The SFC of the microstructure alsoincreased from 24.8 to 31.5 g/100 g from day 1 to day 3 and thenremained constant (±32.1 g/100 g) until day 28.

At 25 �C, the photomicrograph shows larger crystals aggregatedto form larger clusters as compared to themicrostructures at 15 and20 �C as shown in Plate 1c. These larger clusters lead to less clus-terecluster interactions and the crystal networks can be easilybroken down by an external force. Hence, the crystal network ofRBDPO at 25 �C is a weak crystal network and is less structured.After 1 week of storage, the clusters became more spherulitic andincrease in size (the white spot) as shown in Plate 1c. However, theSFC (±13.0 g/100 g) remained the same throughout the storage.

Fig. 4 shows the fractal dimensions (Db) of the crystal networksof RBDPO measured using the box counting method. It wasobserved that the Db of the crystal networks increased from 1.20 to1.57, 1.14 to 1.48 and 1.05 to 1.19 at 15, 20 and 25 �C, from day 1 toweek 4 of storage, respectively. Thus, the number of clustereclusterinteractions of the crystal networks continuously increased uponstoragewith the formation of more new crystals in between and onthe surfaces of existing crystals causing the whole system becamecompact and dense.

3.4. Fractal dimension by rheological measurements

The result showed that the fractal dimension (Df) of RBDPOcrystal network increased from day 1 toweek 4 of storage as shownin Fig. 4. The fractal dimension of the three dimensional crystalnetworks (Df) by rheological measurements increased from 2.19 to2.59, 2.09 to 2.50 and 2.04 to 2.21 at 15, 20 and 25 �C, respectively.This observation was in agreement with the fractal dimensions ofoils and fats crystal networks as reported by Vreeker et al. (1992). Itwas reported that upon aging, further crystallisation wouldcontinue while aggregation and network formationwould give riseto a denser, more compact and tightly packed structure with ahigher fractal dimension (Fig. 5).

4. Conclusion

Based on this study, it was shown that post-crystallisation hadoccurred upon storage without transformation of polymorphicform from b0 to b. However, more work on polymorphism need tobe carried out using a sensitive instrument to study the subtlechanges in polymorphic form within the same type of polymorph(e.g. b01, b02) to evaluate whether they will contribute to post-crystallisation too. The post-crystallisation could also be due tothe secondary nucleation that formed slowly on the surface of theprimary crystals upon storage which will change the viscoelasticityof the end-products. Fractal dimension by image analysis (Db) canbe a good indicator for viscoelasticity and mass distribution of thefat crystal networks for palm-based products. The fractal dimen-sion is able to detect small changes in hardness of the fat micro-structures with no significant changes in their SFCs. Hence, fractaldimension by image analysis (Db) can be a good indicator for thepost-hardening or post-crystallisation of oils and fats.

Acknowledgement

The authors would like to thank the Director General of MPOBand Faculty of Applied Sciences of Universiti Teknologi MARA ShahAlam for permission to publish this paper and to all the Physics and

Chemistry Laboratory staff of MPOB for their assistance. Greatestappreciation and thanks to Prof. Alejandro Marangoni, Departmentof Food Science, University of Guelph for the valuable guidance andarrangement for the experiments in his laboratory. Special thanksto Dr. Amanda, Silvana, Latifeh, Heidi, Dong Ming and Stphanie fortheir kind cooperation.

References

De Clercq, N., Danthine, S., Nguyen, M., Gibon, V., & Dewettinck, K. (2012). Enzy-matic interesterification of palm oil and fractions: monitoring the degree ofinteresterification using different methods. Journal of the American Oil Chemists'Society, 89, 219e229.

Duns, M. L. (1985). Palm oil in margarines and shortenings. Journal of the AmericanOil Chemists' Society, 62, 408e410.

Fauzi, S. H. M., Norizzah, A. R., & Zaliha, O. (2013). Effects of chemical inter-esterification on the physicochemical, microstructural and thermal propertiesof palm stearin, palm kernel oil and soybean oil blends. Food Chemistry, 137,8e17.

Heertje, I. (1993). Microstructural studies in fat research. Food Microstructure, 12,77e94.

Heertje, I., van Eendenburg, J., Cornelissen, J. M., & Juriaanse, A. C. (1988). The effectof processing on some microstructural characteristics of fat spreads. FoodMicrostructure, 7, 189e193.

Johansson, D., Bergenst�ahl, B., & Lundgren, E. (1995). Water-in-triglyceride oilemulsion. Effect of fat crystals on stability. Journal of the American Oil Chemists'Society, 72, 8939e8950.

Litwinenko, J. W., Rojas, A. M., Gerschenson, L. N., & Marangoni, A. G. (2002).Relationship between crystallisation behavior, microstructure, and mechanicalproperties in a palm oil-based shortening. Journal of the American Oil Chemists'Society, 79, 647e654.

Litwinenko, J. W., Singh, A. P., & Marangoni, A. G. (2004). Effects of glycerol andTween 60 on the crystallisation behavior, mechanical properties and micro-structure of a plastic fat. Crystal Growth & Design, 4(1), 161e168.

Mandelbrot, B. B. (1982). The fractal geometry of nature. New York: Freeman.Marangoni, A. G. (2005). The nature of fractility in fat crystal networks. In

A. G. Marangoni (Ed.), Fat crystal networks (pp. 322e439). New York: MarcelDekker.

Marangoni, A. G., & Rousseau, D. (1996). Is plastic fat rheology governed by thefractal geometry of the fat crystal networks? Journal of the American OilChemists' Society, 73, 991e994.

Narine, S. S., & Marangoni, A. G. (1999a). Mechanical and structural model of fractalnetworks of fat crystals at low deformations. Physical Review E, 60, 6991e7000.

Narine, S. S., & Marangoni, A. G. (1999b). Relating structure of fat crystal networks tomechanical properties: a review. Food Research International, 31, 227e248.

Pande, G., & Akoh, C. (2013). Enzymatic modification of lipids for trans-freemargarine. Lipid Technology, 25(2), 31e33.

Prism GraphPad. (2003). PRISM user's guide, version 4 software. San Diego, CA.Sahri, M. M., & Omar, Z. (2014). Minimising post-hardening in palm oil/sunflower oil

soft margarine formulation by optimizing processing conditions. Journal of OilPalm Research, 26(4), 340e354.

Smith, R. E. J. W., Finley, G. A., & Leveille. (1994). Overview of salatrim, a family oflow calories fats. Journal of Agricultural and Food Chemical, 42, 432e434.

Souza, D. V., deMan, J. M., & deMan, L. (1990). Short spacings and polymorphicforms of natural and commercial solid fats: a review. Journal of the American OilChemists' Society, 67, 835e843.

van den Tempel, M. (1961). Mechanical properties of plastic disperse systems atvery small deformations. Journal of Colloid and Interface Science, 16, 284e296.

van den Tempel, M. (1979). Rheology of concentrated suspensions. Journal of Colloidand Interface Science, 71, 18e20.

Timms, R. E. (1985). Physical properties of oils and mixtures of oils. Journal of theAmerican Oil Chemists' Society, 62, 241e249.

Timms, R. E. (1990). Crystallisation behavior of palm oil. In Symposium proceeding ofnew developments in palm oil (pp. 38e44).

Vreeker, R., Hoekstra, L. L., den Boer, D. C., & Agterof, W. G. M. (1992). The fractalnature of fat crystal networks. Colloids and Surfaces, 65, 185e189.

Ward, J. (1998). Crystallisation of fats. In N. Garti, & K. Sato (Eds.), Development inoils and fats (pp. 189e226). New York: Marcel Dekker.

Zaliha, O., Chong, C. L., Cheow, C. S., & Norizzah, A. R. (2005). Crystallization andrheological properties of hydrogenated palm oil and palm oil blends in relationto crystal networking on crystallization. European Journal of Lipid Science andTechnology, 107, 635e640.

Zaliha, O., Chong, C.,L., Cheow, C.,S., Norizzah, A. R., & Kellens, M. J. (2004). Crys-tallisation properties of palm oil by dry fractionation. Food Chemistry, 86,245e250.

Zaliha, O., Siti Hazirah, M. F., Zaizuhana, S., & Norizzah, A. R. (2014). Effect ofblending on physic-chemical properties of palm oil and palm oil products withsoyabean oil. Journal of Oil Palm Research, 26(4), 332e339.