Food Chemistry 141 (2013) 3393–3401

Contents lists available at SciVerse ScienceDirect

Food Chemistry

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

Creation of reduced fat foods: Influence of calcium-induced dropletaggregation on microstructure and rheology of mixed food dispersions

0308-8146/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.foodchem.2013.06.044

⇑ Corresponding author. Tel.: +1 413 545 1019.E-mail address: mcclements@foodsci.umass.edu (D.J. McClements).

Bi-cheng Wu a, Brian Degner b, David Julian McClements a,⇑a Department of Food Science, University of Massachusetts, Amherst, MA 01003, United Statesb ConAgra Foods, Six ConAgra Drive, Omaha, NE 68102, United States

a r t i c l e i n f o a b s t r a c t

Article history:Received 4 February 2013Received in revised form 26 April 2013Accepted 11 June 2013Available online 19 June 2013

Keywords:Reduced fatEmulsionAggregationRheologyDropletStarchDispersionsElectrostaticCalcium

The impact of calcium-induced fat droplet aggregation on the microstructure and physicochemical prop-erties of model mixed colloidal dispersions was investigated. These systems consisted of 2 wt% whey pro-tein-coated fat droplets and 4 wt% modified starch granules heated to induce starch swelling (pH 7).Optical and confocal microscopy showed that the fat droplets were dispersed within the interstitialregion between the swollen starch granules. The structural organisation of the fat droplets within theseinterstitial regions could be modulated by controlling the calcium concentration: (i) at a low calcium con-centration the droplets were evenly distributed; (ii) at an intermediate calcium concentration theyformed a layer around the starch granules; (iii) at a high calcium concentration they formed a networkof aggregated droplets. Paste-like materials were produced when the fat droplets formed a three-dimen-sional network in the interstitial region. The properties of fat droplet–starch granule suspensions can bemodulated by altering the electrostatic interactions to alter microstructure.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Over the past few decades there has been an increase in the per-centage of the population that is either overweight or obese (Finu-cane et al., 2011; Novak & Brownell, 2011). Physical inactivity andhigh calorie diets are major factors contributing to this increase inbody weight. Fat has the highest caloric density of the major nutri-ents in foods, and so there has been a considerable interest in thecreation of reduced-fat food products. However, the successfuldevelopment of these products remains challenging because fatplays multiple roles in determining their desirable physicochemi-cal and sensory attributes (Arancibia, Jublot, Costell, & Bayarri,2011; van Aken, Vingerhoeds, & de Wijk, 2011). The fat dropletsin many emulsion-based food products contribute to theirsmooth/creamy/rich texture, milky/creamy appearance, desirableflavour, and satiating effects (Frank, Appelqvist, Piyasiri, Wooster,& Delahunty, 2011; McClements, 2005; van Aken et al., 2011). Itis therefore important to identify commercially viable strategiesthat are capable of reducing the overall fat content of food productswhile maintaining their desirable sensory attributes.

One of the most popular fat replacement strategies currentlyused is to incorporate thickening agents (such as food

hydrocolloids) into emulsion-based products so as to compensatefor the reduction in viscosity that occurs when fat droplets are re-moved (Arancibia et al., 2011; Flett, Duizer, & Goff, 2010; Tarrega &Costell, 2006; Vingerhoeds, de Wijk, Zoet, Nixdorf, & van Aken,2011). An alternative approach is to induce flocculation of the fatdroplets, since this leads to an appreciable increase in shear viscos-ity or elastic modulus (Mao & McClements, 2012a; Mao & McCle-ments, 2012b; Simo, Mao, Tokle, Decker, & McClements, 2012).Consequently, highly viscous or gel-like textures can be createdat lower fat contents in flocculated emulsions. Droplet flocculationcan be induced in a variety of ways depending on the system, e.g.,changing pH or ionic strength to reduce electrostatic repulsion;heating to increase hydrophobic attraction; adding biopolymersto induce depletion or bridging interactions; mixing positive andnegative droplets to induce heteroaggregation (Mao & McCle-ments, 2012b; McClements, 2005). Most previous studies, on theinfluence of droplet flocculation on emulsion properties, have beencarried out using relatively simple model systems, consisting of fatdroplets dispersed within an aqueous medium. In practise, manycommercial food products consist of mixtures of different kindsof particles within an aqueous medium. For example, many com-mon food emulsions consist of mixtures of fat droplets and starchgranules, e.g., dressings, sauces, soups, and desserts (Chung, Deg-ner, & McClements, 2013). There is therefore a need to understandthe influence of particle organisation and interactions on the

3394 B.-c. Wu et al. / Food Chemistry 141 (2013) 3393–3401

microstructure and physicochemical properties of these morecomplex mixed particle systems.

Recently, we showed that mixtures of starch granules and fatdroplets could be treated as bimodal particulate suspensions,whose properties depend on the size and concentration of the dif-ferent kinds of particles present (Chung, Degner, & McClements,2012a; Chung, Degner, & McClements, 2012b; Chung et al.,2013). We also showed that the rheological properties of thesestarch granule – fat droplet mixtures could be modulated by con-trolling the aggregation state of the fat droplets within the aqueousphase through the pH (Wu, Degner, & McClements, 2013). At pHvalues around the isoelectric point of the protein-coated fat drop-lets an extensive droplet aggregation occurred, which led to a largeincrease in viscosity. It was proposed that a controlled aggregationof the fat droplets in mixed systems could be used to formulate re-duced-fat products with textural characteristics similar to those ofhigh-fat products. However, pH-induced fat droplet aggregationcan only be utilised in products that have pH values close to theisoelectric point of protein-coated fat droplets. Consequently, it isuseful to identify alternative means of inducing fat droplet aggre-gation in mixed systems so that this approach can be applied toa wider range of food products.

The purpose of the current study was to investigate the poten-tial of using calcium-induced droplet aggregation to control theproperties of model food dispersions containing mixtures of pro-tein-coated fat droplets and starch granules. In particular, weaimed to develop a better mechanistic understanding of the influ-ence of calcium content on the microstructure and physicochemi-cal properties of these systems. This information may be useful inthe design of reduced fat products with desirable qualityattributes.

2. Materials and methods

Throughout the manuscript we have used the units ‘‘wt%’’ to re-fer to grams of component per 100 g of sample.

2.1. Materials

Commercial whey protein isolate (WPI) powder was providedby Davisco Foods International (Le Sueur, MN, USA). The WPIwas reported to contain 97.9 wt% protein, 0.2 wt% fat, and1.9 wt% ash. Modified starch (hydroxypropyl distarch phosphate)made from waxy corn starch was provided by ConAgra Foods(Omaha, NE, USA). The modified starch was reported to contain0.6% ash. Calcium chloride dihydrate (CaCl2�2H2O), hydrochloricacid (HCl), sodium azide and technical grade Nile red dye (CAS#7385-67-3) were purchased from Sigma–Aldrich (St. Louis, MO,USA). Sodium azide was used to prevent microbial growth (but isnot suitable for food use). Double distilled water was used to pre-pare all solutions.

2.2. Preparation of model dispersions

Aqueous WPI solution (0.033 wt% WPI) was prepared by dis-persing a weighed amount of WPI powder into pH-adjusted aque-ous solution (0.02 wt% sodium azide, pH 7.0). The WPI solution wasstirred for at least one hour at room temperature to ensure hydra-tion. A coarse oil-in-water emulsion (25 wt% oil) was prepared bymixing appropriate amounts of WPI solution and canola oil (1:10protein-to-oil) using a high-speed blender (Tissue Tearor Model985370-395, BiosPec Products Inc., Bartlesville, OK, USA) at15,000 rpm for 2 min. The resulting coarse emulsion was thenpassed through a high-pressure homogenizer (Microfluidizer Mod-el 110 L, Microfluidics, Newton, MA, USA) three times at a chamber

pressure of 10,000 psi to further reduce the size of the fat droplets.After homogenisation, the stock emulsion had a pH of �7.2, whichwas then adjusted to pH 7.0 using a HCl solution.

A series of 2 wt% oil-in-water emulsions with different CaCl2

concentrations (0, 0.5, 1, 1.5, 2 and 3 mM) were prepared by mix-ing different ratios of stock emulsion, stock CaCl2 solution (25 mM,pH 7.0), and aqueous solution (pH 7.0). The emulsions were stirredcontinuously (for at least one hour) prior to further treatment toensure they were homogeneous. Starch granule dispersions(4 wt%) with different calcium concentrations (0, 0.5, 1.0, 1.5, 2.0,3.0 mM) were prepared by mixing different ratios of modifiedstarch powder, stock CaCl2 solution, and aqueous solution, andthen stirring until further use.

Mixed emulsion–starch dispersions (model sauces) with differ-ent calcium concentrations were prepared that contained 2 wt% fatand 4 wt% starch. Different ratios of stock emulsion, stock CaCl2

solution, and aqueous solution were mixed together followingthe same procedure as for preparing the pure emulsion samples,and then modified starch was added.

Samples (pure emulsions, pure starch suspensions or mixedemulsion–starch dispersions) (120 g) were placed in a 250 mlbeaker and heated to 90 �C and then held for 5 min in a waterbath (MGW Lauda KS6, LAUDA, Lauda-Königshofen, Germany).Unheated samples with similar compositions were collected forcomparison. The samples were manually stirred occasionally dur-ing heating to ensure a uniform heat distribution. The mass of thesamples was measured before and after this procedure, and wasadjusted back to the original moisture content (by adding aque-ous solution) prior to analysis to avoid losses due to waterevaporation.

2.3. Particle charge measurement

A particle micro-electrophoresis instrument (Zetasizer NanoZS,Malvern Instruments, Ltd., Worcestershire, UK) was used to mea-sure the electrical charge (f-potentials) of the samples before andafter heating. Samples were diluted 200-times using an aqueoussolution (pH 7.0) and then injected into a folded capillary cell (Mal-vern Instruments, Ltd., Worcestershire, UK) for analysis. The refrac-tive index, viscosity, and relative dielectric constant of thecontinuous phase were set at 1.330, 0.8872 mPa s and 78.5, respec-tively as provided by the software database (version 6.30, Zetasiz-er, Malvern Instruments, Ltd., Worcestershire, UK). The instrumentsoftware was used to calculate the f-potentials from the measuredelectrophoretic mobility data.

2.4. Particle size measurement

The particle size distribution of the samples before and afterheating was measured using static light scattering (Mastersizer2000, Malvern Instruments, Ltd., Worcestershire, UK). Modelsauces and pure starch suspensions after heat treatment were di-luted 2-fold with aqueous solution (pH 7.0) prior to analysis toeliminate multiple scattering effects. Other samples were directlymeasured without pre-dilution. A few drops of the samples werepipetted into a water diluting accessory (Hydro 2000 SM, MalvernInstruments, Ltd., Worcestershire, UK) to obtain a light obscurationlevel of 11–14%. The instrument software (Mastersizer 2000, ver-sion 5.60) calculated the particle size distribution based on Mietheory (ISO, 2009). A refractive index of 1.33 was used for the con-tinuous (water) phase (ISO, 2009), and a refractive index of 1.47was used for the dispersed phase, which is the value for a pureoil (Weast, 1985).

The particle sizes determined by light scattering on mixed par-ticulate systems should only be treated with caution due to a num-ber of limitations of this method: (1) the refractive index of the fat

B.-c. Wu et al. / Food Chemistry 141 (2013) 3393–3401 3395

droplets and starch granules will be different, and it was assumedthat they are similar; (2) the refractive index and morphology ofthe particles varies from sample to sample due to swelling ofstarch granules and flocculation of fat droplets; and, (3) dilutionand stirring of samples during measurements may alter the struc-tural organisation of the dispersed phase. Consequently, particlesize results obtained by static light scattering should only be con-sidered to provide a rough approximation of the true size of theparticles in highly aggregated multicomponent systems.

2.5. Rheological properties

2.5.1. Experimental measurementsThe rheological properties of the samples were characterised

using a dynamic shear rheometer (Kinexus, Malvern Instruments,Ltd., Worcestershire, UK) with a cup-and-bob geometry at 25 �C.The bob had a diameter of 25 mm and the cup had a diameter of27.5 mm. The shear stress was measured as a function of shear rate(0.01–100 s�1) for the samples. The rheological data was presentedas shear stress versus shear rate profiles, and as the apparent vis-cosity at a constant shear rate (20 s�1) selected to mimic oral con-ditions (Rao, 2007; Shama & Sherman, 1973). Native starchgranules tended to rapidly settle during the rheological measure-ments causing inaccurate readings, and so only the heat-treatedsamples containing swollen starch granules were analysed byrheology.

2.5.2. Analysis of rheology dataSome of the samples behaved as Newtonian or non-Newtonian

fluids, whereas others behaved as non-ideal plastics (exhibiting ayield stress). Different rheological models were therefore used toanalyse the shear stress versus shear rate profiles depending onthe nature of the sample. A power-law model was used to describethe rheological properties of the fluids:

Power � law model : s ¼ kcn ð1Þ

Here s is the shear stress (Pa), c is the shear rate (s�1), K is theconsistency index (mPa sn), and n is the power law index. The con-sistency index provides a measure of the viscosity of materials atvery low shear rates (a high consistency index indicates a high vis-cosity of the sample). The power law index is a measure of thedeviation of the fluid from ideal behaviour: n = 1 is ideal (Newto-nian); n < 1 is shear thinning; and n > 1 is shear thickening. TheHershel–Bulkley (HB) model was used to describe the rheologicalproperties of non-ideal plastics that showed a yield stress:

HB model : s� s0 ¼ Kcn ð2Þ

Here s is the shear stress (Pa), s0 is the yield stress (Pa), c is theshear rate (s�1), K is the consistency index (mPa sn), and n is thepower law index. The HB model was selected to represent the rhe-ological behaviour of these systems since it contains analogousparameters to the power-law model (i.e., K and n). The unknownparameters, in the above equations, were found by finding thebest-fit (least squares analysis) of the models to the experimentalmeasurement of shear stress versus shear rate using the instrumentsoftware (rSpace, version 1.40, Malvern Instruments, Ltd., Worces-tershire, UK). The power-law model was chosen if the value of yieldstress obtained by the HB model fitting was unrealistic, i.e. s0 < 0.The regression coefficients (r2) of the model fittings were all >0.99.

For the lowest viscosity samples (emulsions), reliable rheologi-cal measurements could only be made at shear rates from 5 to100 s�1. Therefore, the power law model was only applied to datawithin this shear rate range.

2.6. Optical properties and storage stability determination

A colorimeter (ColorFlez EZ, Model 45/0 LAV, HunterLab, Res-ton, VA, USA) with a geometry of 45�/0� and pulsed xenon as thelight source was used to measure the lightness (L⁄) of the samplesafter heating. In general, lightness values range from 0% (black) to100% (white). A fixed amount of sample was placed within a glasssample cup, which was then covered with a white cover prior tomeasurement.

All heated samples were placed in sealed tubes and were storedat ambient temperature for 30 days. The images of these sampleswere then recorded by digital photography (Panasonic DMC-ZS8,Panasonic, NJ, USA) to provide an indication of their storagestability.

2.7. Microstructure analysis

The microstructure of the samples was examined by DifferentialInterference Contrast (DIC) and confocal microscopy (Nikon micro-scope D-Eclipse C1 80i, Nikon Corporation, Melville, NY, USA). ForDIC images, samples were mixed in a glass test tube using a vortexto prepare a homogeneous mixture and then a small aliquot of thissample was placed on a microscope slide and covered by a glasscover slip. A drop of Type A immersion oil (Nikon, Melville, NY,USA) was placed on top of the cover slip, and the specimens wereobserved using an oil immersion objective lens (60�, 1.40 NA)along with a 1.0� camera zoom. DIC microscopy images of emul-sions were then obtained using a CCD camera (CCD-300-RC,DAGE-MTI, Michigan City, IN, USA) and the images were processedby a digital image processing software (Micro Video InstrumentsInc., MA, USA).

For confocal images, oils were dyed prior to emulsion forma-tion. The hydrophobic dye Nile red was added to the oil phase ata concentration of 0.1 mg per gram of oil. The mixture was thenstirred overnight in dark to completely dissolve the dye. The dyedoil was then used to prepare model dispersions as described in Sec-tion 2.2. A small amount of sample was pipetted onto a microscopeslide, covered with a cover slip, and then sealed with clear nail pol-ish. An oil immersion objective lens (60�, 1.40 NA) was used toview the specimen. After the image was focused, Nile red was ex-cited by an air cooled argon ion laser (Model IMA 1010 BOS, MellesGriot, Carlsbad, CA, USA) at 488 nm, its emission spectra were de-tected in the 605 channel equipped with a long pass (LP) filter (HQ605LP/75 m). All images were taken and processed by the imageprocessing software (EZ-CS1 version 3.8, Nikon, Melville, NY, USA).

2.8. Statistical analysis

For the particle size and rheology measurements each individ-ual sample was analysed twice, while for the particle charge andcolour measurements each individual sample was analysed threetimes. The whole experiment was repeated twice using newly pre-pared samples. The mean and standard deviations were then calcu-lated from this data.

3. Results and discussion

3.1. Influence of calcium on particle characteristics

3.1.1. Fat droplets in simple emulsionsThe influence of calcium chloride and thermal treatment on the

microstructure and physiochemical properties of simple emulsions(WPI-coated-fat droplets in water) was initially investigated. In theabsence of calcium chloride and heating, the fat droplets had a highnegative charge (��70 mV) because the pH was well above the

0.1

1

10

100

1000

0 0.5 1 1.5 2 2.5 3

Mea

n P

arti

cle

Dia

met

er (

µ µm)

CaCl2 concentration (mM)

E (U) E (H) Mix (U) Mix (H) S (U) S (H)

Fig. 2. Influence of calcium chloride content and thermal treatment (90 �C, 5 min)on the mean particle diameter of (d32) of simple emulsions (2 wt% fat), mixeddispersions (2 wt% fat + 4 wt% starch) and starch suspensions (4 wt% starch). Here:E = Emulsion; Mix = Fat droplets + Starch granules; S = Starch granules;U = Unheated; and H = Heated.

3396 B.-c. Wu et al. / Food Chemistry 141 (2013) 3393–3401

isoelectric point (�pH 5) of the adsorbed WPI molecules (Charoenet al., 2011; Demetriades, Coupland, & McClements, 1997a). Therewas little change in the charge on the fat droplets (��69 mV)when they were heated in the absence of calcium (Fig. 1). Withoutheating, there was a slight decrease in the negative charge (from�70 to �62 mV) on the fat droplets when calcium was increasedfrom 0 to 3 mM, which can be attributed to electrostatic screeningand ion binding effects (McClements, 2005). With heating, therewas a more pronounced decrease in the negative charge on thefat droplets (from�69 to �44 mV) with increasing calcium (partic-ularly between 0.5 and 1 mM), which suggests that calcium ionsmay have been able to bind to absorbed proteins more stronglyafter they are thermally denatured. A slight decrease in pH (from7.0 to 6.6) occurred when the calcium concentration was increasedfrom 0 to 3 mM, which may have been due to interactions of themultivalent calcium ions with carboxyl groups on adsorbed pro-teins: –COOH + Ca2+ ? –COO�Ca2+ + H+.

Without heating, the mean particle diameter (Fig. 2) and parti-cle size distribution (data not shown) of the emulsions was inde-pendent of the calcium concentration (0–3 mM). Presumably, thehigh negative charge on the droplets resulted in a strong electro-static repulsion that stabilized them against flocculation. In addi-tion, there must have been insufficient calcium ions present topromote bridging flocculation. At higher calcium concentrations(>5 mM) droplet flocculation was observed in the unheated emul-sions as demonstrated by an increase in particle size and creaminginstability (data not shown). After heating, there was a strongdependence of particle size on the calcium content. At low calciumconcentrations (60.5 mM Ca2+), the mean particle diameter (Fig. 2)and particle size distribution (data not shown) was unaffected byheating, being similar to the emulsions containing no calcium. Atintermediate to high calcium concentrations (1–3 mM), therewas a large increase in mean particle diameter (Fig. 2) and a pop-ulation of large particles (>100 lm) in the particle size distribution(data not shown) of the heated emulsions. This pronounced

-80

-70

-60

-50

-40

-30

-20

-10

00 0.5 1 1.5 2 2.5 3

ζ-po

tent

ial (

mV

)

CaCl2 concentration (mM)

E (U) E (H)Mix (U) Mix (H)S (U) S (H)

Fig. 1. Effect of calcium chloride concentration and thermal treatment (90 �C,5 min) on the electrical characteristics (f-potential) of particles in simple emulsions(2 wt% fat), mixed dispersions (2 wt% fat + 4 wt% starch) and starch suspensions(4 wt% starch). Here: E = Emulsion; Mix = Fat droplets + Starch granules; S = Starchgranules; U = Unheated; and H = Heated.

increase in particle size was attributed to fat droplet flocculationinduced by a combination of heating and calcium ions.

When whey protein-coated fat droplets are heated above theirthermal denaturation temperature the layer of adsorbed proteinsundergoes conformational changes that lead to exposure of non-polar and sulfhydryl groups, thereby increasing the hydrophobicattraction between the droplets (Demetriades, Coupland, & McCle-ments, 1997b; McClements, Monahan, & Kinsella, 1993). At rela-tively low salt levels, the electrostatic repulsion between thedroplets is sufficiently strong to overcome the attractive forces(van der Waals and hydrophobic), and so the emulsion remainsstable to flocculation. However, once a critical salt concentrationis exceeded the electrostatic repulsion is no longer strong enoughto overcome the attractive forces (van der Waals, hydrophobic,and ion bridging), and so the emulsion flocculates. Our results sug-gest that both thermal denaturation (to increase the hydrophobicattraction) and calcium addition (to reduce the electrostatic repul-sion) is required to promote droplet flocculation in the systemsstudied.

The influence of calcium concentration on the microstructure ofthe emulsions was observed by DIC microscopy (Fig. 3a). Fat drop-lets remained un-aggregated at low calcium concentrations(60.5 mM), which was in agreement with the static light scatteringmeasurements. At high calcium concentrations (2 and 3 mM), themajority of droplets were present in large flocs. At intermediatecalcium concentrations (1 and 1.5 mM) large flocs were also ob-served, but there also appeared to be some individual fat dropletsand small clusters of fat droplets present. These smaller particleswere not detected by the light scattering instrument, probably be-cause the larger aggregates dominated the signal or because theydissociated prior to analysis. In general, light scattering resultson highly flocculated systems should be treated with caution be-cause: (i) flocs have ill-defined refractive indices; (ii) flocs haveirregular (non-spherical) shapes; (iii), the size, number, and shapeof flocs may change during measurements due to stirring anddilution (Chantrapornchai, Clydesdale, & McClements, 2001b).

Fig. 3. DIC images of simple emulsions with different calcium concentrations after heat treatment. The red arrow shows small particles (non-aggregated fat droplets or smallflocs). Black arrows show large aggregates. The scale bars are 50 lm in length. (For interpretation of the references to colour in this figure legend, the reader is referred to theweb version of this article.)

B.-c. Wu et al. / Food Chemistry 141 (2013) 3393–3401 3397

Consequently, it is important to support light scattering data withmicroscopy data on these types of systems.

3.1.2. Starch granules in starch suspensionsIn this series of experiments, the influence of calcium and heat

treatment on the microstructure and physicochemical propertiesof starch suspensions was examined. The starch granules hadslightly negative charges (�4 to �6 mV) that were not strongly af-fected by calcium addition (0–3 mM) or thermal treatment(Fig. 1). A slight negative charge on native starch granules at aneutral pH has also been reported in other studies (Considineet al., 2011; Takeuchi, 1969; Wu et al., 2013). Presumably, thenegative charge on the granules was due to ionisation of phos-phate groups (–PO�3 ) and/or hydroxyl groups (–OH) on the modi-fied starch. The relatively weak negative charge on the starchgranules may account for the fact that we did not observe a largechange in f-potential (indicative of ion binding) when cationic cal-cium ions were added.

The particle size of the starch granules was mainly affected byheating, rather than calcium addition. The mean diameter of theheated starch granules (�37 lm) was considerably higher thanthat of the unheated ones (�17 lm), which can be attributed towater absorption and swelling during starch gelatinization. Thestarch granules used in our study were cross-linked with phos-phate and substituted with hydroxypropyl, which accounts forthe fact that they did not disintegrate during the gelatinizationprocess (Wurzburg, 2006). The presence of calcium (0–3 mM) didnot have a major impact on the size of the starch granules beforeor after heating, suggesting that it did not promote aggregationor alter swelling.

3.1.3. Fat droplet – starch granules in mixed systemsIn this section, the impact of calcium and heating on the micro-

structure and physicochemical properties of model sauces contain-ing a mixture of fat (2 wt%) and starch (4 wt%) was examined.Before heating, the charge on the particles in the mixed systemwas similar to that of the fat droplets in the emulsions (Fig. 1),

which suggested that the f-potential was dominated by the drop-lets. As in the emulsions, the charge on the particles in the un-heated mixed systems became slightly less negative as thecalcium concentration was increased from 0 to 3 mM (Fig. 1), sug-gesting that there may have been some electrostatic screening andion binding. The magnitude of the f-potential in the mixed systemswas somewhat lower than in the simple emulsions at equivalentcalcium concentrations, which may have been due to the presenceof the starch granules (Wu et al., 2013).

After heating, the measured charge on the particles in the mixedsystems was strongly dependent on calcium concentration (Fig. 1),which may have occurred for a number of reasons. First, calciumion binding to fat droplet surfaces may have increased after heat-ing due to protein unfolding (Section 3.1.1), which would reducetheir negative charge. This effect may have been accentuated inthe presence of the starch since the calcium concentration in theaqueous phase would have increased after starch granule swelling(since there was less free water remaining). Second, heated mix-tures containing high calcium levels were highly viscous aggre-gated systems, and therefore there may have been someproblems making reliable f-potential measurements.

The influence of calcium and heating on the mean particlediameter (Fig. 2) and particle size distribution (data not shown)of mixed systems was also investigated. Before heating, all sampleshad bi-model distributions, with a small peak around 0.2 lmattributed to the fat droplets and a large peak around 17 lm attrib-uted to the starch granules. Calcium did not affect the particle sizeof the mixed systems before heating, which was in agreement withthe earlier observations for pure emulsion droplets and starchgranules (Fig. 2). After heating, the mean particle diameter of themixtures increased appreciably, and a single monomodal peakwas observed in the particle size distribution at high particle diam-eters (30–40 lm). As reported previously, this peak was presum-ably due to the presence of swollen starch granules thatdominated the light scattering signal (Wu et al., 2013). Appreciablefat droplet aggregation occurred in the simple emulsions afterheating in the presence of P1 mM calcium (Fig. 2). We would

3398 B.-c. Wu et al. / Food Chemistry 141 (2013) 3393–3401

therefore have expected fat droplet aggregation to also have oc-curred in the heated mixed systems containing high calcium levels.However, droplet aggregation could not be observed in the lightscattering data, presumably because the swollen starch granulesdominated the overall signal. For this reason, we used optical(DIC) microscopy to provide additional information about the orga-nisation of the fat droplets in the mixtures (Fig. 4a).

All the mixed systems showed similar microstructures beforeheating regardless of the initial calcium concentration. As an exam-ple, a representative image showing the microstructure of the un-heated mixed system containing no calcium is shown in Fig. 4a.The fat droplets were homogeneously dispersed throughout theaqueous phase that surrounded the native starch granules. Afterheating, the microstructure of the mixed systems was highlydependent on the calcium concentration. At a low calcium concen-tration (0.5 mM), the fat droplets in the mixed system appeared tobe individual particles that diffused freely within the spaces be-tween the swollen starch granules, similar to what was observedin the mixed system containing no calcium. At an intermediate cal-cium concentration (1.0 mM), confocal microscopy showed thatthe fat droplets adhered to the surfaces of the swollen starch gran-ules, and formed a fat-starch complex that appeared to link adja-cent starch granules together (Fig. 4b). Additionally, some fatdroplets that moved freely in the spaces between these complexeswere observed. As the calcium concentration increased to 1.5 mM,no free fat droplets were seen in the system. Instead, the dropletsaggregated into large flocs that co-existed with those coating thestarch granule surfaces. At a high calcium concentration (2 and3 mM), the fat droplets formed a three-dimensional network thatoccupied the entire region surrounding the swollen starchgranules.

These results suggest that there was an attractive interactionbetween the fat droplets and starch granules after heating in thepresence of sufficiently high calcium levels, which caused the fatdroplets to form a coating around the swollen starch granules. Thisattraction may have been due to van der Waals interactions be-tween the flocs and starch granules, electrostatic attraction be-tween negatively charged groups on starch (e.g. phosphategroups) and positively charged groups on protein (e.g., aminogroups) (Noisuwan, Hemar, Bronlund, Wilkinson, & Williams,

Fig. 4a. DIC images of mixed dispersions at different calcium concentrations. Symbol ‘‘Sscale bars are 50 lm in length. (For interpretation of the references to colour in this fig

2007; Zaleska, Ring, & Tomasik, 2001), or ion bridge formationinvolving anionic fat droplets, cationic calcium ions, and anionicstarch granules. Indeed, previous studies have reported that cal-cium ions can bind to hydroxypropyl distarch phosphate (Hood &Oshea, 1977). At an intermediate calcium concentration (e.g.,1 mM), the attraction between the fat droplets is relatively weak,and so they can move freely through the aqueous phase and attachto starch granule surfaces. At a high calcium concentration (e.g.,3 mM), the attraction between fat droplets is relatively strong,and so they form a three-dimensional network in the aqueousphase and are therefore not free to move to starch granule surfaces.

In summary, these results show that the aggregation state of thefat droplets in the interstitial region separating swollen starchgranules can be modulated by calcium addition.

3.2. Influence of calcium on rheology

The influence of calcium on the rheology of heated mixed dis-persions (2 wt% fat, 4 wt% starch), starch suspensions (4 wt%starch), and simple emulsions (2 wt% fat) was compared. Samplescontaining native starch granules were not analysed because theywere highly unstable to sedimentation and therefore reliable rhe-ology measurements could not be made. The rheological behaviourof the other samples was characterised by measuring their shearstress versus shear rate profiles. Representative rheological profilesof heated mixed dispersions containing different calcium levels areshown in Fig. 5a, and the apparent viscosities (at 20 s�1) of all thesamples are compared in Fig. 5b. Rheological parameters were cal-culated using the mathematical models described in Section 2.5.2(Table 1). Reliable shear stress versus shear rate measurementscould not be made on heated simple emulsions containing highcalcium levels due to extensive droplet aggregation and creaming.

The simple emulsions at low calcium concentrations had rela-tively low consistencies (<1 mPa at 1 s�1) and exhibited Newtonianfluid behaviour (n � 1) (Table 1), which can be attributed to the factthat they contained a relatively low concentration of non-aggre-gated fat droplets (Pal, 2011). The starch suspensions had relativelyhigh consistencies (240–370 mPa at 1 s�1) and exhibited non-New-tonian (shear thinning) behaviour (n = 0.78–0.85), but their rheol-ogy did not depend strongly on the calcium concentration

’’ denotes gelatinized starch granules. Red arrows show location of fat droplets. Theure legend, the reader is referred to the web version of this article.)

Fig. 4b. DIC and Confocal microscopy images of mixed dispersions at 1 mM calcium concentration. Symbol ‘‘S’’ denotes gelatinized starch granules. Arrows show location offat droplets. All images were obtained at the same location of the microscope slide. The two confocal images were obtained at different focal points. The scale bars are 50 lmin length. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

0

10

20

30

40

50

0 20 40 60 80 100

Shea

r str

ess (

Pa)

Shear rate (1/s)

0 mM

0.5 mM

1 mM

3 mM

Fig. 5a. Shear stress versus shear rate profiles of heated mixed dispersionscontaining different calcium chloride levels.

0

200

400

600

800

1000

1200

1400

1600

1800

0 0.5 1 1.5 2 3

App

aren

t Vis

cosi

ty (P

a.s)

Calcium Concentration (mM)

E

S

Mix

Fig. 5b. Influence of calcium content on the apparent shear viscosity at a fixedshear rate (20 s�1) of emulsions, starch suspensions, and mixed dispersions afterthermal treatment (90 �C, 5 min). Here: E = Emulsion (2 wt% fat droplets), S = Starchdispersion (4 wt% starch granules) and Mix = Mixed dispersion (2 wt% fat drop-lets + 4 wt% starch granules).

B.-c. Wu et al. / Food Chemistry 141 (2013) 3393–3401 3399

(Table 1, Fig. 5b). As discussed earlier, calcium addition did not havea major effect on the size or aggregation state of the starch granules(Section 3.1.2) and therefore would not be expected to have a majorinfluence on the rheological properties of starch suspensions. Con-versely, the rheology of the mixed dispersions containing fat drop-lets and starch granules depended strongly on the calcium content(Table 1). The mixed dispersions all had high consistencies (260–3,900 mPa at 1 s�1) and exhibited shear-thinning behaviour(n = 0.47–0.82). At low calcium levels the mixed systems behavedas non-Newtonian (shear-thinning) liquids, but at higher levelsthey behaved as non-ideal plastics with a yield stress (Fig. 5a).Interestingly, the mixed systems (2 wt% oil and 4 wt% starch) con-taining intermediate calcium levels (1–3 mM) prepared in thisstudy had higher viscosities than higher fat mixed systems(10 wt% oil and 5 wt% starch) containing non-aggregated fat drop-lets reported previously (Chung et al., 2012b). This suggests that acontrolled droplet aggregation may be a useful strategy to create re-duced fat products.

Differences in the rheological behaviour of mixed dispersionscontaining different calcium levels were related to differences intheir microstructures. The mixed dispersions containing a low cal-cium concentration (0.5 mM) had fairly similar flow characteristicsto those containing no calcium (Fig. 5a), which can be attributed tothe fact that the fat droplets were not strongly aggregated (Figs. 4a

and b). The mixed systems containing an intermediate calciumconcentration (�1–3 mM) had high yield stresses and apparentviscosities (Fig. 5b, Table 1), which may have been due to the for-mation of a coating of fat droplets around the starch granules thatlinked them together (Figs. 4a and b). In this case, the starch gran-ules may have formed a three-dimensional particle network thatgave the overall system some elastic properties (Dickinson, 2012;Dickinson & Chen, 1999). When the calcium concentration was in-creased further, there was a decrease in the apparent viscosity andyield stress, which may be associated with changes in the struc-tural arrangement of the fat droplets in the system. Rather thanaccumulating on the starch granule surfaces, the fat dropletsformed a fairly uniform three-dimensional network of aggregateddroplets throughout the interstitial region separating the starchgranules. This network may have provided some elastic propertiesto the mixed dispersion, but the mechanical strength of the net-work formed by the fat droplets may have been weaker than theone formed by the starch granules. Indeed, the starch granulesmay even have acted as inactive fillers that did not contribute tothe overall strength of the fat droplet network (Dickinson, 2012;Dickinson & Chen, 1999).

Table 1Summary of rheological parameters determined by fitting mathematical models tothe shear stress versus shear rate data, and of the instrumental lightness.

Calcium(mM)

Rheology model parameters Lightness(%)

Yield stress(Pa)

K(mPa sn) n

Emulsions0 – 0.67 ± 0.03P 1.11 ± 0.01P 91.1 ± 0.00.5 – 0.57 ± 0.03P 1.15 ± 0.01P 91.4 ± 0.01 – 4.2 ± 3.8P 0.79 ± 0.28P 89.0 ± 0.41.5 – ND ND 80.6 ± 0.02 – ND ND 79.7 ± 5.53 82.6 ± 3.5

Starch suspensions0 – 288 ± 93P 0.82 ± 0.04P 32.8 ± 0.80.5 – 370 ± 180P 0.78 ± 0.07P 33.3 ± 1.71 – 226 ± 9P 0.85 ± 0.01P 34.4 ± 2.21.5 – 260 ± 11P 0.83 ± 0.01P 33.4 ± 3.12 – 237 ± 4P 0.84 ± 0.01P 33.3 ± 2.53 – 290 ± 100P 0.82 ± 0.06P 33.4 ± 1.7

Mixed systems0 – 260 ± 30P 0.82 ± 0.03P 90.4 ± 0.30.5 – 510 ± 80P 0.65 ± 0.05P 91.0 ± 0.91 22.6 ± 1.5H 1670 ± 40H 0.63 ± 0.03H 89.8 ± 0.21.5 15.8 ± 3.4H 1230 ± 40H 0.76 ± 0.01H 88.4 ± 1.22 9.0 ± 0.8H 1960 ± 370H 0.64 ± 0.04H 88.1 ± 1.43 2.9 ± 1.7H 3900 ± 150H 0.47 ± 0.02H 86.4 ± 0.9

Note: Letters indicates the type of rheology models (P: Power-law model; N:Newtonian model and H: Hershel–Bulkley model) that used to obtain the corre-sponding values.

3400 B.-c. Wu et al. / Food Chemistry 141 (2013) 3393–3401

3.3. Influence of calcium addition on appearance

The appearance of an emulsion-based food is one of its mostimportant sensory attributes, since it strongly influences a con-sumer’s judgment of product quality. The influence of calciumaddition on the optical properties (lightness) of the simple emul-sions, starch suspensions, and mixed dispersions were thereforedetermined using a colorimeter immediately after heat treatment(Table 1). In the absence of calcium, the simple emulsions had ahigh lightness (�91%), which can be attributed to the strong lightscattering efficiency of the small fat droplets they contain (McCle-ments, 2002). At low calcium levels (0.5 mM), the lightness of theemulsions (�91%) was similar to that of the control emulsions dueto the fact that the droplets were not strongly aggregated andtherefore their light scattering efficiency was not changed appre-ciably. At higher calcium levels there was an appreciable decreasein the lightness of the emulsions, which suggests that there was achange in the light scattering pattern after the fat droplets aggre-gated (Chantrapornchai, Clydesdale, & McClements, 2001a; McCle-ments, 2002).

The lightness of the starch suspensions was relatively indepen-dent of the calcium concentration, and was much lower than thatof the simple emulsions (Table 1). This is because starch granulesabsorb water during gelatinization, which reduces their refractiveindex contrast and increases their particle size thereby reducingtheir light scattering efficiency (McClements, 2002). Similar resultshave been reported in other recent studies of gelatinized starchsuspensions (Chung et al., 2013; Wu et al., 2013). The influenceof calcium addition on the optical properties of the mixed disper-sions followed a similar trend to that of the simple emulsions:the lightness decreased as the calcium concentration increased.Again, this effect can be attributed to the lower light scattering effi-ciency of the flocculated fat droplets compared to non-flocculatedones (Chantrapornchai, Clydesdale, & McClements, 2001a).

Finally, we examined the influence of calcium addition on theoverall visual appearance of the samples after they were storedfor 30 days at ambient temperature (data not shown). When there

was no or low (60.5 mM) calcium present, the simple emulsionsremained stable throughout the observation period. At intermedi-ate calcium levels (1 mM), part of the fat droplets formed a visiblecream layer on the surface of the emulsions, while the rest re-mained uniformly dispersed throughout the system. This con-firmed the observations made using optical microscopy (Figs. 4aand b), which also showed the presence of some individual fatdroplets and small clusters at similar calcium levels. As discussedearlier, at this calcium concentration the attractive interactions be-tween the fat droplets will be relatively weak and so a fraction ofthem may not aggregate. At a higher calcium concentration(P1.5 mM), the simple emulsions were visibly unstable to cream-ing with a white layer of fat droplets forming on their surface with-in a day or so. Eventually, all the fat droplets were incorporatedinto the cream layer, leaving a transparent continuous phase atthe bottom of the samples after 30 days of storage (data notshown). The increased instability of the emulsions to creaming inthe presence of high calcium levels can be attributed to the greaterrate of gravitational separation of larger particles (Charoen et al.,2011; Demetriades et al., 1997b; McClements, 2005). The mixeddispersions were visibly stable to separation after 30 days of stor-age at all calcium concentrations, which can be attributed to theability of the starch granule network to inhibit fat droplet move-ment. After the storage stability test, the tubes containing themixed dispersions were inverted. The mixed dispersions contain-ing low calcium levels (60.5 mM) were highly viscous materialsthat slowly flowed out of the tubes, which can be attributed tothe fact they had no appreciable yield stress (Table 1). On the otherhand, the mixed dispersions containing higher calcium levels(P1 mM) did not flow, and appeared to have a relatively stronggel-like texture, which can be attributed to the fact they did havea yield stress (Table 1).

The surface appearance of the mixed dispersions also dependedon the calcium content: the systems containing 1 mM calciumwere grainy, whereas those containing higher calcium contentswere smooth. This macroscopic difference in appearance was prob-ably due to microscopic differences in the structural organisationof the particles in the mixed dispersions: at a 1 mM calcium con-centration, the individual starch granules were coated by fat drop-lets, whereas at higher calcium levels the starch granules weretrapped within a network of aggregated fat droplets (Figs. 4a andb).

4. Conclusions

This study was designed to determine the impact of calciumaddition on the microstructure and physicochemical properties ofmixed dispersions containing starch granules and fat droplets,which are models for many commercial products such as sauces,dressings, and desserts. Initially, the influence of calcium contenton protein-coated fat droplets and starch granules was studiedseparately. The properties of the starch granule suspensions wereprimarily determined by thermal processing, rather than by cal-cium addition. On the other hand, the properties of the emulsionswere strongly affected by both calcium addition and thermal pro-cessing. The degree of droplet aggregation in the emulsions couldbe changed by altering the calcium ion concentration, which wasattributed to charge neutralisation, ion binding, and salt bridgeformation.

Light scattering, optical microscopy, and rheology measure-ments showed that droplet aggregation in mixed dispersions con-taining fat droplets and starch granules also depended on thecalcium content. At low calcium levels, no droplet aggregationwas observed and the mixed dispersions were non-ideal viscousfluids. At intermediate calcium levels, the fat droplets aggregated

B.-c. Wu et al. / Food Chemistry 141 (2013) 3393–3401 3401

onto the surfaces of the swollen starch granules and appeared toform a starch granule network that gave the system some mechan-ical strength. At high calcium levels, the fat droplets formed athree-dimensional network in the aqueous phase that appearedto trap the swollen starch granules, which caused a decrease inmechanical strength. This study provides an improved understand-ing of the influence of the structural organisation of fat dropletsand starch granules in complex mixed dispersions on their physi-cochemical properties, which may be useful for the rational devel-opment of reduced fat products with improved physicochemicalproperties and sensory attributes.

Acknowledgements

We thank ConAgra foods for financially supporting this work,and Cheryl Chung, Gordon Smith, Kerstin Olson, and Ware Florafor valuable scientific and technical discussions about the work.

References

Arancibia, C., Jublot, L., Costell, E., & Bayarri, S. (2011). Flavor release and sensorycharacteristics of o/w emulsions. Influence of composition, microstructure andrheological behavior. Food Research International, 44(6), 1632–1641.

Chantrapornchai, W., Clydesdale, F. M., & McClements, D. J. (2001a). Influence offlocculation on optical properties of emulsions. Journal of Food Science, 66(3),464–469.

Chantrapornchai, W., Clydesdale, F. M., & McClements, D. J. (2001b). Influence ofrelative refractive index on optical properties of emulsions. Food ResearchInternational, 34(9), 827–835.

Charoen, R., Jangchud, A., Jangchud, K., Harnsilawat, T., Naivikul, O., & McClements,D. J. (2011). Influence of biopolymer emulsifier type on formation and stabilityof rice bran oil-in-water emulsions: Whey protein, gum arabic, and modifiedstarch. Journal of Food Science, 76(1), E165–E172.

Chung, C., Degner, B., & McClements, D. J. (2012a). Instrumental mastication assayfor texture assessment of semi-solid foods: Combined cyclic squeezing flow andshear viscometry. Food Research International, 49(1), 161–169.

Chung, C., Degner, B., & McClements, D. J. (2012b). Rheology and microstructure ofbimodal particulate dispersions: Model for foods containing fat droplets andstarch granules. Food Research International, 48(2), 641–649.

Chung, C., Degner, B., & McClements, D. J. (2013). Physicochemical characteristics ofmixed colloidal dispersions: Models for foods containing fat and starch. FoodHydrocolloids, 30(1), 281–291.

Considine, T., Noisuwan, A., Hemar, Y., Wilkinson, B., Bronlund, J., & Kasapis, S.(2011). Rheological investigations of the interactions between starch and milkproteins in model dairy systems: A review. Food Hydrocolloids, 25(8),2008–2017.

Demetriades, K., Coupland, J. N., & McClements, D. J. (1997a). Physical properties ofwhey protein stabilized emulsions as related to pH and NaCl. Journal of FoodScience, 62(2), 342–347.

Demetriades, K., Coupland, J. N., & McClements, D. J. (1997b). Physicochemicalproperties of whey protein-stabilized emulsions as affected by heating andionic strength. Journal of Food Science, 62(3), 462–467.

Dickinson, E. (2012). Emulsion gels: The structuring of soft solids with protein-stabilized oil droplets. Food Hydrocolloids, 28(1), 224–241.

Dickinson, E., & Chen, J. (1999). Heat-set whey protein emulsion gels: Role of activeand inactive filler particles. Journal of Dispersion Science and Technology, 20(1–2),197–213.

Finucane, M. M., Stevens, G. A., Cowan, M. J., Danaei, G., Lin, J. K., Paciorek, C. J., et al.(2011). National, regional, and global trends in body-mass index since 1980:Systematic analysis of health examination surveys and epidemiological studieswith 960 country-years and 9.1 million participants. Lancet, 377(9765),557–567.

Flett, K. L., Duizer, L. M., & Goff, H. D. (2010). Perceived creaminess and viscosity ofaggregated particles of casein micelles and kappa-Carrageenan. Journal of FoodScience, 75(5), S255–S262.

Frank, D., Appelqvist, I. A. M., Piyasiri, U., Wooster, T. J., & Delahunty, C. (2011).Proton transfer reaction mass spectrometry and time intensity perceptualmeasurement of flavor release from lipid emulsions using trained humansubjects. Journal of Agricultural and Food Chemistry, 59, 4891–4903.

Hood, L. F., & Oshea, G. K. (1977). Calcium-binding by hydroxypropyl distarchphosphate and unmodified starches. Cereal Chemistry, 54(2), 266–271.

ISO (2009). ISO 13320:2009 Particle size analysis – Laser diffraction methods.Mao, Y. Y., & McClements, D. J. (2012a). Fabrication of viscous and paste-like

materials by controlled heteroaggregation of oppositely charged lipid droplets.Food Chemistry, 134(2), 872–879.

Mao, Y. Y., & McClements, D. J. (2012b). Influence of electrostatic heteroaggregationof lipid droplets on their stability and digestibility under simulatedgastrointestinal conditions. Food & Function, 3(10), 1025–1034.

McClements, D. J. (2002). Theoretical prediction of emulsion color. Advances inColloid and Interface Science, 97(1–3), 63–89.

McClements, D. J. (2005). Food emulsions: Principles, practices, and techniques (2nded.). Boca Raton: CRC Press.

McClements, D. J., Monahan, F. J., & Kinsella, J. E. (1993). Disulfide bond formationaffects stability of whey-protein isolate emulsions. Journal of Food Science, 58(5),1036–1039.

Noisuwan, A., Hemar, Y., Bronlund, J. E., Wilkinson, B., & Williams, M. A. K. (2007).Viscosity, swelling and starch leaching during the early stages of pasting ofnormal and waxy rice starch suspensions containing different milk proteiningredients. Starch-Starke, 59(8), 379–387.

Novak, N. L., & Brownell, K. D. (2011). Obesity: A public health approach. PsychiatricClinics of North America, 34(4), 895.

Pal, R. (2011). Rheology of simple and multiple emulsions. Current Opinion in Colloid& Interface Science, 16(1), 41–60.

Rao, M. A. (2007). Rheology of fluid and semisolid foods: Principles and applications (2ed.). New York, NY, USA: Springer Science + Business Media, LLC.

Shama, F., & Sherman, P. (1973). Identification of stimuli controlling the sensoryevaluation of viscosity II. Oral methods. Journal of Texture Studies, 4(1), 111–118.

Simo, O. K., Mao, Y., Tokle, T., Decker, E. A., & McClements, D. J. (2012). Novelstrategies for fabricating reduced fat foods: Heteroaggregation of lipid dropletswith polysaccharides. Food Research International, 48(2), 337–345.

Takeuchi, I. (1969). Interaction between protein and starch. Cereal Chemistry, 46(5),570.

Tarrega, A., & Costell, E. (2006). Effect of composition on the rheological behaviourand sensory properties of semisolid dairy dessert. Food Hydrocolloids, 20(6),914–922.

van Aken, G. A., Vingerhoeds, M. H., & de Wijk, R. A. (2011). Textural perception ofliquid emulsions: Role of oil content, oil viscosity and emulsion viscosity. FoodHydrocolloids, 25(4), 789–796.

Vingerhoeds, M. H., de Wijk, R. A., Zoet, F. D., Nixdorf, R. R., & van Aken, G. A. (2011).How emulsion composition and structure affect sensory perception of lowviscosity model emulsions (vol 22, pg 631, 2008). Food Hydrocolloids, 25(5).1407–1407.

Weast, R. C. (1985). CRC handbook of chemistry and physics (66th ed.). Boca Raton, FL:CRC Press, Inc..

Wu, B., Degner, B., & McClements, D. J. (2013). Microstructure & rheology of mixedcolloidal dispersions: Influence of pH-induced droplet aggregation on starchgranule–fat droplet mixtures. Journal of Food Engineering, 116(2), 462–471.

Wurzburg, O. B. (2006). Modified starches. In A. M. P. G. O. W. P. A. Stephen (Ed.),Food Polysaccharides and Their Applications, 2nd ed. (pp. 87–118).

Zaleska, H., Ring, S., & Tomasik, P. (2001). Electrosynthesis of potato starch-wheyprotein isolate complexes. Carbohydrate Polymers, 45(1), 89–94.