microfluidization of dairy model emulsions. 1.preparation of

Lait (1992) 72,511-531© Elsevier/INRA

511

Original article

Microfluidization of dairy model emulsions.1.Preparation of emulsions and influence

of processing and formulationon the size distribution of milk fat globules

o Robin, V Blanchot, Je Vuillemard, P Paquin

Université Laval, Sainte-Foy, centre de recherche en sciences et technologies du lait (STELA),département de sciences et technologie des aliments, (Québec), Canada G1K7P4

(Received 1st July 1992; accepted 17 September 1992)

5ummary - The influence of certain process variables (pressure and temperature) as weil ascomposition variables (fat, protein and low molecular weight emulsifier concentrations) on the sizedistribution of milk fat globules was studied in a dairy model emulsicn (oil-in-water) produced bymicrofluidization, a mechanical emulsification technique. The use of a central compositeexperimental design allowed us to obtain 2 nonlinear multiple regression equations relating thevolume-weighted average diameter of the fat globules (d-:) as weil as the relative size distributionwidth (cv) to the emulsification pressure (7.8-76.3 MPa) and temperature (35-100 oC), and tosodium caseinate (0.5-3.9 wt%), butter oil (5.2-14.7 wt%) and monoglyceride (0.08-0.88 wt%)concentrations. These 2 functions account respectively for 93.7 and 81.7% of the variation in theaverage diameterand in the size distribution width of the microfluidized fat globules and were usedto explain certain interactions between the different variables affecting the size of the microfluidizedfat globules. They were also used to demonstrate the existence of the optimal conditions thatcorrespond to the extremes of the average particle diameter and of the distribution width of the fatglobules. Finally, these 2 functions allowed us to predict fat globule size parameters as a function ofprocess and formulation conditions.

emulsification 1microfluidization 1oil-in-water emulsion 1fat globule

Résumé - Microfluidisation d'émulsions laitières modèles. 1. Préparation des émulsions etinfluence des facteurs de procédé et de formulation sur la distribution de la taille des glo-bules de gras. L'influence de certaines variables opératoires (pression et température) et des va-riables de composition (teneurs en huile de beurre, en protéine et en émulsifiant de faible poids mo-

Symbols used and SI units: [BD] = butter oil concentration (wt%); cv= coefficient of variation or rela-tive standard deviation of the volume-weighted size distribution (%); if" = volume-weighted averagefat globule diameter (m); D = diffusion coefficient (m2.s-1); F = number of factorial points of thedesign and Fisher ratio; k = number of independent variables or conductivity of a solution (S.m-1 =m--3.kg-1. s3.A2); [MGS] = monoglyceride concentration (wt%); [Pro~ = protein concentration (wt%);P = emulsification pressure (Pa = m-1.kg.s-2); Q = flow rate (m3.s-1); S = interfacial area (m2); T =emulsification temperature (K); a = distance separating a star point from a factorial point in terms ofcoded variables (dimensionless); e = power density (W.m--3 = m-1.kg.s--3); tP = volume fraction of the

. disperse phase (dimensionless); y= interfacial tension (N.m-1 = kg.s-2); p = mass density (kg.m--3).

512 o Robinetai

léculaire) sur la distribution de la taille des globules de gras d'une émulsion laitière modèle (huiledans l'eau) a été étudiée pour une technologie d'émulsification mécanique donnée: la microfluidisa-tion. L'utilisation d'un dispositif expérimental central composite a permis l'obtention de 2 équations derégression multiple non linéaires reliant le diamètre moyen des globules de gras (d,,) ainsi que l'éten-due relative de la dispersion des tailles (cv)à la pression (7,8-76,3MPa) et à la température d'émut-sification (35-100 OC), aux teneurs en caséinate de sodium (0,5-3,9% de la masse totale), en huilede beurre (5,2-14,7%) et en monoglycéride (0,08--0,88%). Ces 2 fonctions expliquent respectivement93,7% et 81,7%des variations du diamètre moyen des globules de gras microfluidisés et de l'étenduerelative de la distribution des tailles dans les valeurs considérées des paramètres. Ces fonctions onten outre permis d'expliquer certaines interactions entre les différentes variables précitées sur la distri-bution de la taille des globules de gras microfluidisés. Elles ont également permis de mettre en évi-dence l'existence de conditions optimales correspondant à des extremums du diamètre particulairemoyen et de l'étendue de la distribution de la taille des globules de gras. Finalement, ces 2 fonctionsnous ont permis de prédire la distribution de la taille des globules de gras en fonction des conditionsde procédés et de formulation.

émulsification / microfluidisation / émulsions huile dans l'eau / globules de gras

INTRODUCTION

For a number of industries, the homogen-ization, and more generally the mechani-cal emulsification of Iiquids, dietary as weilas others, has become over the past manyyears an important technological process.The use of this type of process in the food,pharmaceutical and cosmetic industries ingeneral and in certain dairy operations inparticular is essential for the production ofemulsions with certain desirable rheologi-cal properties (creams and ointments), tex-tures (ice cream, mayonnaise) and de-grees of stability (milk, saiad dressing)(Dickinson and Stainsby, 1988). The prin-ciple of homogenization and mechanicalemulsification processes which, accordingto Gaulin, must allow "de fixer la composi-tion des liquides" remains simple: in thecase of milk, the homogenization processresults in a considerable decrease in thesize of the fat globules found in the initiàlsuspension, and has the effect of increas-ing the creaming stability of the emulsion(Tadros and Vincent, 1983). Aside fromconventional homogenizers, relatively fewmechanical emulsification processes havebeen and are used in an industrial setting

(Walstra, 1983). During the 1980s, how-ever, other mechanical emulsification pro-cesses appeared. One of them, called mi-crofluidization (Cook and Lagace, 1985;Washington, 1987), was initially used prin-cipally in the cosmetic and pharmaceuticalindustries (Chandonnet et al, 1985). Morerecently, microfluidization was suggestedas an alternative method for the productionof milk fat microcapsules (Vuillemard,1991), alcoholized creams (Paquin andGiasson, 1989) and for milk homogeniza-tion (Pouliot et al, 1991). If the result of theprocess is similar to that of homogeniza-tion (reduction in the size of the fat glob-ules), the means used to achieve such aresult are different. As is the case for con-ventional homogenization processes, theIiquid is forced under high pressure into achamber. However, in the case of the mi-crofluidization process, the Iiquid is then di-vided into 2 microstreams that are project-ed against one another at high speed andat an angle of 1800 (Cook and Lagace,1985).

Other than a considerable change in thedispersion state of the fat, the emulsificationof dietary liquids also results in a considera-ble rearrangement of the oil-water interface

Microfluidization of dairy emulsions. 1 513

(Walstra and Oortwijn, 1982) where mainly2 classes of molecules can be adsorbed:amphiphilic macromolecules (mainly pro-teins) and low molecular weight emulsifiers(Iecithins, monoglycerides, Tweens, Spans,etc) (Dickinson et al, 1990). Proteins andlow molecular weight emulsifiers affect theproduction and stabilization of emulsions.Proteins play 2 major roles: on the onehand they lower surface tension betweenthe interfaces that are fonned du ring theemulsification process, and on the otherhand, they form a macromolecular layersurrounding the dispersed particles whichstructurally stabilizes the emulsion by reduc-ing the rate of recoalescence (Fisher andParker, 1988). Low molecular weight emul-sifiers affect the production and stabilizationof emulsions in several ways: during emulsi-fication, by reducing interfacial tension morerapidly than would the proteins alone, theyallow the production of smaller-sized dis-persed particles (Darling and Birkett, 1987).Furthennore, while proteins generally stabi-lize oil-in-water emulsions, low molecularweight emulsifiers tend to destabilize them(Dickinson et al, 1990) by removing proteinfrom the surface of the dispersed particles.This ability of low molecular weight emulsifi-ers to dislodge macromolecules is due totheir greater energy of adsorption comparedto individual segments of macromolecules.Finally, if the distribution of these 2 classesof molecules between the surfaces of thedispersed particles and the bulk phase is di-rectly affected by the competitive adsorptionbetween the macromolecules and the emul-sifiers at the oil-water interface, it is also aresult of the nature of macromolecule-emulsifier interactions at the interface and inthe bulk phase (Dickinson, 1986).

Opinions are divided as to the usefulnessof using experimental model systems (sys-tems containing a single protein sourceand/or low molecular weight emulsifier)rather than real food systems (Darling andBirkett, 1987; Dickinson et al, 1990). Even if

the model systems allow relatively preciseinfonnation to be obtained, they can rarelybe extrapolated to real food systems. An al-temative wou Id be to use systems that canbe called "quasi-foods", that is systems thatcontain a known mixture of proteins and lowmolecular weight surfactants and whoseconcentrations are controlled.

The aim of the study presented herewas to examine the influence of certainprocess (pressure and temperature) andcomposition (butter oil, protein and surfac-tant concentrations) variables on the sizedistribution of fat globules in a model dairyemulsion (oil-in-water) using a given me-chanical emulsification technology, le mi-crofluidization.

MATERIALS AND METHODS

Experimental design

Description

A central composite rotatable design made up ofk = 5 factors with 3 levels and 1 repetition wasdeveloped. This type of design, the most widelyused for fitting experimental results to a second-order model (Piggot, 1986; Box and Drapper,1987), is called rotatable because the varianceof the estimated response fat point x is unique-Iy a function of the distance separating this pointfrom the centre point and not of the direction.These designs are composed of 2k factorialpoints (usually coded ± 1) with 2k star pointscoded [(± a, 0, 0, ... ,0), (0, ± a, 0, ... ,0), ... , (0,0, 0, .. ,' ± a)] and no centre points coded (0, 0,... , 0). The value of a, on which the rotability ofthe design depends, is such that a = {FJ1/4where F is the number of factorial points of thedesign: in the present case F = 2k• The complexprinciples for the development of these designshave been presented by various authors includ-ing Montgomery (1976), Gacula and Singh(1984), and Box and Drapper (1987).

The design used was therefore composed of25 factorial points, with 10 star points (with a =

514

2.38) and 10 centre points which results in a to-tal of 52 experimental units or emulsions.

The 5 factors that were varied in the prepara-tion of the emulsions are the following: 2 pro-cess factors-emulsification pressure (7.8, 20.7,41.4, 62.1 and 76.3 MPa) and temperature (35,55, 70, 85 and 100 OC),and 3 formulation fac-tors-protein concentration (0.5, 0.9, 1.5, 2.5 and3.9 wt%), butter oil concentration (5.2, 8, 10, 12and 14.8 wt%) and monoglyceride concentra-tion (0.08, 0.2, 0.4, 0.6 and 0.88 wt%).

Statistical analysis

An analysis of variance was first conducted onthe values of the volume-weighted average di-ameters of the emulsified fat globules (dv) andon the relative width (cv) of the particie sizedistribution. This analysis allowed the split ofthe total variation of the measured parameters(av and cv) according to the principal effects ofthe treatments, the interactions and the error.The sum of the squares of the differences ofthe principal effects was split in unitary degreeof freedom using an orthogonal comparisonmethod (Box and Drapper, 1987). This calcula-tion allowed an individual estimation of effectswithout mutual interference. Two tests werethen carried out: a Fisher test allowed a diffe-rentiation between the treatments and the in-teractions which had a significant effect on agiven parameter, and an adjustment test (Net-er et al, 1985) confirmed that the second-ordermodel allowed an adequate description (to ±1%) of the variances in the experimental re-sults.

The effects of the principal treatments and/orinteractionswhich were not significant at the level95% have been included in the error term. Thispooling, which has been justified by Little (1981),increases the degree of significance of the vari-ance due to an increase in the number of degreesof freedom. Although a lower probabilitythresholdcould be used in the selection of the effects to beconsidered in the determination of the regressionequation, a probability threshold of at least 95%was used in order to simplify the expression anduse of the regressionequations.

Finally, following this second analysis of vari-ance, each of the 2 measured parameters (avand cv) was analyzed as a function of the princi-pal effects and interactions of process and com-position variables. The nullity of the regression

a Robin etai

coefficients was tested using Student's test (Net-er etai, 1985).

This method for interpreting results is sup-ported and recommended by Little (1981) andBox and Drapper (1987). According to these au-thors, ail the treatment levels in the experimen-tal range are significantly different in their effectsafter a significant tendency has been estab-lished. The best estimates of the effects of thetreatments are those values calculated using theregression equation.

Statistical procedure and graphie

The analysis of variance and regression calcula-tions were carried out using the General LinearModel procedure (Proc GLM and Proc REG) ofthe SAS (1990).

The 3-dimensional picture of the equationswas carried out using Mathematica (Wolfram,1991).

Emulsion preparation

Ingredients

Sodium caseinates (89.3% protein) were obtainedfrom ICN Nutritional Biochemicals, Canada Ltd(Dorval, Que, Canada). Butter oil (99.4% anhy-drous) was purchased from a local dairy coopera-tive (Agropur, Granby, Quebec, Canada) and wasstored at 4 oC. Distilled monoglycerides (EXCEL,T-95) with a hydrophile-lipophilebalance (HLB) of4.5 (Shinodaand Kunieda, 1983) were purchasedfrom Atkemix (Bratford,Ontario, Canada). Sodiumazide was purchased from Fisher Scientific (Que-bec, Que, Canada).

Emulsion production

Fifty-two emulsions (600 g) containing 0.5, 0.9,1.5, 2.5 or 3.9 wt% (based on protein content) so-dium caseinates, 5.2, 8, 10, 12 or 14.8 wt% but-ter oil, and 0.08, 0.2, 0.4, 0.6 or 0.88 wt% mono-glyceride were prepared in the following manner:various sodium caseinate solutions, hydrated andsolubilized in deionized water (k = 1.1 IlS.cm-1)for 90 min, and butter oil, preheated to 50 "C,were mixed using a magnetic stirrer. The solu-tions were then brought to the appropriate emul-

Microfluidization of dairy emulsions. 1

Table 1. Process variables (protein, butter oiland monoglyceride concentrations, emulsifica-tion pressure and temperature) and responsesof the central composite rotatable design interms of volume-weighted average diameterand relative width of the emulsified fat globulesdistribution.Variables de procédés (concentrations en pro-téine, huile de beurre et monoglycéride, pres-sion et température d'émulsification) et ré-ponses en termes du diamètre pondéré envolume et du coefficient de variation de la distri-bution de la taille des globules de gras émulsi-fiés.

Trials [Prot] [BO] [MGS] P T (dv) cv(wt%) (wt%) (wt"lo) (MPa) (oC) (nm) (%)

Factorial points123456789

101112131415161718192021

0.50.50.50.50.50.50.50.50.50.50.50.50.50.50.50.52.52.52.52.52.5

88888888

121212121212121288888

0.2 20.70.2 20.70.2 62.10.2 62.10.6 20.70.6 20.70.6 62.10.6 62.10.2 20.70.2 20.70.2 62.10.2 62.10.6 20.70.6 20.70.6 62.10.6 62.10.2 20.70.2 20.70.2 62.10.2 62.10.6 20.7

55 584 5185 594 4955 480 5685 466 5555 427 3985 382 3855 317 4885 326 3855 806 6985 793 6955 808 5285 777 5355 460 6785 491 6455 407 4785 419 4355 509 4185 482 4755 439 4285 358 5055 423 41

sification temperature (35, 55, 70, 85 or 100 OC)and the emulsifiers were added. The dispersions,maintained at the emulsification tempe rature (±1.5 "C), were then mixed for 2 min using a stirrer(Braun CDN Ud, Model MR7, Mississauga, Ont,Canada). The oil-in-water dispersions were thenimmediately emulsified using a Microfluidizer (M-110™, Microfluidic Corporation, Boston, MA,USA) at its maximum power setting.

2223242526272829303132

2.52.52.52.52.52.52.52.52.52.52.5

888

1212121212121212

0.6 20.70.6 62.10.6 62.10.2 20.70.2 20.70.2 62.10.2 62.10.6 20.70.6 20.70.6 62.10.6 62.1

515

85 407 4755 361 4685 300 5155 574 5685 593 5955 538 5285 539 4055 609 4585 653 4655 563 2885 509 32

70 420 4670 427 4570 451 4570 448 4570 415 46·70 441 4670 465 4370 433 4570 471 4170 444 46

442 50403 29291 31633 52549 53324 30942 39461 66655 55513 58

0.4 41.40.4 41.40.4 41.40.4 41.40.4 41.40.4 41.40.4 41.40.4 41.40.4 41.40.4 41.4

10 0.4 41.1 7010 0.4 41.4 705.2 0.4 41.4 70

14.8 0.4 41.4 708 0.08 41.4 708 0.88 41.4 708 0.4 7.8 708 0.4 76.31 708 0.4 41.1 352

8 0.4 41.4 1003

For the design rotability to be pertect (a = 2.378), the va-lues 01some Independent parameters should 00: 1 emul-sion 50, P = 90.6 MPa (in lact, the maximum pressuregiven by the air lead system was 76.5 MPa); 2 Emulsion51, T=34.3°C; 3 Emulsion 52, T=105.7°C.

The microfluidization procedure was per-formed in 2 stages: the first processing was at7.8,20.7,41.4,62.1 or 76.3 MPa and the sec-ond at 4.8 MPa in order to eliminate aggregatesformed during the first processing procedure(Ogden, 1973; Walstra, 1975). The compositionof the emulsions is given in table 1.

Sodium azide, an antibacterial agent, wasadded to each emulsion (0.2% vol/wt), and 2 pH

Centre points33 1.5 1034 1.5 1035 1.5 1036 1.5 1037 1.5 1038 1.5 1039 1.5 1040 1.5 1041 1.5 1042 1.5 10

Star points43 0.944 3.945 1.546 1.547 1.548 1.549 1.550 1.551 1.552 1.5

516

measurements (pH-meter, Corning 140, elec-trode 476530, Canlab, Montreal, Que, Canada)were carried out.

Size distributionof the emulsified fat globules

Sampling

A sam pie of each emulsion was taken andstored at 4 oC for 24 h in a 20 ml bottle com-pletely filled. Before analysis, the samples wereheated (T = 20 oC) and manually stirred. Thisprocedure, proposed by Walstra (1975), wasused in order to prevent a change in the sizedistribution of the fat globules due to churning.

Dissociation of protein structures

The samples were mixed with a buffer (urea,EDTA and ~-mercaptoethanol, pH 7) designedby Dalgleish et al (1987) to dissociate proteinstructures. Despite the presence of sodiumcaseinates in the emulsions, the composition ofthe dissociating buffer was not modified. Meas-urements were carried out immediately after vig-orous manual stirring.

Photon correlation spectrophotometry

The size distributions of the fat globules weredetermined by photon correlation spectrometry(PCS) using the method proposed by Robin andPaquin (1991). The measurements were takenusing the Nicomp multibit (7 bit) 64-channelphoton correlation system (Pacific Scientific,Hiac/Royce Instruments Division, Model 370,Manlo Park, CA, USA); the correlation functionswere measured on the Iight that was diffused atan angle of 90° by the particles in suspension.The correlation functions were analyzed by thecumulants method (Koppel, 1972). The first cu-mulant yielded an' average diffusion coefficient(0), and the second yielded a mean squareddeviation of the average diffusion coefficient(s2). The volume-weighted average particle di-ameters (d--:) were calculated using the Stokes-Einstein law, and polydispersity was expressedby the relative standard deviation or coefficientof variation (cv) of the volume-weighted distribu-tion calculated as the distribution width divided

o Robin etaI

by the average diameter. The accuracy (= 1%),the reproducibility (= 1%) as weil as the length ofthe analysis (2.5 x 106 photo-impulses or '" 20min) were determined on latex sphere solutions(Robin and Paquin, 1991).

Three measurements of the size distributionof the fat globules were carried out on eachsample. The values of the average diametersand relative widths of the size distribution of thefat globules are shown in table 1.)

Average size distributions

A logarithmic transformation was applied to theemulsified fat globule volume-weighted averagediameter (d:) and to the homogenization pres-sure. Indeed, using the theory of isotropic turbu-lence developed by Kolmogorov (Walstra, 1969,1983) showed that the maximum size of a fatglobule undergoing the homogenization processis a function of the energy density E (energy dis-sipated per unit of volume and time) or of thepressure P (E ocfX3I2 fTll2), of the interfacial ten-sion 'Y, and of the mass density p acc.ording to:

1 3d oc (e2 y3 fT1)-oc P-_ [1]

max 5 5or

3log10 (dmax) a. - - IOg10 (P) [2]

5

Goulden and Phipps (1964), Walstra (1975) andothers have experimentally confirmed these re-sults.

RESULTS

Accuracy and reproducibilityof the PCS technique

The 10 centre points serve as an indicationof the reproducibility of the PCS techniquefor determining the parameters of the sizedistribution of the fat globules. These centrepoints correspond to values obtained fromthe same emulsion. Using these results, thereproducibility of the evaluation of the dv andcv parameters was approximately 4.2, and


3.2% respectively. However, even thoughthe reproducibility of the results seem salis-fying, that is, comparable with those ob-tained by other methods, their accuracymust be interpreted with some caution. Eventhough the validity of the PCS technique, interms of precision and reproducibility, hasbeen demonstrated by scattering fromknown mixtures of relatively monodispersedpolystyrene latex beads (Bargeron, 1974;Brown and Pusey, 1975), results becomeless reliable wh en distributions are largerand polydispersed (Weiner and Tschamuter,1987). Moreover, if the second cumulant isgenerally dominated by polydispersity ratherthan by statistical noise in the autocorrela-tion function, in large distributions the analy-sis of the cumulant can lead to an underesti-mation of the variance (Chu and Dinapoli,1983). Higher order terms should be takeninto account, but this remains a difficult prob-lem (Nicoli, personal communication) ..

Influence of process variableson the average diameter of fat globulesand on the distribution widthof the diameters

An analysis of variance was carried out onthe values given in table 1 using orthogonalcomparisons between the principal treat-ments and between the interactions (Boxand Drapper, 1987). An F test was carriedout to differentiate the treatments and theinteractions which had a significant effect onthe volume-weig!}!ed average diameter ofthe fat globules (dv) and on the size distribu-tion of the relative width (cv). An adjustmenttest of the experimental results for a sec-ond-order model was also carried out.

Average diameterof the emulsified fat globules

Table Il shoy!s the analysis of variance ofthe log10 (dv) as a function of process

517

Table Il. Analysis of variance of the logarithm ofthe volume-weighted average fat globule diame-ter versus process and composition variables.Analyse de variance du logarithme du diamètremoyen pondéré en volume des globules degras, en fonction des variables de procédé et decomposition.

df1 552 r24

ModelLinear effects

[Pro~[BO][MGS]10glO (P)T

Quadratic effects[Pro~2 1[B0]2 1[MGS]2 110glO 2 (P) 1J2 1[Pro~ x [BD) 1[Pro~ x [MGS) 1[Pro~ x log10(P) 1[Pro~ x T 1[BD) x [MGS] 1[BD) x 10glO (P) 1[BO] x T 1[MGS] X log10(P) 1[MGS] x T 1Tx log10(P) 1

ErrorLack of fitPure error

Total

ModelErrorTotal

20 0.624

o.ooa0.2170.1470.1030.006

<10-,30.002<10--30.03a0.025<10-,30.064< 10--30.001<10--30.0060.0020.001<10-,30.002

31 0.02722 0.0029 0.003

51 0.651

36.0a 0.959.

9.6C

251Aa170Aa119.1a

6.6d

< 0.11.9

< 0.143.5a

2a.9a0.5

73.aa0041.3

< 0.17.02.6104

< 0.12.0

3AC

a 0.610 so.s- 0.93743 0.04151 0.651

1 dl = degree01 lreedom;2 S5 = sum01 squares;3 F =Fisherratio;4 t2 = coefficient01 determination.Signili-caneelevels• p:s; 0.0001;b p:s; 0.001;s p « 0.01;d p:s;0.05.

518

variables. According to the adjustmenttest, for p s 0.01 (F < 4.77), the second-order model allowed an adequate descrip-tion of the variance of log10 (dv)' Further-more, the linear effects of ail the principalfactors were significant (from 95% to99.999%). Butter oil ([BD]) and monogly-ceride ([MGS]) concentrations had a majoreffect on the variance of the fat globule di-ameter: they explained 33.4% and 22.6%,respectively, of the total variance. Theemulsification pressure (P) and tempera-ture (T) represented 21.7% and 4.8% re-spectively, of the variance (the sum of thelinear and quadratic effects: IgglO (P) +10916 (P) or T + T2) of log10 (dv)). The ef-fect of the protein concentration was signif-icant (p :;;; 0.001) but represented only1.3% of the variance of the IOg10 (dv)' Thelast significant effect given by the analysisof variance was the [Pro~ x [MGS] interac-tion which represented 9.8% of the vari-ance of the log10 (dv)' Other effects werenot significant (p > 0.05). The error termrepresented "" 4% of the variance of log10(dv)·

A regression equation calculated fromthe significant effects (p:;;; 0.05 or more) ofthe analysis of variance of log10 (dv) isshown in table III. This second-order mod-el was highly significant (p s 0.0001) andexplains 93.7% of the total variance oflog10 (dv). Ali of the regression coefficientswere highly significant (p s 0.0001). lt is,however, worth noting that the existenceof a minimum average diameter as a func-tion of pressure (regression equation) wasnot significant, and resulted from an arte-fact praduced by one of the extreme point(star point No 48). A separate regressionanalysis revealed that the increase in sizebeyond P == 50 MPa (the emulsificationpressure from the regression model forwhich the derivative equals 0) was not sig-nificant.

Figure 1 is the response surface ob-tained trom a second-order model relating

o Robin etaI

Table III. Analysis of regression and second-arder model of the logarithm of the volume-weighted average fat globule diameter versus si-gnificant variables and interactions.Analyse de régression et modèle de secondordre du logarithme du diamètre moyen pondéréen volume des globules de gras émulsifiés, enfonction des variables et des interactions signifi-catives.

Estimate 1 STD2

Parameter:Intercept 14.293

[Pro~ -0.102[BO] 0.035[MGS] -0.63310910 (P) -4.755T -0.016109102 (P) 0.506J2 1.110-4[Pro~ x [MGS] 0.223

1.61.210-2210-3510-20.7310-30.1210-52.710-2

9.1a

-B.5a15.2a

-13.4a-6.9a-5.5a

6.6a5.2aa.2a

10910 (dv) = 14.293 - 0.102 [Pro~ + 0.035 [BO]- 0.633 [MGS]- 4.75510910 (P)- 0.016 T + 0.506 109102(P)+ 10-4 J2 + 0.223 [Pro~ x MGS

Validity intervals:7 soo s p:s; 76310 kPa 0.5 :s;[Pro~ s 3.9 wt%35:S;rs 100 -c 5.2:s; [BO]:s;14.a wt%

o.os s [MGS] s o.aa wt%

1 Estimate = parameter estimated under the regressionmodel (table Il). 2 STO = standard error 01 the estimate;3 1= value 01 Student's r-test. Signilicance level: a: p :s;0.0001.

the values of dv as a function of the butteroil concentration ([BO], wt%) and of theemulsification pressure (P, MPa) for afixed temperature (50 "C) and for fixed pro-tein (1.5 wt%) and monoglyceride (0.4wt%) concentrations. As has been report-ed elsewhere, the average diameter of thefat globules decreases and reaches a pla-teau as emulsification pressure increases,


1000

i1.,5

800

519

14

Fig 1. Response surface of the volume-weighted average fat globule diameter as a function of emulsi-fication pressure, P (MPa), and butter oil concentration, [BD] (wt%), for a fixed tempe rature (T = 50 "C),a fixed protein concentration ([Pro~ = 1.5 wt%), and a fixed monoglyceride concentration ([MGS] = 0.4wt%).Surface de réponse du diamètre moyen pondéré en volume des globules de gras en fonction de lapression d'émulsification, P (MPa), et de la concentration en huile de beurre, [BO] (%), pour une tem-pérature donnée T = 50 OC), et des concentrations fixées en protéines ([prot] = 1.5% et en monogly-cérides ([MGS] = 0.4%).

600

but increases as the fat content increases(Walstra, 1975; Phipps, 1985).

Figure 2 shows the complex influencesof emulsifier type and concentration on fatglobule size, for emulsions produced underconditions of constant pressure (50.0MPa), temperature (50 oC), and butter oilconcentration (10 wt%). Our results are inagreement with previous reports in that, ifthe presence of a single surfactant resultsin a dramatic lowering of the size of the fatglobules, the near absence of surfactants([Pro~min = 0.5 wt%, and [MGSJmin = 0.08

wt%) results in a maximum value for a;(678 nm).

Relative width of the size distribution. of the emulsified fat globules

A preliminary statistical analysis has re-vealed that the variations of cv as a func-tion of various independent variables couldnot be adequately described by a second-order model (r2 = 0.71 after pooling; theadjustment test was not significant) in theoriginal validity interval. A reduction of the

520 o Robin etaI

Ê 600.s~ 500

400

300

Fig 2. Response surface of the volume-weighted average fat globule diameter as a function of pro-tein, [Pro~ (wt%) , and monoglyceride, [MGS] (wt%) , concentrations for a fixed temperature (T =50 OC), a fixed pressure (P = 50 MPa), and butter oil concentration [Ba] = 10 wt%).Surface de réponse du diamètre moyen pondéré en volume des globules de gras en fonction desconcentrations en protéines, {prot] (%), et monoglycérides, [MGS} (%), pour une température (T =50 OC), une pression (P = 50 MPa), et une concentration en huile de beurre ([BO) = 10%) fixées.

interval corresponding to the elimination ofan extreme point (No 49) has consequent-Iy been effected.

Table IV shows the analysis of varianceof cv as a function of the process vari-ables. According to the adjustment test, forp s; 0.01 (F < 4.77), a second-order modelcould adequately describe the variances ofcv in this new validity interval. As men-tioned previously, the linear effects of theprincipal factors, except the temperatureone, were significant (from 95% to99.999%). The protein and monoglycerideconcentrations had an important effect onthe variance of cv: they explained respec-

tively 16.0% and 13.4% of the total vari-ance. The emulsification pressure andtemperature represented respectively16.8% and 2.7% (the sum of the linear andquadratic effects: P + pl and 72) of thevariance of cv. The effect of the butter oilconcentration (p s 0.0001) represented7.7% of the variance of cv. The other high-Iy significant interactions ([Pro~ x [BO] and[Ba] x Pl represented 28.5% of the totalvariance. The error represented approxi-mately 10.2% of the variance of cv.

A regression equation, calculated usinghighly significant effects (p s 0.000 1) fromthe analysis of variance of cv, is shown in


Table IV. Analysis of variance of the standarddeviation of the volume-weighted size distribu-tion of fat globules versus process and composi-tion variables.Analyse de variance de la déviation standard dela distribution des tailles pondérées en volumedes globules de gras en fonction des variablesde procédé et de composition.

df1

Model 20 4196.5 13.1a 0.897

Linear effects[Pro~ 749.1 46.9a

[BD] 361.2 22.6a

[MGS] 626.7 39.2a

P 156.9 9.8a

T 4.6 0.3

Quadratic effects[Pro~2 17.9 1.1[BO]2 39.4 2.5[MGS]2 20.8 1.3~ 630.9 39.5a

T2 123.9 7.8a

[Pro~ x [BD] 359.4 22.5a

[Pro~ x [MGS] 23.5 1.5[Pro~ x P 0.2 < 0.1[Pro~ x T 75.0 4.7[BO] x [MGS] 19.7 1.2[BO] x P 974.8 61.0a

[BO] x T 2.1 < 0.1[MGS] x P 0.5 < 0.1[MGS] x T 7.2 0.4TxP 3.0 0.2

Error 30 479.3Lack of fit 21 3592.7 4.1cPure error 9 358.5

Total 50 4675.8

Model 9 3823.0 20.4a 0.818Error 41 852.9Total 50 4675.9

See table Il for statistical abbreviations.

table V. The second-order model, whichwas highly significant (p s 0.0001), wasable to explain 81.8% of the total variance

Table V. Analysis of regression and second-order model of the variation coefficient of the vo-lume-weighted size distribution of fat globulesversus significant variables and interactions.Analyse de régression et modèle de secondordre du coefficient de variation de la distribu-tion des tailles pondérées en volume des glo-bules de gras, en fonction des variables et desinteractions significatives.

Estimate 1 STO 2 t 3

Parameter:Intercept 25.127 18.2 1.4[Prot] 12.532 4.1 3.1[BD] 9.477 1.1 8.9a

[MGS] -19.278 3.6 -5.48

P +2.81Q-4 31Q-4 0.9T -1.133 0.4 -2.6~ 10-7 1{)~ 4.18

T2 0.0083 0.03 2.7[Pro~ x [BO] -1.676 0.4 -4.28

[BO]xP -1.310-4 210-5 -6.8a

cv= 25.127 + 12.532 [Pro~ + 9.477 [BD]-19.278 [MGS]':' 2.8 1Q-4 P-1.133 T+ 10-7 ~ + 0.008 T2 -1.676 [Pro~ x [BD]-1.310-4 [BO] x P

Validity intervals20400:5 P:5 76 300 kPa 0.5 :5 [Pro~ :53.9 wt%35:5 T:5 100 -c 5.2:5 [BO] :514.8 wt%

0.08:5 [MGS] :50.88 wt%

See table III for statlsûcal abbrevialions.

cv. The value of the regression coefficientsof the P term was not significantly differentfrom 0 tp » 0.05).

Figure 3, obtained from a second-ordermodel, corresponds to the response sur-face of the standard deviation of the vol-ume-weighted size distribution of fat glob-ules as a function of emulsificationpressure, P (MPa), and butter oil concen-tration [BO] (wt%) , for a fixed temperature

522

70

>U

o Robin etaI

P (!v[Pà)

Fig 3. Response surface of the coefficient of variation standard of the volume-weighted size distribu-tion of fat globules as a function of emulsification pressure, P (MPa), and butter oil concentration,[BD) (wt%), for a fixed temperature (T = 50 OC), a fixed protein concentration ([Pro~ = 1.5 wt%), and afixed monoglyceride concentration ([MGS] = 0.4 wt%).Surface de réponse du coefficient de variation de la distribution des tailles pondérées en volume desglobules de gras en fonction de la pression d'émulsification, P (MPa), et de la concentration en huilede beurre, [BD] (%), pour une température donnée (T = 50 OC), et des concentrations fixées en pro-téines ([Prot] = 1.5%) et en monoglycérides ([MGS] = 0.4%).

(T = 50 OC), a fixed protein concentra-tion ([Pro~ = 1.5 wt%), and a fixed mono-glyceride concentration ([MGS] = 0.4wt%). An increase in the emulsificationpressure and/or butter oil concentration re-sulted in an increase in the value of cv.This figure also shows the minimal valuesfor cv that are a function of both P and[BO] (table IV).

Figure 4 shows the response surfaceobtained from a second-order model of thevolume-weighted size distribution of fatglobules as a function of protein and

monoglyceride concentration, for a fixedtemperature (50 OC), pressure (50 MPa)and butter oil concentration (10 wt%). Thisfigure indicates the cooperative effect ofsurfactant content on the reduction in cv ofthis volume-weighted distribution.

pH of the emulsions

The average pH of the 52 emulsions was7.06 with a standard deviation of ± 0.04. Aliemulsions therefore had similar acidlbaseproperties.


~u

[Prot.) (%)

Fig 4. Response surface of the coefficient of variation of the volume-weighted size distribution of fatglobules as a function of protein, [Pro~ (wt%) , and monglyceride, [MGS] (wt%) , concentrations for afixed tempe rature (T = 50 oC), a fixed pressure (P = 50 MPa), and butter oil concentration ([BD) = 10wt%).Surface de réponse du coefficient de variation de la distribution des tailles pondérées en volume desglobules de gras en fonction des concentration en protéines, [prot] (%), et monoglycérides, [MGSJ(%), pour une température fT = 50 OC), une pression (P = 50 MPa), et une concentration en huile debeurre ([BOJ = 10%) fixées.

DISCUSSION

Emulsification is a dynamic process wherethe disruption and recombination or coa-lescence of the fat globules take place si-multaneously, each with its own rate con-stant or time scale (Walstra, 1983;Tornberg et al, 1990). Consequently, notonly the final droplet size distribution butalso other emulsion properties such asrheological behavior will be determined bythe conditions leading to an equilibrium be-tween breakdown and coalescence. The

probability of newly formed droplets coa-lescing depends on the time available forthe interfaces to be covered by the surfac-tants. The time available depends on emul-sifying conditions (eg emulsifying machine,power density), which are influenced bythe fat content, the ratio of surfactant to fat,and the nature of the surfactants. In orderto be active (prevent coalescence), surfac-tants must not only be transported to theinterfaces, they must also be able to ad-sorbo Adsorption depends on the numberof surfactant collisions with fat particles,

524

and on the probability of the surfactantsadsorbing after colliding with the interface.Adsorption is likely influenced by molecu-lar properties such as hydrophobicity, f1exi-bility and charge density (Kato, 1991; Lo-rient et al, 1991). However, the adsorptionprocess occurs on a time scale of lessthan a millisecond during the homogeniza-tion process (Walstra, 1983), so it is veryunlikely that an adsorption equilibrium isobtained.

For a given process, the emulsificationefficiency is the result of an optimal combi-nation of process variables and of thecomposition of the Iiquid to be emulsified.Six major variables are generally consid-ered: emulsification pressure (P) and tem-perature (T), Iiquid flow rate (0), the frac-tion (cP) of the disperse phase and thesurfactant concentration(s) (Walstra, 1983;Phipps, 1985). In the present study, the ef-fect of flow rate on the size distribution ofthe fat globules was not considered. It has,however, been shown for oil-in-wateremulsions (Phipps, 1975, 1982, 1983;Walstra, 1975) and for conventional ho-mogenizers (Manton-Gaulin and/or Ran-nie), that there is little or no influence ofthe flow rate of the liquid to be emulsifiedwhen cP s 30%.

Influence of process variableson the size distribution of fat globules

Effect of emulsification pressure

It was empirically shown (Goulden andPhipps, 1964; among others) and then vin-dicated by Walstra (1969, 1975) that clas-sic homogenization processes (the sys-tems of Manton-Gaulin or Rannie), atmoderate emulsification pressures (0.25MPa s P s 40.5 MPa), using milk, creamwith a fat concentration of cP ~ 12% or anyother dilute emulsion, could be quantifiedby a relation such as d oc pm where the ex-

o Robin etai

ponent m has a value of -3/5 ([1]). The ex-ponent_m in the descending slope of thecurve dv = f (pm) (fig 1) is of the order-0.53 ± 0.04. These results, therefore,seem to corroborate the theories linked tothe breakdown of fat globules during turbu-lent flow conditions. However, there is nogeneral agreement on the functioning ofconventional homogenizers at high pres-sure. According to Walstra (1969, 1983)and Davies (1985) the breakdown of thefat globules can be described by the turbu-lence theory put forward by Kolmogorov.According to Phipps (1975, 1985), thebreakdown of the globules at the entranceto the homogenization valve can be de-scribed by the Taylor analysis of shearingforces. The design of the microfluidizationchamber (microchannels, mixing zone, etc)differs fundamentally from an homogeniza-tion chamber (basically a valve and aseat). Even if the forces that lead to thebreakdown of the fat globules are thesame, their respective influence on thebreakdown is apparently not the same._ Although the convex nature of the curvedv = f (P~Bo] (fig 1) is the result of an ex-perimental artefact and of its statisticalconsequence (the minimum is very shal-low), sorne authors (Becher, 1967; Phipps,1975; Tornberg, 1980) have reported thatan increase in homogenization intensity(E), and/or fat concentration above optimalyields (P ~ 40 MPa or E ~ 60 W and/or cP 2:12%) could result in an increase in dropletsize. This increase, called overprocessingby Tornberg (1980), should indicate thatcoalescence is the dominant factor govern-ing the final droplet size of the emulsion.However, the validity of the methods thathave been used for producing emulsionsand determining average size is questiona-ble. In the present study, it is reasonable tothink that an increase in collision frequencywith increasing pressure would at mostlead to no further decrease in globule sizeas the minimum is very shallow. An in-


crease in emulsification intensity above op-timal conditions, which causes a relativelysm ail change in average droplet size, re-sults in unnecessary energy consumption.The optimal condition corresponded to P ==50 MPa. The reduction in fat globule sizehave various consequences. Emulsionswhich contain small droplets have a ten-dency to be more stable with respect tocreaming (Stokes' law), and to coalescemore than those containing large droplets(MacRitchie, 1976). According to the Mac-Ritchie approach, which correlates film sta-bility to an energy barrier, the corn pres-sional free energy barrier can be, undersorne conditions, proportional to 1/cP. How-ever, in practice the effects are not asgreat as predicted by these equations(Walstra and Oortwijn, 1975; Dickinsonand Stainsby, 1982).

If the average particle size can providesorne information about the behavior ofemulsions, most attention should, how-ever, be paid to the top end of the size dis-tribution as most types of instability areusually first manifested by the behavior ofthe largest droplets. Although the distribu-tion width of fat globules is an extremelyimportant parameter when characterizingan emulsion, it has not been studied ingreat depth. Walstra (1975) demonstratedthat the relative width of fat globules in ho-mogenized milk sharply increases with thehomogenization pressure (0 < P < 5 MPa),reaches a maximum (P == 5 MPa) beforeremaining fairly constant and attaining alimit (P == 30 MPa). Thèse results werepartly confirmed by Tornberg (1980) study-ing the behavior of various protein stabi-lized emulsions made by a sonifier. In thepresent study, valid for 20.4 MPa s P :,;76.3 MPa, the results revealed a behaviorthat differed as a function of butter oil con-centration ([BO], wt%). At quite high [BO],in conformity with previously mentioned re-sults, the relative width of the fat globuledistribution diminished as a function of

emulsification pressure. However, at lowerconcentrations, the behavior was reversed.Although these trends should be interpret-ed with sorne caution (Iack of accuracy inthe determination of the standard devia-tion, low r2 value), it appears that trendsnear central points are similar to those re-ported elsewhere. At the endpoints of theexperimental design, trends are muchmore difficult to establish, particularly whenthe decrease in cv as a function of [BO] athigh P should arise from the standgrd devi-ation(s) being less affected th an dv by theemulsification intensity.

Effect of emulsification temperature

Since Gaulin, it has been recognized thatthe efficiency of the emulsification pro-cess is very low in the presence of solidfat (Kessler, 1981). Consequently, theproduction of oil-in-water emulsions isnormally carried out at tempe raturesabove the final melting temperature of thefat (7j == 40 "C, Jenness and Walstra,1984). Several authors, studying the ho-mogenization of milk, have shown that anincrease of 10°C in temperature between40-70 oC decreases the average diame-ter of the fat globules by 6-8% (Walstra,1975) to 10-15% (Sweetsur and Muir,1983). This effect weakens or disappearsabove 80 oC. As expected, an increase intemperature between 35-82 oC resultedin a decrease in the average diameter of= 8%/10 oC. Although equation [1] doesnot predict that the viscosity of eitherphase can have an effect on the dropletsize resulting from the emulsification pro-cess in turbulent f1ow, a decrease of vis-cosity will understandably affect the rateof passage through the emulsifier and theease of disruption of the globules. Walstra(1974, 1983) suggested that an increasein the viscosity of the dispersed phase(T/o) should correspond to a deformationtime of a droplet larger than the character-

526

istic time of an eddy. As the smallest ed-dies are presumably the most effective(they have thehighest kinetic energy), anincrease in 7Jo should lead to a largerspread in flow conditions and conse-quently in droplet size distribution (cv), aswas indeed the case. Moreover, an in-crease in temperature changes ail thecomposition variables that determine ad-sorption; fat changes to oil and the mac-romolecular penetration into the oil be-cornes possible, hydrophobic interactionsat the interface probably become more in-tense (at least up to 60-65 OC), and themolecular structure of water weakens,which affects its quality as a solvent forsurfactants and for hydrophobic interac-tions. Above 82 oC, the average particlediameter increased slightly. Although thisresult has not been explained, it shouldbe noted that the influence of the temper-ature on the average diameter of the fatglobules is relatively low: temperatureonly explains 4.8°-i (table Il) of the totalvariance of 10glO(dv) and between 82 and100 "C it represents < 1% which is muchlower than the experimental error('" 6.3%). ~oreover, the statistical in-crease in dv is a consequence of themeasurements obtained from an emul-sion that was not repeated (No", 48) andcorresponds to an extreme point (starpoint) in the experimental design. cv wasinfluenced in a similar manner by the tem-perature (table V).

Effect of composition variableson the size distribution of fat globules

Effect of butter oil concentration

The concentration of fat affects the ho-mogenization efficiency of dairy productswhen it is > 10% (Walstra, 1975; Phipps,1985). The microfluidization of the modelemulsion demonstrated that an increase in

o Robin etaI

the butter oil concentration (5.2 s [BO] s 14wt%) resulted in a very significant increasein_the average diameter of the fat globules(dv) (table Il, fig 1) and in the size distribu-tion (cvat low pressure) (table III, fig 3). Anincrease in the fat concentration resulted ina relative decrease in the concentration ofsurfactant available to cover the new interfa-cial surface that was formed during theemulsification process. The fat content andthe ratio of surfactants to fat affected the ex-tent of coalescence of the newly formedglobules by goveming the probability thatthe latter collide before they recover by asurfactant layer. Also, with higher oil con-centrations, distances between fat globulesdecrease and the probability of bridgingcould increase.This results in a general ten-dency to increase "particle" diameter.

Effect of surfactant concentration

The addition of surfactants, which resultsin a decrease in interfacial tension (y), re-duces the interfacial free energy (~G) ofthe system,

[3]

where ilS is the change in the interfacialareaand thereby the Laplace pressure which isbeneficial in reducing both the energy re-quirement to form emulsions and the drop-let size that can be obtained. However, thisonly holds if the ratio (surfactantlinterfacialsurface) is large enough to cover the newinterfacial surface formed during the emul-sification process and leading to minimizethe coalescence or polymeric bridging(Halling, 1981; Tadros and Vincent, 1983).For minimal protein and monoglycerideconcentrations ([Profj = 0.5 wt%, [MGS] =0.08 wt%, with P = 50 MPa, T = 50 oC,[BO] = 10 wt%), the average diameter ofthe fat globules was at a maximum (d'v =678 nm).


Although certain general rules can bededuced from simple thermodynamic prin-ciples, the degree and, more importantly,the consequences of the adsorption ofprotein to the interface can vary consider-ably with the type of protein and withemulsification conditions. Thus, whileOortwijn and Walstra (1979) reported thatwhey protein concentration (0.01-2.0%)had relatively Iittle effect on the size of fatglobules in homogenized emulsions,Pearce and Kinsella (1978) demonstrat-ed, using homogenized emulsions stabi-lized with various proteins, that increasingthe protein concentration (0.5-5%) couIddecrease the average diameter of the fatglobules by a factor of 2.5. Thus, the na-ture of the surfactant (hydrophile-lipophilebalance, molecular weight, molecular f1ex-ibility) as weil as its relative concentration(with respect to other surfactants) directlyinfluence the size of the fat globules. Fig-ure 2 iIIustrates the complex influences ofthe chemical structure and of protein andmonoglyceride concentration on the aver-age diameter of emulsified fat globules;the other variables were kept constant(P = 50 MPa, T = 50 oC, [80] = 10 wt%).During the_emulsification process, smallerdroplets (dv = 259 nm) are produced inthe presence of low concentrations of pro-teins and high concentrations of monogly-cerides ([Pro~ = 0.5 wt% and [MGS] =0.88 wt%, respectively) because monogly-cerides are better able to lower interfacialtension than proteins alone. With highprotein concentrations ~nd low monogly-ceride concentrations, dv= 350 nm. Dick-inson et al (1989) reported similar trendswhen studying the effect of octaoxyethy-lene dodecyl ether (C12Ea) with 0.1 wt%caseinate on the droplet diameter of O/Wemulsion (20 wt% n-tetradecane, pH 7,25 OC). However, care is needed whenanalysing these results as the surfactantsused were different and hydrocarbonshave a higher interfacial tension with wa-

527

ter than do mixtures of triglycerides withwater (Fisher et al, 1985).

Moreover, the increase in the concen-tration of surfactant (for a given [80], Pand 7) results in a decrease in the averageparticle diameter (Dickinson et al, 1989) ifthe monoglycerides to protein ratio is loweror equal to 0.15, eg [Pro~ = 2.85 wt% and[MGS] = 0.46 wt% (fig 2). Although theseconcentrations are independent of the[80], which is rather surprising, it is sug-gested that an increase in the size of thefat globules above these concentrationscan be attributed to,- a competition between fat and monogly-cerides to bind proteins due to the highnumber of monoglyceride molecules perresidue of caseinate and to the formation ofhydrophobic bonds between a polar amineacids and the hydrocarbon groups of themonoglycerides (Dickinson and Woskett,1989).These weak complexes, if they form,should be rather surface-inactive, as the hy-drophobic areas of protein and nonionic sur-factant bind together, and- to protein displacement from the inter-face by monoglycerides. In another studyusing casein and MGS, Paquin et al (1987)found that surface pressure isotherms inthe high-pressure region are essentiallythe same as for the monoglycerides alone,suggesting protein displacement from theinterface. lt is likely, as suggested byDoxastakis and Sherman (1986), thatmono- and diglycerides present in com-mercial glycerol monostearate form com-plexes with caseinate at the oil-water inter-face. It would appear that this (mixed)emulsifier does not behave simply as anonionic, noninteracting surfactant.

Surfactants (proteins and low molecularweight surfactants) seem to have a coop-erative effect on the reduction of cv (fig 4).As was mentioned in the discussion on thedetermination of cv, these results must beinterpreted with caution.

528 o Robin etai

1000 Homogenized milk (A2)

b..!800

~

~Uomogenized cream (B)

1.,,' 600

"'--- 7400 IBO)=6%

200mogenizedmi1k(Al)

0 20 40 60 80 20 40 60 80 100 4 10 13 16 0 1 2 3 4 o 0.2 0.4 0.6 0.8 1.0P(MPa) T ('C) [BOJ (wt%) [ProtJ (wt%) [MGSJ (wt%)

80

~ 70~ 60

50

Homogenizedmilk(Al)

4030

020406080P(MPa)

20 40 60 80 100T ('C)

~LL4 7 10 13 16 0 1 2 3 4 0 0.2 0.4 0.6 0.8 1.0

[BO] (wt%) [Prot] (wt%) [MGSJ (wt%)

Fig 5. Effect of process and composition variables on fat globule size distribution parameters. Foreach graph, values of 4 variables were either fixed or specified. Values of fixed variables were: P = 50MPa, T = 50 oC, [BO] = 10 wt"lo, [Pro~ = 1.5 wt"lo, [MGS] = 0.4 wt"lo.Effets des variables de procédés et de composition sur les paramètres de la distribution des taillesdes globules de gras. Pour chaque figure, les valeurs des 4 variables étaient soit fixées soit spécifi-ciées. Lorsque les variables étaient fixées, leur valeur était: P = 50 MPa, T = 50 "O, [BO] = 10%,[prot] = 1.5%, [MGS] = 0.4%.(A): Walstra. 1975; Aj, P 282; A2, P 289. (B): Phipps, 1985, p 12.

CONCLUSIONS

The process and composition variableswhich were studied affect, in varying pro-portions, the efficiency of the emulsificationprocess called microfluidization. Some ofthese influences are summarized in figure5. The variables that had the greatest influ-ence on the size distribution of the fat glob-ules are the butter oil concentration, theconcentration of low molecular weight sur-factants (monoglycerides), the emulsifica-tion pressure, the protein concentration (so-dium caseinates) and the emulsificationtemperature. However, if the average sizesare fairly reliable, the results on relative dis-tribution width must be taken with caution.

The utilization of an experimental de-sign, applied to a dairy-type emulsion,demonstrated the complexity of the influ-ences and interactions (depending on thearea of study, pressure, temperature orconcentrations) of process and composi-tion variables on the size distribution of thefat globules. If the usefulness of mechanis-tic and thermodynamic principles is not indoubt, it is still impossible to completelypredict ail aspects of the behavior of anemulsion.

Finally, although this work was carriedout on a model system and although it wasdemonstrated that for other conditions (theexistence of interaction terms) the effectswould be different, the results should pro-


vide a guide for optimizing the emulsifica-tion process. '

ACKNOWLEDGMENTS

We are indebted to the first reviewer for his help-fui comments.

REFERENCES

Bargeron CS (1974) Analysis of intensity corre-lation spectra of mixtures of polystyrene latexspheres: a comparison of direct leastsquares fitting with the method of cumulants.J Chem Phys 60,2516-2519

Becher P (1967) Effect of preparation parame-ters on the initial size distribution functions inoil-in-water emulsions. J Col/oid Interface Sci24,91-96

Box GE, Drapper NR (1987) Empirical ModelBuilding and Response Surfaces. John Wileyand Sons, NY

Brown JC, Pusey PN (1975) Photon correlationstudy on the initial size distribution function inoil-in-water emulsions. J Chem Phys 62,1136-1144

Chandonnet S, Korstvedt H, Siciliano AA (1985)Preparation of microemulsions by microfluidi-zation. Soap Cosmet Chem Spec (Feb) 37-38

Cook EJ, Lagace AP (1985) Apparatus for forrn-ing emulsions. US Pat 4 533 254

Chu B, Dinapoli A (1983) Extraction of distribu-tions of decay times in photon correlation ofpolydisperse macromolecular solutions. In:Measurement of Suspended Particles byQuasi-Elastic Light Scattering (Dahneke BE,ed) John Wiley and Sons, NY, 81-105

Dalgleish DG, Pouliot Y, Paquin P (1987) Stud-ies on heat stability of milk. II. Associationand dissociation of particles and the effect ofadded urea. J Dairy Res 54, 39-49

Darling DF, Birkett RJ (1987) Food colloids inpractice. In: Food Emulsions and Foams(Dickinson E, ed) R Soc Chem, London, 1-29

Davies JT (1985) Drop sizes of emulsions relat-ed to turbulent energy dissipation rates.Chem Eng Sei 40, 839-842

529

Dickinson E (1986) Mixed proteinaceous emulsi-fiers: review of competitive protein adsorptionand the relationship to food colloid stabiliza-tion. Food Hydroco/1, 3-23

Dickinson E, Stainsby G (1982) Food Col/oids.Elsevier Appl Sci Publ, London

Dickinson E, Stainsby G (1988) Emulsion stabili-ty. In: Advances in Food Emulsions andFoams (Dickinson E, Stainsby G, eds) Elsevi-er Appl Sci Publ, London, 1-44

Dickinson E, Woskett CM (1989) Competitiveadsorption between proteins and small-molecule surfactants in food emulsions. In:Food Col/oids (Bee RD, Richmond P, Min-gins J, eds) R Soc Chem, London, 74-96

Dickinson E, Mauffret A, AoIfe SE, Woskett CM(1989) Adsorption at interfaces in dairy sys-tems. J Soc Dairy Technol42, 18-22

Dickinson E, Euston SR, Woskett CM (1990)Competitive adsorption of food macromole-cules and surfactants at the oil-water inter-face. Prog Col/oid Polym Sci82, 65-75

Doxatakis G, Sherman P (1986) The interactionof sodium caseinate with monoglyceride anddiglyceride at the oil-water interface and itseffect on interfacial rheological properties.Col/oid Polym Sci 264, 254-259

Rsher LR, Parker NS (1988) Effect of surfac-tants on the interactions between emulsiondroplets. In: Advances in Food Emulsionsand Foams (Dickinson E, Stainsby G, eds)Elsevier Appl Sci Publ, London, 45-90

Rsher LR, Mitchell EE, Parker NS (1985) Inter-facial tensions of commercial vegetable oilswith water. J Food Sci 50, 1201-1202

Gacula MC, Singh J (1984) Statistical Methodsin Food and Consumer Research. Food Sci-ence and Technology. Academic Press, NY,214-272

Goulden JDS, Phipps LW (1964) Factors affect-ing the fat globule size during the homogen-ization of milk and cream. J Dairy Sci 31,195-200

Halling PJ (1981) Protein-stabilized foams andemulsions. CRC Crit Rev Food Sci Nutr 13,155-203

Jenness R, Walstra P (1984) Dairy Chemistryand Physies. John Wiley and Sons, NY, 88-89

530

Kato A (1991) Significance of macromolecularinteraction and stability infunctional proper-ties of food proteins. In: Interaction of FoodProteins (Parris N, Barford R, eds) ACSSymp Ser 454, Am Chem Soc, Washington,DC,13-24

Kessler HG (1981) Food Engineering and DairyTechnology. Verlag A Kessler, Freising, 119-138

Koppel DE (1972) Analysis of macromolecularpolydispersity in intensity correlation spec-troscopy: the method of cumulants. J GhemPhys 57,4814-4820

Little TM (1981) Interpretation and presentationof results. Hortic Sei 16, 19-31

Lorient D, Closs B, Courthaudon JL (1991) Con-naissances nouvelles sur les propriétés fonc-tionnelles des protéines du lait et de ses déri-vés. Lait 71, 141-171

MacRitchie F (1976) Monolayer compressionbarrier in emulsion and foam stability. J Gol-loid Interface Sci 56, 53-56

Montgomery DC (1976) Design and Analysis ofExperiments. John Wiley and Sons, NY,341-368

Neter J, Wasserman W, Kutner M (1985) Ap-plied Linear Statistical Models: Regression,Analysis of Variance and Experimental De-signs. Irwin, Homewood, 2nd edn

Ogden LV (1973) The homogenization-inducedclustering of fat globules in cream and modelsystems. Ph Dissertation, Univ Minnesota, StPaul, MN

Oorwijn H, Walstra P (1979) The membranes ofrecombined fat globules. 2. Composition.Neth Milk Dairy J 33, 134-154

Paquin P, Britten M, Laliberté MF, Boulet M(1987) Interfacial properties of milk caseinproteins. In: Proteins at Interfaces: Physico-cnemicel and Biochemical Studies (BrashJL, Horbett TA, eds) ACS Symp Ser 343,Washington DC, 677-686

Paquin P, Giasson J (1989) La microfluidisationcomme procédé d'homogénéisation d'uneboisson à base de matière grasse laitière.Lait 69, 491-498

Pearce KN, Kinsella JE (1978) Emulsifyingproperties of proteins: evaluation of a turbidi-metric technique. J Agric Food Ghem 26,716-723 .

o Robin etaI

Phipps LW (1975) The fragmentation of oildrops in emulsions by high-pressure homog-enizer. J Phys 0: Appl Phys 8, 448-462

Phipps LW (1982) Homogenizing valve designand its influence on milk fat globule disper-sion. 1. Low rate of flow (100 1.h-1 Re :>3000). J Dairy Res 49, 309-315

Phipps LW (1983) Effects of fat concentrationon the homogenization of cream. J Dairy Res50,91-96

Phipps LW (1985) The High Pressure Dairy Ho-mogenizer. Tech Bull 6. Nat Inst Res Dairy-ing, Reading, UK

Piggott JR (1986) Statistieal Procedures in FoodResearch. Elsevier Appl Sci Publ, London,101-123

Pouliot Y, Paquin P, Robin 0, Giasson J (1991)Étude comparative de l'effet de la microfluidi-sation et de l'homogénéisation sur la distribu-tion de la taille des globules de gras du laitde vache. Int Dairy J 1, 39-49

Robin 0, Paquin P (1991) Evaluation of the par-ticle size of fat globules in a milk model emul-sion by photon correlation spectroscopy. JDairy Sci74, 2440-2447

SAS® (1990) SAS User's Guide: statistics Ver-sion 6.06. SAS Inst Inc, Cary, NC

Shinoda K, Kunieda H (1983) Phase propertiesof emulsions: PIT and HLB. In Encyclopediaof Emulsion Technology. 1. Basic Theory(Becher P, ed) Marcel Dekker, NY, 337-367

Sweetsur MA, Muir DD (1983) Effect of homog-enization on the heat stability of milk. J DairyRes 50, 291-300

Tadros TF, Vincent B (1983) Emulsion stability.ln: Encyclopedia of Emulsion Technology. 1.Basic Theory (Becher P, ed) Marcel Dekker,NY,129-285

Tomberg E (1980) Functional characteristics ofprotein stabilized emulsions: emulsifying be-havior of proteins in a sonifier. J Food Sci 45,1662-1668

Tornberg E, Oison A, Persson K (1990) Structu-rai and interfacial properties of food proteinsin relation to their function in emulsions. In:Food Emulsions (Larsson K, Friberg SE, eds)Marcel Dekker, NY, 247-326

Vuillemard JC (1991) Recent advances in thelarge scale production of Iipid vesicles for use


in food products: microfluidization. J Microen-capsula, 547-562

Walstra P (1969) Preliminary note on the mech-anism of homogenization. Neth Milk Dairy J23,290-292

Walstra P (1974) Influence of rheological prop-erties of both phases on droplet size of O/Wemulsions obtained by homogenization andsimilar processes. Dechema Monogr77, 87-94

Walstra P (1975) Effect of homogenization onthe fat globule size distribution in milk. NethMilk Dairy J 29, 279-294

Walstra P (1983) Emulsion formation. In: Encyclo-pedia of Emulsion Technology. 1. Basic Theo-ry(Becher P, ed) Marcel, Dekker, NY, 57-127

Walstra P, Oortwijn H (1982) The membranes ofrecombined fat globules. 3. Mode of forma-tion. Neth Milk Dairy J 36, 103-113

Washington C (1987) Emulsion production bymicrofluidizer. Lab Equip Dig 85, 69-71

Weiner BB, Tschamuter WW (1987) Uses andabuses of photon correlation spectroscopy inparticle sizing. In: Particle Size Distribution:Assessment and Characterization (ProvderT, ed) ACS Symp Ser 332, Am Chem Soc,Washington OC, 48-64

Wolfram S (1991) Mathematica: A System forDoing Mathematics by Computer. Addison-Wesley Publ Co Inc, Champaign, IL, 2nd edn

microfluidization of dairy model emulsions. 1.preparation of

Documents