The rheology of aqueous emulsions prepared by direct emulsification and phase inversion from a high...

Colloids and Surfaces

A: Physicochemical and Engineering Aspects 196 (2002) 121–134

The rheology of aqueous emulsions prepared by directemulsification and phase inversion from a high viscosity

alkyd resin

David J. Watson, Malcolm R. Mackley *Department of Chemical Engineering, Uni�ersity of Cambridge, Pembroke Street, Cambridge CB2 3RA, UK

Received 20 October 1999; accepted 14 June 2001

Abstract

A comparison of the rheological properties of dispersions of an alkyd resin in water prepared by directemulsification and phase inversion has been conducted. These dispersions have application as the base dispersion inthe manufacture of the new generation of water based gloss paints. An experimental investigation of the effect ofdispersed phase fraction, droplet size and dispersion age on the rheological properties of dispersions produced by eachemulsification route has been carried out. The droplet size distributions of the dispersions are also characterised. Itwas found that phase inversion always gave a smaller droplet size distribution than direct emulsification. Surprisinglyit was found that for a given dispersed phase fraction, the dispersions with a smaller droplet size, produced by thephase inversion route, have a lower zero shear rate viscosity than those produced by the direct emulsification route.The rheology of the direct emulsification samples was also observed to change with age, whilst with phase inversionsamples the rheology was stable. It is proposed that this and other differences observed between the two methods ofmanufacture can be associated to the presence, or otherwise, of excess surfactant in the aqueous phase. © 2002Elsevier Science B.V. All rights reserved.

Keywords: Emulsions; Phase inversion; Droplet size; High viscosity; High dispersed phase

www.elsevier.com/locate/colsurfa

1. Introduction

Alkyd resin dispersions are used in the paintindustry as the basis for the new generation ofenvironmentally friendly water based gloss paints(see, for example, [1,2]). These dispersions, of ahighly viscous resin (above 200 Pa s at 20 °C)dispersed in water, can be manufactured by directemulsification or via phase inversion to produce a

final product dispersion generally containing 50–60% dispersed phase volume. When manufactur-ing dispersion by direct emulsification the alkydresin, oil, phase is simply added to the aqueousphase. Energy then imparted to the system bymixing is used to create the new surfaces of thedroplets and thus form an oil-in-water dispersion.During manufacture by phase inversion, however,water is initially added to the agitated alkyd phaseuntil a critical dispersed phase fraction is reachedand the dispersion inverts to produce the requiredoil-in-water dispersion.* Corresponding author.

0927-7757/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved.

PII: S0927 -7757 (01 )00813 -5

D.J. Watson, M.R. Mackley / Colloids and Surfaces A: Physicochem. Eng. Aspects 196 (2002) 121–134122

Commercial dispersions are further processedfollowing initial manufacture to add the remain-ing constituents of the finished paint, including,for example, pigments. During both the formationprocess and subsequent processing of the disper-sions, knowledge of the rheological properties isimportant for effective processing to be achieved.The high viscosity of the alkyd resin presentsdifficulties during dispersion manufacture. There-fore, the choice of manufacturing route can becrucial to controlling the resulting droplet sizedistribution and thus rheological properties of thedispersion, and in the case of a finished paint, thewetting and film forming characteristics.

The rheological properties of emulsions are ingeneral complex and are governed by dropletinteractions (see, for example, [3,4]) and these arein turn affected by numerous factors includingdispersed phase fraction, continuous phase viscos-ity, droplet size distribution, droplet deformabil-ity, surfactant concentration, temperature anddispersion age. The viscosity enhancement causedby the suspension and interaction of dropletswithin an immiscible liquid has been the subjectof numerous studies and [5–7] for example, givegood reviews on this subject.

As the dispersed phase fraction of a dispersionis increased, the dispersion rheology changes sig-nificantly as the frequency of droplet interactionincreases. [8] developed an expression to describethis viscosity enhancement at low concentrationsfor hard spheres, which was subsequentlymodified by [9] to account for droplet deformabil-ity. Further studies have lead to expressions forviscosity enhancement at higher dispersed phasefractions, for example, the semi-empirical equa-tion of [10]. Over the wide range of dispersedphase fraction that can be achieved, the frequencyof droplet interactions changes greatly [11]. Belowa volume fraction of order 0.3, droplets can essen-tially move freely past each other, however, as thedispersed phase fraction is increased droplet inter-actions increase, until above fractions of 0.74,when the droplets become tightly packed andmovement of droplets past one another becomesseverely impaired. As the dispersed phase fractionchanges through this range, the rheological char-acteristics of the dispersion also consequently

change from Newtonian, to shear thinning, to aviscoelastic type of behaviour [7].

The interfacial rheology of the droplets cansignificantly influence droplet interactions, whichare enhanced by the presence of a surfactant (see,for example, [3,12]). At high surfactant concentra-tions, excess surfactant micelles, which can formin the continuous phase also influence the rheo-logical characteristics of dispersions by inducingdepletion flocculation [13,14], Pal and Rhodes,1989). For a given dispersed phase fractionsmaller droplets undergo a greater number ofdroplet interactions and therefore decreasing thedroplet size of a monodispersed dispersion in-creases the viscosity, an effect which decreases asthe dispersion droplet size distribution broadens(see, for example, [15,12]).

It is well known that the manufacturing routecan affect the droplet size distribution of disper-sions (see, for example, [16,17]). Direct emulsifica-tion will occur provided sufficient energy can beimparted to the system to break-up the dispersedphase and suspend it as droplets in the continuousphase. Phase inversion, however, can be achievedthrough two routes. One route involves changesto the relative phase concentrations and is termedcatastrophic phase inversion. Alternativelychanges to the hydrophilic lypophilic balance(HLB) value of the surfactant [18], which is ameasure of the surfactant’s affinity for theaqueous or oil phase, can also cause phase inver-sion and is known as transitional phase inversion.The former route has been found to generate asimilar droplet size distribution to formation bydirect emulsification [29] whereas in the latter casea much smaller droplet size distribution can beachieved. This difference has been attributed tothe formation of a co-continuous phase duringthe moment of phase inversion that has a verylow interfacial tension [19].

In this paper, investigation of how manufactur-ing route affects the rheological properties ofdispersions has been conducted for a single alkydresin system. The rheology, as a function of defor-mation, age history and dispersed phase fractionhas all been explored.

D.J. Watson, M.R. Mackley / Colloids and Surfaces A: Physicochem. Eng. Aspects 196 (2002) 121–134 123

2. Experiments and modelling

Rheological tests were carried out using aRheometrics dynamic spectrometer (RDSII),which is a controlled strain rheometer. Experi-ments were conducted using either a 50 or 25 mm(for viscoelastic samples) parallel plate geometry.Mackley et al. [20] have earlier demonstrated theapplicability of a KBKZ type constitutive equa-tion with a Wagner damping function, k, to de-scribe the rheological characteristics of a range ofcomplex fluids and this method has been usedhere to model the rheological response of thedispersions under test.

The so-called KBKZ model is a strain depen-dant integral constitutive equation that uses adiscrete set of Maxwell elements to give a spec-trum of relaxation times that describe the fluidslinear viscoelastic behaviour. Non-linearity is thenintroduced via a strain dependent damping func-tion, k, the Wagner damping function. This classof constitutive equations is described well by Lar-son [21].

The procedure used to obtain the rheologicalcharacteristics of a material consists of severalsequential tests. Initially, a strain sweep is carriedout to investigate the extent of any viscoelasticbehaviour. This test is then repeated to ensurethat a consistent test response is achieved. Fromthe strain sweep data a strain is chosen at whichto conduct a frequency sweep. A strain is selectedat which the material response is independent ofstrain and the torque reading is well within therange of the force transducer. A discrete relax-ation spectrum is then calculated from a best fit tothe frequency sweep data using a rheometricssoftware package, and its accuracy checked byusing the result to re-calculate the variation of thedynamic moduli with frequency. For viscoelastic

dispersions, step strain experiments are then per-formed. Where these tests were possible, a Wag-ner type damping coefficient could be obtainedand used to predict the steady shear response,which is compared with the experimental steadyshear curve. For full details of the experimentalprotocol used see [20].

The constituent components of the dispersionsto be used are given in Table 1 and their respec-tive Newtonian viscosities specified.

3. Material samples

Samples of dispersions with a range of dis-persed phase fraction were manufactured by di-rect emulsification and via phase inversion (by thecatastrophic route [29]). The mixing techniques ofoscillatory flow mixing (OFM) (see for example,[22]) and a conventional stirred tank were utilisedfor this, as described by Watson [23]. The disper-sions contain a surfactant system (consisting of anon-ionic surfactant and a co-solvent) at concen-trations (10% by mass of surfactant and 5% co-solvent) that result in a significant excess ofsurfactant. In general this leads to the formationof dispersions with a stable droplet size dis-tribution in unsheared material where droplet co-alescence is considered to be negligible. Represen-tative samples of these dispersions were readilyobtained by simply using a spatula to place asample of the dispersion into the test area of therheometer. A new sample was used for each test,before which thorough cleaning of the sampleplates with white spirit (in which the alkyd resin issoluble) and then distilled water was carriedout. By varying the mixing intensity at whichdispersions were manufactured, dispersions with arange of droplet size distributions were manu-factured.

Table 1Physical properties of the materials used for the production of alkyd resin dispersions

Density at 20 °C (kg m−3) Melting point (°C)Material Viscosity at 20 °C (Pa s)

200Alkyd resin –1050– 40Surfactant 1100

–950Approx. 10−3Co-solvent


Fig. 1. Rheological characterisation of an alkyd resin dispersion containing 50% dispersed phase, manufactured by directemulsification with 10% surfactant and x50=7.3 �m. Tests conducted at 20 °C. (a) Strain sweep �=1 rad s−1, (�) G�, (�) G�(�) �*.(b) Frequency sweep �=30%, (�) G�, (�) G�, (— ) fit, �* (�). (c) Spectrum of relaxation times. (d) Steady shearmeasurements (�) �, (�) shear stress, Wagner prediction (— ) calculated using k=0.

When manufacturing dispersions by directemulsification, samples with 10, 20, 30, 40 and50% dispersed phase were prepared for testing.With the mixing apparatus available dispersedphase volumes greater than 50% could not beachieved. However, for the alkyd resin systemused, phase inversion from a water-in-oil to oil-in-water dispersion occurs at an alkyd resin concen-tration of around 92% [23] and hence, whenutilising the phase inversion route samples withhigher dispersed phase volumes could also beproduced. Thus samples with 20, 30, 40, 50, 60,70, 80 and 90% dispersed phase were prepared foranalysis via phase inversion. A dispersion of 5%water in alkyd was also tested. All these disper-sions were stable over the duration of the testingprocedure.

For the alkyd resin in water dispersions, dropletsize distributions were obtained using a Sympatec

laser light scattering particle size analyser (fordetails of the operation of this type of particle sizeanalyser see, for example, [24]).

4. Results

4.1. Manufacture by direct emulsification

4.1.1. Effect of dispersed phase fractionFig. 1 shows an example of the rheological

characteristics of a fresh dispersion manufacturedby direct emulsification and containing 50% dis-persed phase, in this case with a median dropletsize (x50) of approximately 7 �m. Tests were car-ried out at 20 °C using a 100 g cm force re-bal-ance transducer and a 50 mm diameter parallelplate at a gap of 0.5 mm. Fig. 1(a) shows thedynamic strain sweep, conducted at a frequency


of 1 rad s−1, and shows that the alkyd resindispersion is a dominantly viscous material. Alinear region is observed at the lower strainsinvestigated, with a slight thickening of the disper-sion occurring at strains above 100%. A possibleexplanation for this behaviour may be providedby the mechanism of depletion flocculation causedby the presence of excess surfactant micelles in thecontinuous phase [25,13]. This effect is caused bya decrease in osmotic pressure between approach-ing droplets as the excess surfactant micelles areexcluded from the gap between the droplets, thusforcing the droplets together. The strain sweepdata suggests this may happen as the strain isincreased during the test and the droplets areforced past each other with increasing frequencyand intensity.

The dynamic frequency sweep, shown in Fig.1(b), was conducted at 30% strain and shows thatthe complex viscosity decreases with increasingfrequency of deformation. Prediction of the dy-namic moduli, using the relaxation spectrumshown in Fig. 1(c), is good, however, the structureof the discrete spectrum shows little physical sig-nificance. The dispersion is able to relax signifi-cantly in a very short time and hence step strainexperiments could not be performed.

The relatively high volume fraction of dispersedphase tested in the example of Fig. 1 suggests thatthe material will exhibit shear thinning properties,however, as shown in Fig. 1(d) the experimentallymeasured apparent viscosity is essentially constantover the wide range of shear rates tested. Conse-quently, the non-linear damping parameter in theWagner KBKZ type equation is put to zero [20].The model prediction of the steady shear be-haviour is also shown in Fig. 1(d). The consistentover prediction of the experimental data by themodel suggests that wall slip between the rheome-ter plate and sample of dispersion is occurringduring the experimental simple shear data acquisi-tion. This observation is also consistent with re-cent work carried out in this laboratory [26].

Reducing the dispersed phase fraction decreasesthe number of droplet interactions and thereforethe dispersion viscosity. The essentially Newto-nian behaviour observed for a dispersion with50% dispersed phase, as shown in Fig. 1, is re-

tained over the range of dispersed phase fractionsinvestigated and below 30% dispersed phase theviscosity is independent of deformation ratethroughout the range tested. Fig. 2 shows thevariation of apparent and complex viscosity withdispersed phase fraction for dispersions with amedian droplet size, x50=7.3 �m. In this exam-ple, the value of the complex viscosity is arbitrar-ily taken at a strain of 1000%, however, only inthe case of the highest phase volume of 50% doesthe choice of strain affect the result. The steadyshear data shows near Newtonian behaviour forall shear rates tested, as shown in Fig. 1(d) forexample. In Fig. 2 an increase in dispersed phasefraction causes an increase in the dispersion vis-cosity as expected, however, the apparent experi-mental steady shear viscosity is consistentlygreater than the experimental measured complexviscosity over the range of dispersed phase frac-tions investigated. This difference showed a repro-ducible trend and at present we have no clearinterpretation to explain this effect.

The Krieger–Dougherty equation [10],

�r=�

1−��p

�n− [�]�p

(1)

where: �r, relative viscosity; �, dispersed phasefraction; �p, maximum packing fraction; and [� ],intrinsic viscosity=2.5 for hard sphere dispersion;has been fitted to the data. A maximum packing

Fig. 2. Variation of apparent (�) and complex viscosity (�)measurement with dispersed phase fraction, (— ) Krieger–Dougherty fit �=6.5 for complex viscosity and �=8.75 forapparent viscosity.


Fig. 3. The variation of apparent viscosity at a strain of 1000%with dispersed phase fraction for a range of droplet sizedistributions for dispersions manufactured by direct emulsifi-cation, (�) x50=12.8 �m, (�) x50=9.1 �m, (�) x50=7.7 �m,(�) x50=7.3 �m.

tent variation between samples with varyingdroplet size distribution. The polydispersityshown in the droplet size distribution data of Fig.4 will, however, make it difficult for any signifi-cant variation in rheological characteristics withdroplet size distribution to be observed. Polydis-persity is known to influence the observed rheol-ogy with smaller droplets able to fit in the voidsbetween larger droplets and thus reduce the effec-tive viscosity enhancement of the droplets [14], aneffect that becomes more significant as the poly-dispersity increases. The high degree of polydis-persity shown in Fig. 4, which is essentiallyconsistent in each dispersion tested, will thereforebe expected to have a large influence over thedispersion rheology. This result also suggests thatthe possible influence of excess surfactant on thedispersion rheology, as seen in Fig. 1, is greaterthan any effect resulting from variation in disper-sion droplet size.

4.1.3. Effect of dispersion ageDispersions were left to age at room tempera-

ture for up to a week and the rheological proper-ties tested at various intervals during this time. Ofthe dispersions manufactured by direct emulsifica-tion only those containing 50% dispersed phasewere tested because sedimentation became a prob-lem in dispersions with lower dispersed phaseconcentrations over this time scale.

fraction of 0.92, based on the average dispersedphase fraction required to achieve phase inversionfor this system, has been used [23]. The alkydresin droplets behave as soft polyhedra and aretherefore able to pack together to a high dispersedphase fraction. This is greater than that possibleby dispersions of hard spheres, as assumed by theKrieger–Dogherty equation. The hard spheres as-sumption also results in an intrinsic viscosity of2.5 generally being used with the equation. How-ever, to fit the viscosity data of these alkyd resindispersions, values for the intrinsic viscosity of 6.5and 8.75 have been required for the complex andapparent viscosities respectively.

4.1.2. Effect of droplet sizeIn dispersions containing 50% dispersed phase

there will be significant droplet interactions andtherefore dispersions with different droplet sizedistributions will be expected to exhibit differentrheological properties, the effect diminishing asthe dispersed phase fraction decreases (see, forexample, [7,27]). Fig. 3 shows the effect of dropletsize distribution on the variation of the apparentand complex viscosity of these alkyd resin disper-sions over a range of dispersed phase fractions.The corresponding droplet size distributions, on avolume basis, of the dispersions are shown in Fig.4. Fig. 3 shows that there is generally little consis-

Fig. 4. Comparison of the droplet size distributions of thedispersions used for the rheological characterisations shown inFig. 3, (�) x50=12.8 �m, (�) x50=9.1 �m, (�) x50=7.7 �m,(�) x50=7.3 �m.


Fig. 5. Rheological characterisation of an alkyd resin dispersion manufactured by direct emulsification containing 50% dispersedphase, conducted after 3 days at 20 °C. (a) Strain sweep �=1 rad s−1, (�) G�, (�) G�, (�) �*. (b) Frequency sweep �=1%, (�)G�, (�) G�, (— ) fit, (�) �* (c) Spectrum of relaxation times. (d) Steady shear measurements (�) �, (�) shear stress.

Fig. 5(a) shows a dynamic strain sweep of a50% dispersed phase dispersion aged for 3 dayscarried out at a frequency of 1 rad s−1. Theresponse is more characteristic of a ‘typical’ liq-uid– liquid dispersion with higher dispersed phasefraction than that of the fresh dispersion shown inFig. 1, with a linear region present at low strainsbefore internal structures are broken and the ma-terial viscosity begins to decrease. The elastic andviscous moduli are of comparable magnitude ini-tially and the viscous modulus dominates athigher strains. This response suggests that whenthe dispersions are left to age, the droplets areable to interact to form an internal network offlocs, resulting in a significant increase in the lowstrain rate viscosity of the dispersion. This struc-ture is subsequently destroyed under higher defor-mation rates.

A frequency sweep was conducted within thelinear region at a strain of 1%, and is shown inFig. 5(b). Prediction from the relaxation spectrum

of Fig. 5(c) is generally good although a slightunder prediction of the viscous modulus is seen athigher frequencies. In the relaxation spectrum aninitial decrease of relaxation strength with anincrease in relaxation time is seen, but at higherrelaxation times the relaxation strength increases,indicating the influence of the elastic modulus atlower strains.

Under steady shear, the dispersion shows theclassic shear thinning type of behaviour of aliquid– liquid dispersion, as shown in Fig. 5(d).Meaningful data from step strain experimentscould not be obtained as the dispersion relaxesbefore the deformation has been fully applied.Therefore fitting of the Wagner model is notpossible.

Fig. 6 shows how the variation of complexviscosity with strain changes with dispersion age.A large increase in viscosity is seen over the first 3days as internal structures develop within thedispersion, however, only a refinement of this


Fig. 6. Change in the variation of complex viscosity with strainat 20 °C for a dispersion manufactured by direct emulsifica-tion containing 50% dispersed phase aged over 7 days, �=1rad s−1, (�) 0 days, (�) 3days, (�) 7 days.

Fig. 8. Variation of complex viscosity with strain for 7-day-old50% dispersed phase dispersions with a range of mediandroplet sizes manufactured by direct emulsification, (�) x50=1 4 �m, (�) x50=8 �m, (�) x50=5 �m.

structuring occurs subsequently, as further viscos-ity increases are slight and the shape of the re-sponse remains the same. The presence of thesurfactant prevents coalescence from occurringand over this period of ageing the droplet sizedistribution remains constant, as shown in Fig. 7.Therefore, the evolving rheological characteristicis a result of the formation of an inter-dropletnetwork. Weak inter-droplet interactions are seenin the non-linear region at low strains in tests onthe fresh dispersion, Fig. 1(a), indicating the po-tential for the formation of flocs seen in the ageddispersions.

Fig. 8 shows how the complex viscosity of

dispersions, with the range of droplet size distri-butions shown in Fig. 9, change with strain. Thisshows that the droplet size distribution of thedispersion has a large effect on the viscosity foraged dispersions compared with the freshly manu-factured dispersions. This is despite the significantpolydispersity of the dispersion shown in Fig. 9. Alinear region is present for dispersions withsmaller droplets, whereas the dispersion with amedian droplet size of 14 �m thins to a constantviscosity after small increases in strain. All thedispersions have the same dispersed phase frac-tion of 50%, but in the sample with the smallestmedian droplet size there will be a greater number

Fig. 9. Droplet size distribution of dispersions used to obtaindata for Fig. 8, (�) x50=14 �m, (�) x5o=8 �m, (�) x50=5�m.

Fig. 7. Variation of droplet size distribution with dispersionage of 50% dispersed phase dispersions used for Fig. 6, (�) 0days, (�) 7 days.


of inter-droplet interactions, thus enabling morecomplex internal structures to develop within thedispersion and a greater increase in viscosity gen-erated. The increase in droplet packing of a morepolydispersed sample decreases the effective vis-cosity enhancement of the droplets in comparisonwith dispersion with a more monodisperse sizedistribution [14]. Thus, the difference betweensamples shown in Fig. 9 is enhanced by the largevariation in polydispersity. This result shows thatthe droplet size has an influence when flocs havehad time to form, probably by depletion floccula-tion, but, as shown in Fig. 3, does not when theseflocs are not present.

4.2. Manufacture by phase in�ersion

Fig. 10 shows the rheological characteristics of90% dispersed phase dispersion manufactured viaphase inversion. These tests were conducted at20 °C using a 25 mm parallel plate at a gap of 0.8mm and a 100 g cm torque transducer. A dynamicstrain sweep conducted at a frequency of 1 rads−1 is shown in Fig. 10(a). The dispersion isdominantly elastic at lower strains with a signifi-cant linear region. Once a sufficiently large strain,of around 10% is reached, the dispersion viscositybegins to decrease and the viscous modulus be-comes gradually more influential, becoming domi-nant above strains of 100%. This type of responseis typical of a liquid– liquid dispersion with a highdispersed phase fraction (for example, [7,28]).

A frequency sweep, carried out at 0.1% strain,is shown in Fig. 10(b). The complex viscositydecreases considerably with frequency of defor-mation and the elastic modulus is dominant forthe majority of the frequency range investigated.Prediction of the viscous modulus is good whenusing the relaxation spectrum shown in Fig. 10(c)but the elastic modulus is under predicted. Fromthe step strain data of Fig. 10(d) damping coeffi-cients of 1.3 at 100% strain and 0.84 at 300%strain were found. The step strain tests do notproduce parallel responses and consequently esti-mation of the damping factor is unreliable. Thisdependency of the damping factor on strainmakes prediction of the steady shear data of Fig.10(e), which shows that the dispersion has charac-

teristic shear thinning properties, difficult. Thelarge discrepancy in the prediction of experimen-tal steady shear data is believed to be a result ofslip between the dispersion and rheometer platesduring measurement, as reported by [26] for cer-tain emulsion systems.

The rheological response of 50% dispersedphase dispersion manufactured by phase inversionis shown in Fig. 11. Fig. 11(a) shows a strainsweep carried out at a frequency of 1 rad s−1

indicating the general Newtonian type behaviourof dispersions manufactured by phase inversion atthis dispersed phase fraction. A frequency sweepat 30% strain is shown in Fig. 11(b) and shows aslight decrease in complex viscosity with increas-ing frequency of deformation, with the viscousmodulus again dominant. Increasing relaxationstrength increases the relaxation time linearlythroughout the range, as shown in Fig. 11(c),which predicts the frequency sweep data well. Aslight shear thinning is observed under steadyshear with a constant viscosity approached abovea shear rate of 100 s−1. It was not possible toobtain meaningful step strain data due to the fastrelaxation of the sample before deformation hadbeen fully applied. Therefore, fitting of the Wag-ner model for this sample has not been possible.

The rheological response described in Fig. 11 issignificantly different to that for a dispersion pro-duced by direct emulsification and is more typicalof that expected of dispersion behaviour. Thelinear strain sweep and shear thinning behaviourwas not seen in dispersions produced by directemulsification at the same dispersed phase frac-tion. The smaller droplets generated by phaseinversion reduce the level of excess surfactantpresent, thus decreasing the possibility of the for-mation of flocs by depletion flocculation. Thisallows the droplets to move past each other morefreely as the shear rate increased during a steadyshear test.

As with direct emulsification, below 50% dis-persed phase, dispersions retain a Newtoniancharacteristic. However, as the dispersed phaseincreases above 50% the viscoelastic propertiesgradually increase until those of 90% dispersedphase, as shown in Fig. 10, are observed.


Fig. 10. Rheological characterisation, at 20 °C, of a dispersion manufactured via phase inversion containing 90% dispersed phaseand with x50=1.2 �m. (a) Strain sweep, �=1 rad s−1 (�) G�, (�) G�, (�) �*. (b) Frequency sweep, �=0.l%, (�) G�, (�) G�,(�) �*. (c) Spectrum of relaxation times. (d) Step strain data for strains of 10, 30, 100 and 300 from top to bottom, respectively,from which values of k=1.3 for 100% strain and k=0.8 for 300% strain were calculated. (e) Steady shear measurements (�) �, (�)shear stress and predicted response for each value of k (k=1.3 top line).

4.2.1. Effect of droplet sizeDispersions with a narrow range of droplet size

distributions are produced by the phase inversionroute. Over this range, from about 0.8 to 1.5 �m,no influence of droplet size on the rheologicalcharacteristics of the dispersions was observed.This result is similar to that seen for dispersionsmanufactured by direct emulsification.

4.2.2. Effect of dispersion ageIn contrast to direct emulsification the rheology

of dispersions manufactured by phase inversiondoes not change with age. This is illustrated inFig. 12 which shows the change of strain sweepfor a dispersion with 80% dispersed phase agedover 2 days. There is little difference in responseover this time and if anything the moduli show a


Fig. 11. Rheological response, at 20 °C, of a dispersion with 50% dispersed phase manufactured by phase inversion. (a) Strainsweep, �=1 rad s−1, (�) G�, (�) G�, (�) �*. (b) Frequency sweep, �=30%, (�) G�, (�) G�, (�) �*. (c) Spectrum of relaxationtimes. (d) Steady shear measurements (�) �, (�) shear stress.

slight decrease in magnitude, suggesting a reduc-tion in internal structure rather than the forma-tion of flocs seen in dispersions produced by thedirect emulsification route. Further ageing of thedispersions did not have any further effect on theresponse.

At lower dispersed phase fractions the situationis the same. This is illustrated in Fig. 13, whichshows the change of strain sweep for a 50%dispersed phase dispersion, which has been agedfor 12 days.

This age stability is a significantly differenttrend to that seen in dispersions manufactured bydirect emulsification. As indicated above, thedroplet sizes produced by phase inversion aresignificantly smaller than those generated by di-rect emulsification, with median sizes around 1�m compared with 7–14 �m, respectively, for thedispersions used in this investigation. The same

Fig. 12. Change in strain sweep response, at 20 °C, withdispersion age for an 80% dispersed phase dispersion manufac-tured by phase inversion, (�) G�, (�) G�, (�) �*, (�) G�after 2 days, (�) G� after 2 days, (�) �* after 2 days.


Fig. 13. Change in strain sweep response, at 20 °C, withdispersion age for a 50% dispersed phase dispersion manufac-tured by phase inversion, (�) G�, (�) G�, (�) �*, (�) G�after 2 days, (�) G� after 2 days, (�) �* after 2 days.

Fig. 14. Variation of complex and apparent viscosities, at20 °C, with dispersed phase fraction of dispersions manufac-tured by direct emulsification (DE) and phase inversion (PI).The Krieger–Dougherty (K–D) equation has been fitted toeach set of data. The viscosity of the base alkyd resin mix anda 5% water-in-alkyd dispersion are also shown. (�) �* (at�=1000%) DE, (— ) fit for �=6.5; (�) �* PI, (— ) fit for�=5.5;(�) � (at 100s-1) DE, (— ) fit for �=8.75; (�) � (at100s-1) PI, (— ) fit for �=4; (�) 5% water in alkyd.

concentration of surfactant was used during man-ufacture by both routes and therefore there isconsiderable less excess surfactant present in dis-persion produced by phase inversion than thosemanufactured by direct emulsification. This de-crease in excess surfactant can be expected toreduce the occurrence of depletion flocculation.Figs. 12 and 13 suggest that there is not sufficientexcess surfactant to cause any flocculation to oc-cur and the droplets are able to move past eachother in both the fresh and aged dispersions.

4.3. Comparison between manufacturing methods

There are several differences in the rheologicalresponse of dispersions manufactured by eachemulsification route. For a given dispersed phasefraction (Figs. 1 and 11) the viscosity of thedispersion produced by phase inversion is signifi-cantly lower, by almost an order of magnitude.Shear thinning behaviour in the steady shear testis seen with phase inverted dispersions, but not byfresh dispersions manufactured by directemulsification.

Fig. 14 shows the change in complex viscosityand apparent viscosity at 20 °C with dispersedphase fraction for dispersions manufactured byboth direct emulsification and phase inversion.The Newtonian viscosity of the dispersion wasfound where possible. This only affects the results

for dispersions with high dispersed phase fractionsmanufactured by phase inversion. The droplet sizedistributions of these dispersions are shown inFig. 15 and show that dispersions manufacturedvia phase inversion contain significantly smallerdroplets with a much narrower distribution. Theproduction of smaller droplets by the phase inver-sion route is unexpected as catastrophic inversionoccurs when the dispersed phase fraction ischanged and this has been shown to generatedroplets of a similar size to those obtained from

Fig. 15. Comparison between the droplet size distribution of adispersion manufactured by direct emulsification (DE) and viaphase inversion (PI). A log-normal distribution has been fittedto the distribution obtained from the phase inversion route. (�)PI, (— ) log-normal distribution fit, (�) DE.


direct emulsification [29]. This indicates that theinfluence of the surfactant on the emulsificationprocess, as well as the rheology, is significant.

The Newtonian viscosity of a 5% water in alkyddispersion was found and is also shown in Fig. 14.All the curves show the same expected trend ofincreasing viscosity with increasing dispersedphase and as the alkyd fraction is increased thedispersion viscosity increases up to the phase in-version boundary after which a decrease in viscos-ity is observed towards that of the neat alkydresin [5]. This figure shows clearly that for anequivalent dispersed phase fraction dispersionsproduced by phase inversion have a significantlylower viscosity than those produced by directemulsification. This is an unexpected result as thesmaller droplet size distribution produced byphase inversion would be expected to increase thedispersion viscosity for a given dispersed phasefraction [7]. Therefore, this suggests that the pres-ence of excess surfactant micelles has a greatereffect on the dispersion viscosity than the decreasein droplet size achieved through utilisation ofphase inversion. Also, whereas dispersions manu-factured by direct emulsification have a greaterapparent viscosity than complex viscosity at agiven dispersed phase fraction, the reverse is seenfor dispersions manufactured by phase inversion.This difference is significant at higher dispersedphase fractions. The difference between the twomixing methods will be enhanced after a period ofageing as flocs form in the dispersions manufac-tured by direct emulsification.

5. Conclusions

Experimental rheological data of alkyd resindispersions manufactured by direct emulsificationand phase inversion together with the associateddroplet size distributions have been presented.

Dispersions manufactured by direct emulsifica-tion for the range studied are weakly viscoelasticwith the Newtonian viscosity following aKrieger–Dougherty relation. Surprisingly, the ap-parent viscosity is consistently greater than thecomplex viscosity over a range of dispersed phasefractions. This is an effect that at present we

cannot explain. Dispersions manufactured byphase inversion show viscoelasticity at higher dis-persed phase fractions (above 80%), but nearNewtonian behaviour at lower dispersed phasefractions (below 50%). A KBKZ type model hasbeen fitted to the data and within the linearviscoelastic region we obtain a good fit, however,the steady shear prediction is poor, especially fordispersions manufactured by phase inversionwhich we believe is a result of wall slip.

For a given dispersed phase fraction and sur-factant concentration, the viscosity of dispersionmanufactured by phase inversion is significantlyless than that produced by direct emulsification.This is surprising, as Pal and Otsubo and Prud’-homme [27,30] for example, have shown smallerdroplets produce dispersions with a higher viscos-ity for a given dispersed phase fraction. The effectof ageing on the rheology of the dispersions issignificant for the direct emulsification route withviscoelasticity increasing with time. Utilisation ofthe phase inversion route yields dispersions withsignificantly smaller droplet size distributions andwith greater ageing stability. A possible explana-tion of this result is that the excess surfactant inthe direct emulsification sample is causing vis-coelastic structuring in the aqueous phase by de-pletion flocculation.

Acknowledgements

The authors wish to thank the EPSRC and ICIPaints, Slough, for financial support for thiswork. We would also like to thank ProfessorChris Gallegos (Huelva University, Spain) foruseful comments on the interpretation of certaindata.

References

[1] G.P.A. Turner, Introduction to Paint Chemistry andPrinciples of Paint Technology (Chap. 12), Third ed.,Chapman & Hall, London, 1988.

[2] J. Bentley, in: R. Lambourne (Ed.), Paint and SurfaceCoatings Theory and Practice, Ellis Horwood Ltd,Chichester, UK, 1987.


[3] Th. F. Tadros, Coll. Surf. A: Physicochem. Eng. 91 (1994)39.

[4] J. Israelachvili, Coll. Surf. A: Physicochem. Eng. Aspects91 (1994) 1.

[5] P. Sherman, Rheology of emulsions, in: P. Sherman (Ed.),Emulsion Science, Academic Press, New York, 1968.

[6] H.A. Barnes, J.F. Hutton, K. Walters, An Introduction toRheology, Elsevier Science Publishers, Amsterdam, 1989.

[7] H.A. Barnes, Coll. Surf. A: Physicochem. Eng. Aspects 91(1994) 89.

[8] A. Einstein, Annalytica Physica 19 (1906) 286.[9] G.I. Taylor, Proc. R. Soc. A138 (1932) 41.

[10] I.M. Kreiger, T.J. Dougherty, Trans. Soc. Rheol. 3 (1959)137.

[11] D.C. Peters, Dynamics of emulsification, in: N. Haruby,M.F. Edwards, A.W. Nienow (Eds.), Mixing in the ProcessIndustries, Butterworth, London, 1992.

[12] Y. Otsubo, R.K. Prud’homme, Rheol. Acta 33 (1994a) 29.[13] R. Pal, Chem. Eng. Comm. 98 (1990) 211.[14] R. Pal, Coll. Surf. A 64 (1992) 207.[15] R. Pal, AIChE J. 42 (1996a) 3181.[16] B.W. Brooks, H.N. Richmond, Coll. Surf. 58 (1991) 131.[17] J.M. Franco, A. Guerrero, C. Gallegos, Rheol. Acta 34

(1995) 513.

[18] W.C. Griffin, J. Soc. Cosmetic Chem. 1 (1949) 311.[19] B.W. Brooks, H.N. Richmond, Chem. Eng. Sci. 49 (1994a)

1053.[20] M.R. Mackley, R.T.J. Marshall, J.B.A.F. Smeulders, F.D.

Zhao, Chem. Eng. Sci. 49 (1994) 2551.[21] R.G. Larson, Constitutive Equations for Polymer Melts

and Solutions, Butterworths, Boston, 1988.[22] C.R. Brunold, J.C.B. Hunns, M.R. Mackley, J.W. Thomp-

son, Chem. Eng. Sci. 44 (1989) 1227.[23] D.J. Watson, The Application of Oscillatory Flow Mixing

to the Manufacture of Alkyd Resin Dispersions, Ph.D.Thesis, University of Cambridge, 1998.

[24] T. Allen, Particle Size Measurement, Fourth ed., Chapman& Hall, London, 1990.

[25] R. Pal, Chem. Eng. Sci. 52 (1997) 1177.[26] C. Bower, C. Gallegos, M.R. Mackley, J.M. Madeido,

Rheol. Acta 38 (1999) 145.[27] Y. Otsubo, R.K. Prud’homme, Rheol. Acta 33 (1994b) 303.[28] P. Taylor, Coll. Polym. Sci. 274 (1996) 1061.[29] B.W. Brooks, H.N. Richmond, Chem. Eng. Sci. 49 (1994b)

1065.[30] R. Pal, Chem. Eng. J. 63 (1996b) 195.[31] R. Pal, E. Rhodes, J. Rheology, 33 (1989) 1021.

The rheology of aqueous emulsions prepared by direct emulsification and phase inversion from a high...

Documents

Transcript of The rheology of aqueous emulsions prepared by direct emulsification and phase inversion from a high...