aggregation, viscosity measurements and direct observation of protein coated latex particles under...
TRANSCRIPT
Aggregation, viscosity measurements and direct observation of proteincoated latex particles under shear
Lars Hamberg, Pernilla WalkenstroÈm, Mats Stading, Anne-Marie Hermansson*
SIK, The Swedish Institute for Food and Biotechnology, PO Box 5401, SE-402 29 GoÈteborg, Sweden
Received 28 July 2000; revised 8 September 2000; accepted 9 November 2000
Abstract
The aggregation under shear, of latex particles coated with whey protein isolate was monitored, in a continuous phase with a complex
behaviour in relation to temperature dependence and shear thinning. The monitoring was done with viscosity measurements and microscopy.
An aggregating dispersion of whey coated polystyrene latex particles, salt, sucrose and gelatine was sheared in a rheometer at shear rates
between 0.05 and 5 s21. The viscosity was monitored as a function of time during a temperature increase from 30 to 608C. The viscosity
curves were interpreted with the aid of additional information from light microscopy micrographs. The aggregation was clearly visible as an
increase in viscosity. Aggregation was observed to initiate at a temperature between 40 and 508C. Unbound protein, i.e. protein not a part of
particle coating, was found to be essential for the aggregation of latex particles. After aggregation, a shear thinning behaviour was detected.
This was due to two phenomena: structural changes of the aggregates and shear thinning behaviour of the dispersion.
The build-up of the aggregates was followed by direct observation in a confocal laser scanning microscope. A sequence of micrographs
was taken, in an unstopped 3-D ¯ow ®eld generated in a four-roll mill, which showed the evolution of the size of the aggregates. The
micrographs were in good agreement with the viscosity measurements. This showed that the four-roll mill and a confocal laser scanning
microscope is a useful tool for studying aggregation in an undisturbed 3-D ¯ow. q 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Whey protein; Aggregation; Flocculation; Coated particles; Shear; Confocal laser scanning microscope
1. Introduction
Aggregating particle systems in general have been of
interest for a long time in a wide area of ®elds, such as
waste-water treatment, optical properties of ink, pulp pump-
ing, mineral recovery, nuclear material in power plants,
blood coagulation and food production (Amirtharajah &
O'Melia, 1990; Al-Jabari, Weber, & van de Ven, 1996;
Isaksson, Rigdahl, Flink, & Forsberg, 1998; Mathur,
Singh, & Moudgil, 2000). Particles coated with protein
can aggregate under certain conditions, which makes them
of special interest for several technological, immunological
and clinical applications (Horbett, 1982; Colowick &
Kaplan, 1987; Ortega-Vinuesa, Molina-BolõÂvar, &
Hidalgo-AÂ lvares, 1996). In this investigation of the aggre-
gation of protein coated latex particles, whey protein isolate
was chosen for the coating.
Common to nearly all of the ®elds mentioned above is
that the aggregation usually occurs in a ¯ow ®eld. The ¯ow
promotes the aggregation constantly by introducing
dynamic forces that act on the particles, e.g. rotation,
break up and reconstruction (Russel, 1980; van de Ven,
1982; West, Melrose, & Ball, 1994). The ¯ow, more speci-
®cally the shear ¯ow, also facilitates the aggregation by
increasing the rate of encounters between particles, which
increases the rate of aggregation (Walstra & Jenness, 1984).
It is not only the rate of ¯ow that is important: different types
of ¯ow, such as elongation, hyperbolic and shear ¯ow, also
introduce various effects on the aggregation and break up
(Higashitani & Iimura, 1998). One way to generate various
¯ow types is to use a four-roll mill, 4-RM, (Taylor, 1934).
The 4-RM can produce mixed ¯ow ®elds varying from shear
¯ow to hyperbolic ¯ow. There is also a stagnation point
where the ¯ow is slow, but where elongation and shear
forces act on the aggregates. Flow is known to affect the
aggregation of whey protein by introducing coarser aggre-
gates, when the suspension is sheared than there are in
unsheared suspension. Also the size of whey protein aggre-
gates is affected by shear and decreases with rising shear
rate (Taylor & Fryer, 1994; WalkenstroÈm, Panighetti, Wind-
hab, & Hermansson, 1998a; WalkenstroÈm, Windhab, &
Hermansson, 1998b; WalkenstroÈm, Nilsen, Windhab, &
Hermansson, 1999).
Food Hydrocolloids 15 (2001) 139±151
0268-005X/01/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved.
PII: S0268-005X(00)00060-6
www.elsevier.com/locate/foodhyd
* Corresponding author. Tel.: 146-31-335-5600; fax: 146-31-83-3782.
E-mail address: [email protected] (A.-M. Hermansson).
The chemical properties of the continuous phase
surrounding the particle are critical for the mechanism and
the outcome of the aggregation. Most of the work to inves-
tigate the mechanism of aggregation, has monitored it by
light scattering under static conditions for differing ion
concentrations, pH and temperatures (Quesada, Puig,
Delgado, Peula, Molina, & Hidalgo-AÂ lvarez et al., 1997;
Molina-BolõÂvar, Galisteo-GonzaÂlez, & Hidalgo-AÂ lvarez,
1998; Molina-BolõÂvar, Galisteo-GonzaÂlez, & Hidalgo-
AÂ lvarez, 1999a). In general the aggregates formed during
the aggregation of protein coated particles were found to be
fractal and the stability of the aggregates to be in agreement
with the DLVO (Derjaguin-Landau-Verwey-Overbeek)
theory. At low ion concentration, below 0.1 M, aggregation
increases with increasing concentration. Between concen-
trations of 0.1 M and around 0.5 M, the ion concentration
does not alter the aggregation; however at high concentra-
tion, above 0.5 M, the aggregation slows with rising ion
concentration (Molina-BolõÂvar, Galisteo-GonzaÂlez, &
Hidalgo-AÂ lvarez, 1999b).
The conditions that affect the protein are also important
for the aggregation. Effects of pH on the aggregation of the
protein b-lactoglobulin, the main protein in whey, have
been investigated after gelation. The gels have been found
to be strongest, and are de®ned as aggregate or particle gels,
at a pH interval between 4 and 6 (Stading & Hermansson,
1990, 1991). The rate of heat increase in thermally induced
aggregation of whey protein has also been reported to play a
role in aggregation and structuring of gels by altering the
size of aggregated particulate protein (Stading, Langton, &
Hermansson, 1993; Langton & Hermansson, 1996). Shear
in¯uenced thermal aggregation of whey protein and its
effects have been found for gel structures and viscoelastic
properties (WalkenstroÈn & Hermansson, 1996; Walken-
stroÈm et al., 1998a,b, 1999).
The surface properties, in particular the protein behaviour
at the surface, are another key factor for the aggregation.
Work has been done on the system of oil-in-water dispersion
with protein present at the surface and recently reviewed
speci®cally for milk proteins (Dickinson, 1999). With
milk protoins, the aggregation of oil drops stabilised by
caseinate under shear in an oil-in-water emulsion has been
investigated both by monitoring the viscosity with time and
by measuring the size of the aggregates (Dickinson & Gold-
ing, 1997a,b; Schokker & Dagleish, 1998). The investiga-
tions have shown that the amount of unadsorbed protein in
the continuous phase determines the aggregation and the
result of viscosity measurements. With protein present in
the continuous phase, they found that depletion ¯occulation
was the mechanism underlying the aggregation.
The traditional way to trace aggregation under shear is to
measure the viscosity. Doing this can sometimes lead to
dif®cult interpretations and often complimentary type of
measurement of aggregate size is required. To do this the
aggregates in the dispersions have to be withdrawn from the
¯ow, hence the dynamic of the forces and the balance from
forces acting on the aggregate in the ¯ow are disrupted. If
the aggregates are fragile, this treatment is critical and the
size measurements uncertain (Schokker & Dagleish, 1998).
However, simply placing a part of a process or a selected
dynamic condition under the microscope removes the
necessity to disrupt the dynamics. To our knowledge, moni-
toring the aggregation of particles under shear in micro-
scopes has been performed only for two-dimensional
surface ¯ow or at interfaces (Roussel, Camoin, & Blanc,
1989; Liu & Glasgow, 1997; Velikov, Durst, & Velev,
1998; Hansen & BergstroÈm, 1999a; Hansen, Malmsten,
BergenstaÊhl, & BergstroÈm, 1999b).
The aim of this work is to monitor the aggregation under
shear of latex particles coated with whey protein isolate in a
continuous phase with a complex behaviour in relation to
temperature dependence and shear thinning for example.
The continuous phase included gelatine, sucrose, salt and
water. The strategy was to monitor the aggregation not only
by viscosity measurements, but also to by direct observa-
tions with microscopy. The bene®t of this dual approach
was to enable a more accurate microstructural interpretation
of the viscosity measurements. The system was designed to
be similar to real systems, but also to have the advantages of
a model system. The particles were introduced to make it
possible to monitor the ¯ow in the microscope, as well as to
enlarge the result of the forces introduced by the ¯ow, on the
aggregates. The particles also needed to have a size of
.5 mm to allow the early stages in the aggregation to be
monitored at a low magni®cation. To look at the aggrega-
tion in a three-dimensional ¯ow ®eld, a Confocal Laser
Scanning Microscope, CLSM, was used, which made it
possible to depict a focal plane from inside the ¯owing
suspension.
2. Materials and methods
2.1. Materials
Whey protein isolate, (WPI) speci®cally LACPRODANw
DI-9224: .88% protein, ,6.0% moisture, ,4.5 ash,
,0.2% fat, ,0.2% lactose and ,2.1% minerals), was
obtained from MD Foods Ingredients Amba, Viby,
Denmark. Monodisperse polystyrene latex particles, of
5 mm, known as SOURCE 5RPC, were acquired from
Amersham Pharmacia Biotech, Uppsala, Sweden. The gela-
tine, (bloom strength 250, MW 118 kD) was obtained from
Extraco, Klippan, Sweden. The contrast agent Rhodamin 6
G was obtained from E. Merck AG Darmstadt, Germany.
2.2. Dispersion preparation
The dispersion of whey coated polystyrene latex particles
was prepared by dissolving these particles in a small volume
of a water/ethanol mixture, that contained 25% ethanol. To
reduce the amount of ethanol, the dispersion was stirred and
heated to 908C; this temperature was maintained for 15 min.
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151140
After cooling, sodium chloride, rhodamin dissolved in
water, and whey protein concentrate were dissolved
cautiously in the dispersion. Rhodamin was used to form a
stronger contrast in the microscope micrographs by colour-
ing the protein. The pH was then adjusted to 5.4 with small
amounts of HCl.
A solution containing 10% gelatine, 10% sucrose,
sodium chloride at three concentrations (0.05, 0.125
and 0.2 M), and water was prepared in a small beaker.
The solution was heated to 608C to dissolve the gelatine
and reduce the amount of dissolved gases in the solu-
tion. The gases were reduced to minimise bubbles later
on in the measurements. After 15 min, the temperature
was reduced to around 308C by placing the beaker with
the solution in a temperate water bath. As a ®nal step,
the particle dispersion was mixed with the same amount
of gelatine solution. The dispersion was regulated and
kept at 308C for 10 min during stirring to avoid gelation
of the gelatine. The concentration in the ®nal dispersion
is presented in Table 1.
The reason for adding sucrose to the dispersion was to
increase the density of the continuous phase, thereby to
reduce sedimentation of the particles during the measure-
ments. To raise the viscosity and thereby have more stable
¯ow conditions, gelatine was added to the particle system.
Another bene®t of introducing gelatine was that this would
automatically ®x the aggregates for the microscopy. The
®xation was brought about by gelation of the gelatine
when the temperature was lowered below 308C.
2.3. Experimental design
The experimental design was constructed with two design
variables: salt and shear rate. For the salt, two levels, 0.05
and 0.2 M, and a centre-point were chosen. The salt concen-
tration for the centre-point was 0.125 M. The different
concentrations of salt were used to investigate whether the
aggregation time could be in¯urnced this way (Molina-BolõÂ-
var et al., 1998). For the shear rate, three levels were chosen:
0.05, 0.5 and 5 s21. The middle level was also used to the
centre-point. The levels were chosen with respect to an
assumed logarithmic dependence between viscosity and
shear rate.
2.4. Viscosity measurements
The viscosity measurements were made in a Bohlin VOR
Rheometer (Bolin Rehology, Chichester, UK) equiped with
the Millennium software (Bholin Reologi, OÈ ved, Sweden).
The measuring system consisted of a cup and bob with
double-gap geometry. The inner and outer walls (diameter
21.7 and 27.4 mm) rotated during measuring while the bob
(diameter 23.9±24.9 mm) was ®xed. During the measure-
ments, the Taylor numbers were always in the region of
0.3±6 £ 1027; thus they were far below the critical number
of 3400 for Taylor vortices (Chandrasekher, 1961).
Next, 10 ml of the ®nal dispersion was poured into the
measuring system and covered with a thin layer of low
viscosity silicone oil, 10 mPas, to prevent evaporation
during the measuring. Steady continuous-shear viscosity
measurements were made during a temperature pro®le.
The pro®le started at 308C and rose to 608C with a slope
of 28C min21. The temperature was then held at 608C for
75 min. The three different shear rates were used for the
continuous-shear were 0.05, 0.5 and 5 s21.
2.5. Light microscopy and preparation
Directly after the viscosity measurements, the cup was
taken from the rheometer and gently tilted to the side so
that the silicone oil could be removed and the warm disper-
sion dripped onto four pre-heated slides and immediately
covered with a cover slip. For the ®rst slide a drop was taken
from the top of the cup. For the second, the drop was taken
from the middle, and for the two remaining, the drops were
taken from the very bottom of the cup. When the tempera-
ture fell below 308C, the gelatine in the drops gelled and
®xed the aggregates formed by particles. Three micrographs
from each drop were then taken at a 4 £ magni®cation with
a digitalised light microscope, Microphoto-FXA (Nikon
Corp., Tokyo, Japan); these images were stored digitally.
2.6. Direct microscopy under dynamic conditions
The direct microscopy studies under dynamic conditions
were performed in situ, directly in an unstopped ¯ow-®eld,
without any sample preparation for the microscope. The
microscopy was conducted on a Leica TCS 4D Confocal
Laser Scanning Microscope, CLSM, (Leica, Heidelberg,
Germany) in combination with a four-roll mill 4RM,
made by the workshop at ETH, EidgenoÈssische Technische
Hochschule, Zurich, Switzerland. The light source for the
CLSM was an argon-krypton laser with emissions maxima
at 488, 467 and 647 nm; the signal from the sample was
collected with a Nile red ®lter. The 4-RM, Fig. 1, has a
sample chamber consisting of a double-bottomed, ¯at,
open, dish with a diameter of 200 mm and a height of
20 mm. Four rolls, with a diameter of 45 mm and a height
of 18 mm, are symmetrically attached in a square, with a
centre distance of 55 mm. Two motors control the speeds of
the rolls; each motor controls a diagonally opposite pair of
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151 141
Table 1
Final contents of the prepared dispersion of whey coated polystyrene latex
particles
Material Amount
Particles 0.5 w/w%
WPI 1.0 w/w%
Gelatine 5.0 w/w%
Sucrose 5.0 w/w%
NaCl 0.05, 0.125, 0.2 M
Ethanol , 1 w/w%
rolls. In the middle, there is an optic window at the bottom
of the dish. Warm and cold water are circulated in the
hollow bottom of the dish to control the temperature.
By letting one pair of rolls rotate in one direction and the
other pair in the other direction, a hyperbolic ¯ow-®eld can
be generated in the centre of the chamber. This induces a
stagnation-point at the centre. By placing the scanning laser
at the stagnation-point where the ¯ow is slow, the dispersion
in motion can be observed. If the ¯ow is laminar, the ¯ow
speed is always low near the stagnation-point, irrespective
of the speed of the rolls. A typical ¯ow-®eld at the centre is
shown in Fig. 2. The investigations were made at a roll
speed of at 4 rpm. The dispersion was prepared in exactly
the same way as for the viscosity measurements, it
contained only half of concentration of particles and whey
protein. The sample chamber was ®lled with 300 g of the
dispersion and the surface was covered with a thin layer of
silicone oil. The total depth of the dispersion in the chamber
was 12 mm. The 4RM was placed under the CLSM so that
the laser, could scan near the stagnation point which had a
focal depth of 1.250 mm measured from the oil/dispersion
interface.
3. Results and discussion
3.1. Viscosity as a function of time and temperature
Measurements of viscosity of the dispersion were
continued for a period of 5000 s. During the ®rst
900 s, the temperature was ramped from 30 to 608C.
The temperature was then held at 608C until the end
of the measurements. The viscosity and temperature
were sampled for 10 s every 30 s to facilitate monitor-
ing their evolution. By following viscosity with time, it
is possible to gain understanding of aggregates and their
evolution. The aggregates are built up of the WPI
coated polystyrene particles; typical aggregates are
presented in Fig. 3.
The average curves for the time dependence of the visc-
osity for the three different shear rates 0.05, 0.5 and 5 s21 are
shown in Fig. 4. The curves are point to point averages for a
set of independent series. Salt concentration was varied
within the experimental set, but since no clear difference
was found for the three levels of salt contents, the averaged
series are averaged for these as well. The relatively large
error-bars at the start of the 0.05 s21 curve were due to
oscillation in the measurements at the beginning of the
time series at low shear rates.
All three curves start at different values but show the
same behaviour. Shortly after the start, the viscosities for
all of the measurements decrease to reach a local minimum.
An increase of the viscosity then starts and after a while the
increase levels out, and the viscosity reaches a stable value
towards the end.
The mutual relationship for viscosity, h, between the
curves with different shear rates, _g is well described by a
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151142
Fig. 2. A typical laminar ¯ow-®eld in the centre of the 4-roll mill.
Fig. 3. LM-micrograph of solution taken after the viscosity measurements.
(1% WPI, 0.125 M NaCl, shear rate 0.5 s21).
Fig. 1. A schematic picture over the 4-Roll Mill. (a)±(d) rolls, (e) sample
chamber, (f) optic window.
power law relationship such as:
h � m _g �n21�: �1�
The value of n is the shear thinning index n� 1 for a
Newtonian ¯uid with no shear thinning. The coef®cient m,
normally depends on temperature and concentration
(Macosko, 1994). Extracting the index n as a function of
time from the curves in Fig. 4 gives an initial value of n at
around 0.6. The shear thinning index, n, develops further
with time at around 500 s it reaches its maximum of 0.7,
after which the value starts to decrease. After 1500 s, n is
stable at around 0.3 and the value remains the same until the
end of the measurements. This means that the increase in
shear thinning behaviour of the dispersion does not start
until the temperature has reached approximately 608C.
For the curves in Fig. 4, it is possible to detect three
regions for which all three curves have the same behaviour.
The ®rst is the signi®cant decrease just after the start. The
second region is the rising one until the third region takes
over at the end, when the viscosity levels out and reaches a
stable value. The time intervals of the regions differ for the
three curves. Each of them is discussed in more detail
below.
To study the interior of the aggregates, LM-micrographs
at a high resolution were recorded by ¯uorescent techni-
ques: the protein was stained with rhodamin. In Fig. 5, the
whey protein stained with rodamin is bright, the darker areas
are the continuous phase, and the circular objects are latex
particles. The bright parts of in the micrograph can be seen
both between and edging the circular forms, hence whey
protein is present both as coated protein on the surface of
the latex particles and as free protein aggregates in the
dispersion.
While some of the latex particles clearly touch each other,
with only the coating protein between them, other particles
are solitary within the aggregate with just free protein aggre-
gates connecting them to the rest of the aggregate. This
supports the conclusion that both the coating protein and
the free protein aggregates contribute to the binding
between the particles. From the picture it seems reasonable
to believe that the free protein also forms small aggregates
with a spherical shape in the size order a tenth of the latex
particles. Both the coating protein and the free protein
aggregates should be rather similar and have the same prop-
erties, since hard proteins, such as b-laktoglobulin (the main
protein in whey) have a tendency to adsorb at interfaces
with few or no structural changes (Malmsten, 1998). That
free protein aggregates are essential for the aggregation has
also been found for the rather similar system consisting of
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151 143
Fig. 4. Viscosity shown as a function of time. The viscosity describes the aggregation of polystyrene spheres covered with WPI under shear for three shear
rates, £ 0.05, A 0.5, X 5.0 s21. The temperature increases from 30 to 608C during the ®rst 900 s and is then held at 608C to the end of the measurements.
Fig. 5. LM-micrograph of a small part of an aggregate taken with ¯ores-
cence technique. The light areas between the particles are WPI coloured
with Rhodamin.
drops of oil covered with sodium caseinate (Dickinson &
Golding, 1997a). In dispersions with free protein aggregates
present, the mechanism at early stages of the aggregation,
when two particles form a doublet, is expected to be deple-
tion ¯occulation (Dickinson & Golding, 1997a).
3.1.1. Onset of aggregation
The early decrease in viscosity exhibited by all three
curves characterises the initial region. Most of the decrease
takes place during the temperature ramp from 30 to 608Cduring the ®rst 900 s of the measurements. The ®rst region
of the three curves in Fig. 4 differs in time required for the
decrease. For the upper curve, with a shear rate of 0.05 s21,
the time from the start of measurements to the point of
minimum viscosity, is 490 s. For the middle curve, with
the shear rate 0.5 s21, the time is 720 s, and for the bottom
curve, at a shear rate of 5 s21, the time is 960 s. The devia-
tions of decrease also correspond to different temperatures.
For the top curve, the minimum viscosity is at 468C, for the
middle curve 548C and for the bottom curve 960 s which
means that the temperature has already reached the ®nal
608C.
LM-micrographs taken of a sample at the start tempera-
ture, 308C, Fig. 6(a), show that at the beginning the disper-
sion consists mostly of single latex particles, but that a small
number of doublets are also present. In LM-micrographs of
samples from the 4-RM, at 508C, after 600 s, Fig. 6(b), most
of the particles have formed aggregates consisting of two,
three or more particles. This indicates that the aggregation
starts between 30 and 508C, although the denaturation
temperature of sheared whey protein is around 708C (Ju,
Hettiarchchy, & Kilara, 1999; WalkenstroÈm et al., 1998a,b).
Many examples found in literature show that an aggrega-
tion or an increase in aggregate size increases the viscosity
(Dickinson & Golding, 1997b; Krieger, 1972; Campeanella,
Dorwood, & Singh, 1995; Barnes, 1994) and accurate physi-
cal models have been constructed (Sonntag & Russel, 1986,
1987; Potanin & Uriev, 1991; Potanin, 1991, 1992). This
phenomenon is due to entrapment of a certain volume of the
continuous phase within the aggregated structure; it
increases the effective hydrodynamic volume fraction of
aggregates, which in turn results in an increased apparent
viscosity. Nevertheless, at the same time as these small
aggregates are formed, the total viscosity is decreasing.
The fall in viscosity in the continuous phase is caused by
the well-known Arrenius type of decrease with temperature
and can be expressed by (Macosko, 1994):
h0 / eEh=RT �2�
Since time dependence is unlikely in the temperature
motivated Arrenius decrease in Eq. (2), and the aggregation
of whey protein is known to be triggered by temperature, the
differences in time to minimum viscosity have to be related
to the shear rate. Shear is known to affect aggregation by the
mechanism that greater shear increases the number of
encounters between particles (Walstra & Jenness, 1984)
which raises the rate of formation for new aggregates.
However, it also is known that shear decreases the overall
aggregate growth by introducing more and more break-up
and rearrangements (West et al., 1994). This combination is
the explanation to the shift in minimum viscosity time
towards later times for higher shear rates. Break-up and
rearrangements disrupt the early aggregation of particles
and the effect of this phenomenon increases with shear
rate. Hence, the rise in viscosity by the initial aggregate
build-up is weaker for large shear rates: for longer time
periods, the increase in viscosity at this early state is
outmatched by the decrease of the viscosity in the contin-
uous phase.
Consequently, the time to minimum viscosity would not
reveal the temperature when the aggregation starts, but it
does make it possible to determine the forces between the
particles. The longer time to minimum viscosity, together
with the LM-micrographs, point to the explanation that the
incipient aggregates are fragile in the beginning. The forces
between the particles, due to interactions between proteins
below 608C, have to be relatively weak or of the same order
as the forces acting on the particles due to shear. By measur-
ing and subtracting the viscosity decrease from the contin-
uous phase, it should be possible to obtain information on
when the aggregation starts. When this starting time is
known and, thus, the temperature, as well as the viscosity
for the system, it should be possible to calculate the forces
between the particles, by implementing the assumption that
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151144
Fig. 6. Two micrographs taken with the LM-microscope: (a) at start 308C,
and (b) after 600 s at 50C8. Samples are taken from the 4-RM.
when the aggregation starts, the forces between particles
should equal the drag forces from the surrounding shear.
To investigate and verify the effect of temperature,
experiments have been done with exactly the same
substances and procedure, but with a temperature ramp
between 30 and 408C (instead of the ramp between 30 and
608C). The result from such an experiment is shown in
Fig. 7. The inset curves show the time evolution of the
viscosity at a shear rate of 0.5 s21. In the inset ®gure, a
curve for a particle dispersion with whey protein, and a
temperature rise up to 608C, is compared with a similar
low-temperature dispersion with a temperature rise up to
just 408C. Both the 30±608C and the 30±408C curve start
with an initial decrease of the viscosity. For the 30±408Ccurve, the decrease is smaller than for the 30±608C curve.
After the initial fall, there is only a small rise with time
for the 30±408C curve in comparison with the large and
steep rise that occurs for the 30±608C curve. The differ-
ence in viscosity towards the end and the absence of the
steep rise lead to the conclusion that no aggregation of
particles occurs: hence, the onset of the protein aggrega-
tion is between 40 and 508C. In the literature it was found
that an interesting early aggregation of b-lactoglobulin,
the main protein in whey, occurred even in the tempera-
ture interval of around 408C (Stading & Hermansson,
1990).
3.1.2. Aggregation
The middle region of the viscosity curves in Fig. 4 is
distinguished by the large, steep increase in viscosity that
occurs after the minimum. In this region, the three curves
show more or less the same behaviour; the error-bars are
smaller than in the ®rst region. The viscosity rise for the
curves is relatively steep at ®rst, but levels out with time. At
around 2000 s, the increase has levelled off and viscosity
has reached a stable value. The steep rise of the viscosity
occurs mainly after the dispersion has reached 608C. The
increase in viscosity is probably due to an accelerated aggre-
gation of the whey coated particles. When the temperature
has reached 608C, the forces that hold the particles together
seem to be strong enough to let the aggregates overcome the
resistance even from high shear, and to grow.
To investigate whether or not the aggregation is due to the
presence of protein in the dispersion, experiments with and
without whey protein were made. The dispersion and set-up
were identical as to the centre-point condition, 0.125 M salt,
and the shear rate 0.5 s21, both with and without whey
protein, with the exception that additional salt was used to
compensate for the salt content of the WPI. Viscosity curves
for the measurements, both with and without whey are
shown in Fig. 7. The inset ®gure, displaying the time evolu-
tion of the viscosity at the shear-rate of 0.5 s21, shows that
the curves with and without whey differs even at the begin-
ning. The curve without WPI starts from a lower value and
has only a small decrease in the beginning. In contrast the
curve with whey has its steep increase around 1000 s, but
only a very small increase in viscosity is visible for the
curve without whey. This absence of the characteristic
steep rise in viscosity in the middle region leads to the
conclusion that no aggregation, or very limited aggregation,
occurs without protein. After rising both curves reach a
steady state. The three steady state viscosities are presented
as bars in Fig. 7. In this ®gure bars are given for all three
shear rates, and for the dispersion with whey as well as that
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151 145
Fig. 7. The bars shows steady-state viscosity as a function of shear-rate for three dispersions: S (black) whey 30±608C, B (dark grey) without whey 30±608C
and K (light grey) whey 30±408C. The inset ®gure shows the time evolution of the viscosity for the same dispersions at the shear-rate of 0.5 s21.
without whey protein. For all three shear rates, the bars for
the dispersions with and without whey are clearly different.
The same is true for a micrograph taken of a sample with-
drawn from the dispersion without whey at the end of the
measurements, Fig. 8, in compersion with another micro-
graph taken of a sample from the dispersion with whey, Fig.
3. In the micrograph from the dispersion with whey protein,
large aggregates consisting of 20±200 particles are found,
but in the micrograph without whey protein only singles,
doublets and some triplets are found.
Although the aggregation process could be driven and
controlled by temperature, shear could also have an in¯u-
ence. To investigate whether shear is critical to the process,
experiments were made with no shear present during the
®rst part. The temperature pro®le was the same, with the
increase from 30±608C during the ®rst 900 s. Three curves
with different times for the start of shear are compared in
Fig. 9. The shear, which starts at 0 s, at 900 s and at 1800 s,
has a rate of 0.5 s21. The two curves that start after 900 s and
1800 s correspond well to the curve for shear present from
the start. At the end, all three curves are more or less at the
same level, which indicates that shear and aggregation are
relatively independent. This supports the conclusion that the
in¯uence of increased encounters due to shear seems to be
minor and that the actual aggregation process seems to be
driven and controlled by the temperature. This is surprising,
since the in¯uence of shear on the aggregation is expected to
dominate over temperature by factors of 4, 40 and 400 for
the three shear rates 0.05, 0.5, and 5 s21 respectively
(Walstra & Jenness, 1984). The basis of this expectation
is a simple energy comparison between kinetic and thermal
energy, translated into encounters between the particles.
The theory uses quantities such as viscosity, h , temperature,
T, Boltzmann's constant, k, shear rate, _g , and primary parti-
cle size, d. The numbers above are calculated for latex
particle aggregation in a dispersion at 608C and with
numbers for the viscosity taken from the measurements
according to Eq. (3).
Shear
Temp/ d3h _g
2kT�3�
If the theory is tested with the assumption instead that
the free protein aggregates, i.e. the protein not coated on
the latex particle, control the aggregation, hence using the
protein as the primary particle, changes the result. For the
free protein aggregates, which is expected to be at least 10
times less than that of the latex particles (Langton &
Hermansson, 1996), the same energy comparison instead
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151146
Fig. 8. LM-micrograph of the dispersion taken after the viscosity measure-
ments: 0% WPI, 0.125 M NaCl, shear rate 0.5 s21.
Fig. 9. Viscosity as a function of time. The measurements were not started at the same time in the temperature pro®le: (A) At the beginning, 308C; ( £ ) after
900 s, 608C; and (X) at 1800 s from start, 608C. For all curves the shear rate is 0.5 s21.
demonstrates that the in¯uence of temperature is greater
by factors of 400, 40 and 4.
Both the aggregate build-up and the previously
mentioned shear thinning exponent, n, seem to occur during
the same time period and have an inverse appearance. When
the aggregates increases, the exponent decreases; it is there-
fore possible that the phenomena could be closely connected
in such a way that part of the shear thinning behaviour is due
to aggregate build-up.
3.1.3. Steady state viscosity
In the ®nal region of the viscosity pro®le, measurements
presented in Fig. 4, between 2000 and 5000 s, the viscosity
maintains steady state values. The steady state viscosity
values are 130 ^ 27 mPas for 0.05 s21, 23.1 ^ 2.7 mPas
for 0. 5 s21 and 5.16 ^ 0.40 mPas for 5 s21.
To analyse the phenomena underlying the differences in
viscosity, some dispersions were measured at other shear
rates after 5000 s. The additional viscosity measurements
were also performed at the shear rates of 0.05, 0.5 and
5 s21; the results are presented in Fig. 10. In the ®gure the
bars are shaded according to the shear rate their pro®les had
for the ®rst 5000 s. The shear rate before 5000 s is later in
the text is called pre-shear rate, the shear rates after 5000 s
are called measuring-shear rate. The grouping in the ®gure
are made according to the measuring-shear rates. For all
different pre-shear rates, the viscosity falls with increasing
shear rate. Also for all different measuring-shear rates, visc-
osity falls with increasing pre-shear rate. Comparing the
viscosities within the group with the same measuring
shear rate reveals a viscosity difference that is due to struc-
tural changes in the dispersion. If the viscosities with the
same pre-shear rates were compared instead, this would
reveal the differences due to shear thinning. Hence, the
conclusion is that the clear difference in viscosity value
for the differing shear rates is due to two separate effects:
the shear thinning in the continuous phase and the structural
changes of aggregates.
3.2. Viscosity measurements in a dispersion
When viscosity measurements are made on dispersions, it
is vital to check for any known problems that could affect
the measurements. For measurements as long as 5000 s,
settling could be a problem. Settling is usually omitted if
the time for an aggregate to migrate 10% of the height of a
concentric cylinder is more than the time needed for the
measurements (Macosko, 1994). The time can be obtained
by using of Stokes' law for an aggregate with a radius, a
(2.5 mm), in a concentric cylinder of the height h (55 mm),
under the in¯uence of gravity, g:
t10%h � 0:45hsh
rp 2 rs
��� ���a2g�4�
where r p (1050 kgm23) and r s (1080 kgm23) are the
density for the particles and the continuous phase disper-
sion, and h s is the viscosity of the dispersion. Taking
values for the properties according to the worst case
yields a maximum aggregate radius of 20 times the parti-
cle radius without a settling problem. In reality, more
realistic values allow the aggregate radii to be around
100 times larger than the particle radius. The conclusion
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151 147
Fig. 10. The bars show the steady state viscosity after different shear-histories. They are coloured according to the shear rate of their measurements for the ®rst
5000 s: ( £ ) (black) 0.05 s21, (A) (dark grey) 0.5 s21 and (X) (light grey) 5 s21. The grouping is according to the shear rate of the measurements after the time
5000 s. The inset ®gure shows the time evolution of the viscosity.
drawn from this is that only the very largest aggregate
will be affected. This does not seem to affect the viscos-
ity, since in Fig. 4 viscosity shows no tendency to
decrease with time in the steady state region.
A large aggregate can also alter the velocity ®eld
around itself, which can cause deviations in the viscosity
measurements. To omit this, a particle Reynolds number
should be smaller than 0.1 (Lin, Perry & Schowalter,
1970).
Rep � a2rs _g
hs
! 0:1 �5�
For the shear rates, _g , in the measurements, 0.05, 0.5 and
5 s21, and with the other values as above, Eq. (5) shows that
the ¯ow is not affected by small aggregates for any of the
shear rates. 0.05 s21 gives Re� 2 £ 1029, 0.5 s21 gives
Re� 1 £ 1027 and 5 s21 gives Re� 6 £ 106. However,
when the aggregates become large, e.g. 100 times a particle
radius, the Reynolds numbers become, for 0.05 s21
Re� 2 £ 1025, for 0.5 s21 Re� 1 £ 1023 and for 5 s21
Re� 6 £ 1022. This indicates that the ¯ow could deviate
for the highest shear rate. The result of a deviating ¯ow
®eld is only that the measured viscosity is shifted to a
slightly higher value than for undisturbed conditions. This
will have little effect on the conclusions: it is the evolution
of the viscosity that is of most interest here. Nevertheless
this phenomenon decreases the measured shear thinning
slightly for the highest shear rate, 5 s21.
Another known effect of particles during viscosity
measurements is the particle distribution in the gap (of the
cup and bob) due to wall effects. The particles have a
tendency to migrate towards the centre of the gap if the
Brownian motion cannot keep the particles uniformly
distributed. For equal distribution, K needs to be smaller
than 0.1 (Ho & Leal, 1974), in the following equation:
K � rp �y2a4
bkT, 0:1 �6�
where b is the gap (1.1 mm), T the temperature (608C), k
is Boltzmann's constant and �v is 12
vmax � R _g =2 where R
is the cup radius (12 mm). The value of K exceeds 0.1
for all shear rates when aggregates consist of ®ve parti-
cles or more. The varying distribution and the resulting
higher aggregate concentration in the middle of the gaps
can be expected, since the system was designed to
suppress the Brownian motion. The varying distribution
slightly alters the ¯ow pro®le by a small increase in
shear rate close to the walls and a small decrease in
shear rate at the middle of the gaps. Since this deviation
from ¯ow pro®le increases with shear rate, the actual
differences in shear rate are smaller than those between
the theoretical shear rates of 0.05, 0.5 and 5 s21. This
can have only a minor effect on the shear-thinning
index.
3.3. Direct observations under dynamic conditions
The actual build-up process has been followed in direct
observations under dynamic conditions and visual
evidence of the build-up is presented in Fig. 11, which
is a sequence of CLSM-micrographs taken in the 4-RM.
In these six views the particles are slightly oval because
of motion disturbance due to the continuous ¯ow and also
because the aggregates move during the scanning proce-
dure. The micrographs are from the centre of the 4-RM
where the ¯ow speed is automatically reduced; hence, at
the moment when the aggregates are exposed to lower
stresses than average. The aggregates have free mobility
in all directions and the surface effects are reduced, since
the micrographs are from a focal plane 1.25 mm below
the interface between the dispersion and the silicone oil.
In the ®rst image, a, only singles and occasional doublets
are present as large dots; this one was just after the start
when the temperature in the dispersion is 308C. There is
no notable change from micrograph a±b, where the
temperature is increased to 408C. There are still only
singles and some doublets which is expected since the
protein does not aggregate below 408C. This is also in
good agreement with the explanation that the viscosity
decreases in the beginning are due to viscosity changes
in the continuous phase; they are not due to any structural
changes among the particles. In micrograph c, at 508C, a
weak tendency towards aggregation is visible, since a
number of loose aggregates consisting of 5±10 particles
are visible. In micrograph d, taken after 900 s, when the
temperature had just reached 608C, the aggregates are
larger than in the previous micrograph and they consist
of 10±30 particles. Between the aggregates there are still
many singles. After 1200 s, the aggregates have become
even larger and consist of 50±200 particles. In the ®nal
micrograph, f, the aggregates are of the same order of
magnitude as in e. Although singles are still present,
most of the particles are incorporated in the large aggre-
gate. The evolution of the size of the large aggregates in
the micrograph sequence is in good agreement with the
viscosity pro®le in regions 2 and 3. The good agreement
between the micrographs and pro®le shows that the new
combination of the 4RM together with the CLSM is a
powerful tool for studying the aggregation mechanism
in a continuous undisturbed 3-D ¯ow. The new tool is
useful for situations where the changes in viscosity due
to the structural changes are smaller than the viscosity
changes in the continuous phase. It is also useful when
the aggregates are fragile and the preparation could affect
the aggregates or when the shapes of the aggregates are in
dynamic equilibrium with the undisturbed ¯ow. When
computers become even faster than now, the option to
perform faster scans will become available, so that a
higher resolution becomes possible in the time dimension.
Then, with this technique, it would be possible to follow
smaller and/or faster movements within the dispersion.
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151148
One can also combine the 4-RM and the CLSM with
image analysis and perform quantitative measurements
in the micrographs. This can shift the set-up further,
from a qualitative microscope towards a measurement
tool for dynamic in situ condition.
Acknowledgements
This work is a part of the LiFT program (Future Tech-
nologies for Food Production), ®nanced by the SSF (Swed-
ish Foundation for Strategic Research).
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151 149
Fig. 11. CLSM-micrographs of the dispersion taken during heating and ¯owing: 0.15% particles, 0.30% WPI, 0.125 M NaCl; (a) 308C, 0 s; (b) 408C, 300 s; (c)
508C, 600 s; (d) 608C, 900 s; (e) 608C, 1200 s; (f) 608C, 2000 s.
References
Al-Jabari, M., Weber, M. E., & van de Ven, T. G. M. (1996). Modeling
®nes elutriation from a sprouted bed of pulp ®bers. Chemical Engineer-
ing Communications, 148-150, 465±476.
Amirtharajah, A., & O'Melia, C. R. (1990). Coagulation processes: desta-
bilisation, mixing, and ¯occulation. In F. W. Pontius, Water quality and
treatment (pp. 268±365). New York: McGraw-Hill, Inc.
Barnes, H. A. (1994). Rheology of emulsionsÐa review. Colloids and
surfaces A: Physicochemical and Engineering Aspect, 91 (1), 89±95.
Campeanella, O. H., Dorwood, N. M., & Singh, H. (1995). A study of the
rheological properties of concentrated food emulsions. Journal of Food
Engineering, 25 (3), 427±440.
Chandraskhar, S. (1961). Hydrodynamic and hydromagnetic stability,
London: Oxford University Press.
Colowick, S. P., & Kaplan, N. O. (1987). Methods in enzymology, San
Diego: Academic Press Vol. 136.
Dickinson, E., & Golding, M. (1997a). Depletion ¯occulation of emulsions
containing unadsorbed sodium caseinate. Food Hydrocolloids, 11 (1),
13±18.
Dickinson, E., & Golding, M. (1997b). Rheology of sodium caseinate
stabilised oil-in-water emulsions. Journal of Colloid and Interface
Science, 191 (1), 166±176.
Dickinson, E. (1999). Adsorbed protein layers at ¯uid interfaces: Interac-
tions, structure and surface rheology. Colloids and Surfaces B: Bioin-
terfaces, 15 (2), 161±176.
Hansen, P. H. F., & BergstroÈm, L. (1999a). Perkinetic aggregation of
alkoxylated silica particles in two dimensions. Journal of Colloid and
Interface Science, 218 (1), 77±87.
Hansen, P. H. F., Malmsten, M., BergenstaÊhl, B., & BergstroÈm, L. (1999b).
Orthokinetic aggregation in two dimensions of monodisperse and bidis-
perse colloidal systems. Journal of Colloid and Interface Science, 220
(2), 269±280.
Higashitani, K., & Iimura, K. (1998). Two-dimensional simulation of the
break-up process of aggregates in shear and elongational ¯ows. Journal
of Colloid and Interface Science, 204 (2), 320±327.
Ho, B. P., & Leal, L. G. (1974). Inertial migration of rigid spheres in two
dimensional unidirectional ¯ows. Journal of Fluid Mechanics, 65, 365±
400.
Horbett, T. A. (1982). In S. L. Cooper & N. A. Peppas, Biomaterials:
interfacial phenomena and applications ACS. Advances in chemistry
series (p. 233), Vol. 199. Washington, DC: American Chemical
Society.
Isaksson, P., Rigdahl, M., Flink, P., & Forsberg, S. (1998). Aspects of the
elongational ¯ow behaviour of coating colours. Journal of Pulp and
Paper Science, 24 (7), 204±209.
Ju, Z. Y., Hettiarchchy, N., & Kilara, A. (1999). Thermal properties of
whey protein aggregates. Journal of Dairy Science, 82 (9), 1882±
1889.
Krieger, I. M. (1972). Rheology of monodisperse latices. Advances in
Colloid and Interface Science, 3, 111±136.
Langton, M., & Hermansson, A. -M. (1996). Image analysis of particulate
whey protein gels. Food Hydrocolloids, 10 (2), 179±191.
Lin, C. -J., Perry, J. H., & Schowalter, W. R. (1970). Simple shear ¯ow
round a rigid sphere: inertial effects and suspension rheology. Journal
of Fluid Mechanics, 44, 1±17.
Liu, S. X., & Glasgow, L. A. (1997). Aggregate disintergration in turbulent
jets. Water, Air and Soil Pollution, 95, 257±275.
Macosko, C. R. (1994). Rheology-principles, measurements and applica-
tions, New York: VCH Publishers Inc.
Malmsten, M. (1998). Formation of adsorbed protein layers. Journal of
Colloid and Interface Science, 207 (2), 186±199.
Mathur, S., Singh, P., & Moudgil, B. M. (2000). Advances in selective
¯occulation technology for solid-solid separations. International Jour-
nal of Mineral Processing, 58 (1-4), 201±222.
Molina-BolõÂvar, J. A., Galisteo-GonzaÂlez, F., & Hidalgo-AÂ lvarez, R.
(1998). Cluster morphology of protein-coated polymer colloids. Jour-
nal of Colloid and Interface Science, 208 (2), 445±454.
Molina-BolõÂvar, J. A., Galisteo-GonzaÂlez, F., & Hidalgo-AÂ lvarez, R.
(1999a). Colloidal aggregation in energy minima of restricted depth.
Journal of Chemical Physics, 110 (11), 5412±5420.
Molina-BolõÂvar, J. A., Galisteo-GonzaÂlez, F., & Hidalgo-AÂ lvarez, R.
(1999b). Development of a high sensitivity lgG-latex immunodetection
system stabilised by hydration forces. Polymer International, 48 (8),
685±690.
Ortega-Vinuesa, J. L., Molina-BolõÂvar, J. A., & Hidalgo-AÂ lvares, R. (1996).
Particle enhanced immunoaggregation of F(ab 0)2 molecules. Journal of
Immunological Methods, 190 (1), 29±38.
Potanin, A. A., & Uriev, N. B. (1991). Microrheological models of aggre-
gated suspensions in shear ¯ow. Journal of Colloid and Interface
Science, 142, 385±395.
Potanin, A. A. (1991). On the mechanism of aggregation in shear ¯ow
of suspensions. Journal of Colloid and Interface Science, 145,
140±157.
Potanin, A. A. (1992). On the microrheological modelling of aggregat-
ing colloids. Journal of Dispersion Science and Technology, 13,
527±548.
Quesada, M., Puig, J., Delgado, J. M., Peula, J. M., Molina, J. A., &
Hidalgo-AÂ lvarez, R. (1997). A simple kinetic model of antigen-anti-
body reactions in particle-enhanced light scattering immunoassays.
Colloids and Surfaces B: Biointerfaces, 8 (6), 303±309.
Roussel, J. -F., Camoin, C., & Blanc, R. (1989). Image analysis of kinetics
of aggregation. Journal de Physique France, 50, 3259±3267.
Russel, W. B. (1980). Review of the role of colloidal forces in the rheology
of suspensions. Journal of Rheology, 24 (3), 287±317.
Schokker, E. P., & Dagleish, D. G. (1998). The shear-induced destabilisa-
tion of oil-in-water emulsions using caseinate as emulsi®er. Colloids
and Surfaces A: Physicochemical and Engineering Aspects, 145 (1-3),
61±69.
Sonntag, R. C., & Russel, W. B. (1986). Structure and break-up of ¯ocs
subjected to ¯uid stresses I. Shear experiments. Journal of Colloid and
Interface Science, 113, 399±413.
Sonntag, R. C., & Russel, W. B. (1987). Structure and break-up of ¯ocs
subjected to ¯uid stresses II. Theory. Journal of Colloid and Interface
Science, 115, 378±389.
Stading, M., & Hermansson, A-M. (1991). Large deformation proper-
ties of (b-lactoglobulin gel structures. Food Hydrocolloids, 5 (4),
339±352.
Stading, M., & Hermansson, A-M. (1990). Viscoelastic behaviour of
(b-lactoglobulin gel structures. Food Hydrocolloids, 4 (2), 121±135.
Stading, M., Langton, M., & Hermansson, A. -M. (1993). Microstructure
and rheological behaviour of particulate b-lactoglobulin gels. Food
Hydrocolloids, 7 (3), 195±212.
Taylor, G. I. (1934). The formation of emulsions in de®nable ®elds of ¯ow.
Proceedings Royal Society, A146, 501±523.
Taylor, S. M., & Fryer, P. J. (1994). The effect of temperature/shear on the
history of thermal gelation of whey protein concentrates. Food Hydro-
colloids, 8 (1), 46±61.
WalkenstroÈm, P., & Hermansson, A. -M. (1996). Effects of shear on pure
and mixed gels of gelatine and particulate whey protein. Food Hydro-
colloids, 12 (1), 77±87.
WalkenstroÈm, P., Nilsen, M., Windhab, E., & Hermansson, A. -M. (1999).
Effects of ¯ow behaviour on the aggregation of whey protein suspen-
sions. Pure and mixed with xanthan. Journal of Food Engineering, 42
(1), 15±26.
WalkenstroÈm, P., Panighetti, N., Windhab, E., & Hermansson, A. -M.
(1998a). Effects of ¯uid shear and temperature on whey protein
gels, pure or mixed with xanthan. Food Hydrocolloids, 12 (4),
469±479.
WalkenstroÈm, P., Windhab, E., & Hermansson, A. -M. (1998b). Shear-
induced structuring of particlate whey protein gels. Food Hydrocol-
loids, 12 (4), 459±468.
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151150
Walstra, P., & Jenness, R. (1984). Diary chemistry and physics, New York:
John Wiley & Sons.
van de Ven, T. G. M. (1982). Interactions between colloidal particles in simple
shear ¯ow. Advances in Colloid and Interface Science, 17, 105±127.
Velikov, K. P., Durst, F., & Velev, O. D. (1998). Direct observation of
dynamics of latex particles con®ned inside thinning water-air ®lms.
Langmuir, 14 (5), 1148±1155.
West, A. H. L., Melrose, J. R., & Ball, R. C. (1994). Computer simulations
of the break-up of colloid aggregates. Physical Review E, 49 (5), 4237±
4249.
L. Hamberg et al. / Food Hydrocolloids 15 (2001) 139±151 151