Full Papers

Product Design in Food Industry Using the Example

of Emulsification
By Heike P. Schuchmann* and Helmar Schubert

Designing a product according to the consumers' needs is one of the key targets in food industry. Having once defined the target
product properties, it is of imperative importance to know which microstructure of the product is required (property function)
and how the target microstructure can then be determined by process parameters (process function). Many food products are
emulsions, either during production and/or in the moment of distribution and consumption. This review intends to depict the
design of emulsions according to target needs. Even if we concentrate on typical food emulsions as an example, the concept can
be widely applied, since emulsification in other branches of industry is governed by the same principles.

1 Introduction sense organs. The result of sensorial judging by human beings

As all branches of industry, food industry intends to gain
money by selling products. Most food products are offered to
individual consumers. By choosing a product and buying it at a
certain price the consumer defines its market value. To
maximize sales, one of the key targets in food industry
therefore is designing the products according to the con-
sumers' needs. This review intends to present a concept of
designing food emulsions. First, the principles of product
design will be presented. Then, we will concentrate on
emulsion and emulsification basics. Based on these basics,
emulsion property and emulsification process functions will
be given. While property functions are rarely known for food,
the process functions in emulsification have been widely
investigated within the last years. Using these functions, three
examples of emulsion product design will finally be given for
products of different functionalities.
2 Product Design in Food Industry
In principle, product design involves three steps: defining
the target product properties, establishing the product
property function and, finally, determining the process
functions.
The very first step of designing a product is to define the
target product properties. The definition of target properties of
food products usually is not done by a single person only, but
by several groups: for sure, the consumer wants a special
quality. However, consumers do not judge a food product by
measuring quality parameters, but by using their personal
±
[*] Dr.-Ing. H. P. Schuchmann-Karbstein (author to whom corresponding
should be addressed), Institute of Process Engineering, Federal
Research Center for Nutrition, Haid-und-Neu-Str. 9, D-76131 Karlsruhe;
Prof. Dr.-Ing. habil. H. Schubert, Institute of Food Process Engineering,
Karlsruhe University, D-76128 Karlsruhe, Germany.
rarely depends on objective measurement, but on several
additional factors, such as culture, education, or personal
experience. Thus, ªtheº target quality does not exist.
Considerable sums are therefore spent on market research,
its results being the base for the definition of a target quality
from the consumers' point of view. Besides these studies,
marketing has to take into consideration packaging, logistics,
overall company strategy, as well as cost aspects. In addition, at
this early state of product design, R&D process engineers
should already consider consequences for raw materials,
product handling and processes, and production engineers
should try to foresee future production problems. Taking into
consideration all these different aspects, the target product
will finally be defined by measurable (sensorial and physico-
chemical) quality parameters.
The second step in product design is to determine the
property function, which was suggested by Rumpf [1]. The
property function gives the functionality between product
microstructure and its properties:
product property = f (microstructure)
Property functions are widely investigated by food rheology
specialists [2]. In the case of foods, many property functions
are known only approximately, and highly depend on product
recipe. Thus, they have to be determined anew for each
product composition.
Knowing the product microstructure, the process functions
[3,4] have finally to be determined. Process functions
summarize the impact of process parameters on the product
microstructure:
micro-structure = f (process parameters)
Knowing these functions, the required microstructure can
be adjusted by process parameters. The determination of

Eng. Life Sci. 3 (2003) 2, Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1618-0240/03/0202-0067
0930-7516/03/0202-0067 $ 17.50+.50/0
$ 17.50+.50/0 67
 Full Paper
process functions in food industry usually is the target of
process engineers and food technologists. Process functions
enable engineers to choose the best type of machine, to
optimize process parameters, and to scale up the process.
3 Emulsion and Emulsification Basics
3.1 Emulsions
Emulsions are part of the product range in the chemical or
petrol industry as well as in the pharmaceutical, cosmetic and
food industries. In food industry, typical emulsion products are
milk, butter, mayonnaise, spreads, sauces, and dressings.
An emulsion is a disperse system of at least two immiscible
liquids, e.g. water and oil (see Fig. 1). One of these phases is
finely dispersed and forms droplets in the other one.
Depending on the droplet phase, oil-in-water- (o/w-) type
and water-in-oil- (w/o-) type emulsions exist. In rare cases, the
droplet phase itself is an emulsion. This review will concen-
trate on the simple emulsion type, consisting of two liquids,
only.
w o o w
oil-in-water (o/w) water-in-oil (w/o)
emulsifier: small molecular high molecular
weight: surfactants weight, e.g. protein
hydrophilic lipophilic
surfactant only protein only protein and surfactant
O 3.3 Industrial Emulsification
W In order to produce emulsions, several processes may be
Figure 1. Emulsions: basic structure.
These systems are thermodynamically unstable [5]. If not
stabilized, the droplets tend to flocculate and coalesce, or
sediment and cream, respectively. Other instabilities, such as
Ostwald ripening or phase inversion, are known as well. At
worst, an emulsion will break, i.e., the two phases will separate
completely. A product becoming unstable will loose its quality
within a short period of time and, thus, cannot be commercia-
lized.
Therefore, even in natural emulsion-based products, addi-
tional molecules are found (e.g. lecithin and proteins in egg
yolk or milk). Industry adds natural and artificial emulsifiers
and thickeners. Emulsifiers are amphiphilic molecules that
adsorb at the droplets' interfaces (see Fig. 1), thus stabilizing
them against flocculation and coalescence. Small-molecule
surfactants (in food industry, for example, fatty acids or their
esters, as e.g. Tweens), as well as high molecular weight
68 Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
emulsifiers (in food industry these are most often proteins) are
used. Depending on the molecular structure of an emulsifier,
different adsorption mechanisms are found, both depicted in
Fig. 1. Besides stabilizing an emulsion, the emulsifier often
influences product properties, such as its viscosity. This will be
shown in chapter 5.
Thickeners are high-molecular weight molecules that are
solved in the continuous phase, thus enhancing its viscosity.
They stabilize emulsions by slowing down droplet mobility.
Flocculation, sedimentation or creaming, and coalescence are
either slowed down or inhibited at all. Typical thickeners used
in food industry are (modified) starches and proteins. Since
they increase continuous phase viscosity, they do not only
stabilize the emulsion, but also change its properties (see
chapter 5).
3.2 Emulsion Properties
Important emulsion quality properties are the physical and
microbiological stability, rheological properties (as consis-
tency, viscosity, texture, spread-ability), sensorial qualities
(such as color, taste, and mouth-feel), as well as the
distribution of active ingredients, their dissolution rate and
bioavailability. They are influenced on the one hand by the
formula (type and concentration of main phases and
additives). On the other hand, the droplets' size distribution
has a big impact on the properties of an emulsion. Often, the
formula of an emulsion is already given (by recipe and/or
legislation). Thus, the droplet size distribution might be the
only possibility for engineers to influence the product quality.
Most often, the main process goal therefore is to produce a
required droplet size distribution for a given recipe.
applied. Most often, droplets are formed and disintegrated
mechanically. Besides, emulsions may also be formed, e.g., by
precipitation. However, nonmechanical methods are rarely
used in industrial production. Therefore, this review will
concentrate on mechanical emulsification methods (see Fig. 2).
Rotor-stator machines consist of a rotating and a fixed
machine part. Hundreds of different rotor and stator
geometries are available on the market [6]. The simplest
rotor-stator machine is a vessel with a stirrer, which is used to
produce emulsions batchwise or quasi-continuously. A stirred
vessel has several advantages in emulsification: in a vessel,
different process steps (as e.g. heating for pasteurization, or
cooling before packaging) can be combined with the
emulsification step. It is relatively easy to produce small
quantities of several recipes within one machine, as well.
Therefore, stirred vessels are widely used in food industry,
although there are some drawbacks. In stirred vessels, the
power density is low, and broadly distributed. Therefore, a
1618-0240/03/0101-0068 $ 17.50+.50/0 Eng. Life Sci. 3 (2003) 2

rotor-stator-machines, e.g. colloid mill high-pressure homogenizers
ultrasonic waves microporous membranes
Figure 2. Emulsification machines (mechanical emulsification).
long residence time and, thus, emulsification time is required,
often resulting in broad droplet size distributions. Small mean
droplet diameters (as e.g. < 1 lm) can rarely be produced in
stirred vessels. If processes are driven at higher temperatures
(e.g., to lower viscosities or to melt ingredients), undesired
byproducts may be formed.
In order to solve these problems, the disruption zone has to
be small and well defined, and the power density should be
increased. This is realized in continuously working rotor-
stator machines, such as colloid mills, or toothed-disk
dispersing machines. In a colloid mill, the rotating part is a
truncated cone, most often toothed. It rotates within a stator,
also toothed. In toothed-disk dispersing machines, rotor and
stator consist of one or several disks having pins or teeth of
different design. Rotor-stator-machines are relatively easy to
handle and of low cost. They can be operated at throughputs
from about 50 L/hr to several tons per hour. Droplet
disruption is mainly due to turbulent flow within the
dispersing zone between rotor and stator [6].
High-pressure homogenizers have been traditionally used in
milk industry. These machines are operated continuously at
throughputs from 1 to several 1000 L/hr. They consist
essentially of a high-pressure pump, and a homogenizing
nozzle. The high-pressure pump creates the energy, which is
then used by the nozzle to disintegrate the droplet phase. The
design of the homogenizing nozzle influences the flow pattern
of the emulsion in the nozzle and hence the results of droplet
disruption [7]. Recent developments in high-pressure homo-
genizing concentrate therefore on nozzle design. Examples of
new homogenizing nozzles are opposing jets (Microfluidizer
[8]), jet disperser [9], and a simple orifice valve [10]. Droplet
disruption in high-pressure homogenizers predominantly is
due to inertial forces in turbulent flow and shear forces in
laminar elongational flow [7]; cavitation may also be found
[11,12].
The trend in emulsification is to produce emulsions of very
fine droplets (< 1 lm) and narrow size distributions at low
energy input. Continuous emulsification using ultrasonic
waves, microporous membranes, and microchannel systems
are to be mentioned in this context.
Eng. Life Sci. 3 (2003) 2, Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper
In emulsification by means of ultrasonic waves (waves of
frequency > 18 kHz), droplets are disrupted in zones of locally
high turbulence caused by sound-induced cavitation. A
batchwise operation of a small scale is well known for
laboratory applications whenever fine droplets are required
in low viscous systems [5]. New publications now demonstrate
the possibility to produce finely dispersed emulsions by means
of ultrasonic waves continuously, as well [13,14]. Continuous
applications require a flow chamber of special design into
which the ultrasound waves are introduced. Due to the limited
power of sound inducers, there are technical limits for very
high tonnages (e.g. > 1000 L/hr).
Droplet disintegration usually is done by tensions acting on
droplet interfaces via the continuous phase. In order to
produce the required flow pattern in the continuous phase,
high energy has to be applied. Only a negligible part of the
energy (< 1 %) is used for droplet disintegration, while most of
the energy is transferred into heat. It is therefore more
efficient to produce droplets of the required size directly by a
droplet-forming process. This could be realized, e.g., by
pressing the future droplet phase through pores of small
diameter and dispersing the droplets into the continuous
phase. Industrially, this can be realized by using microporous
membranes [15,16]. In membrane emulsification, the process
starts at a disperse-phase content of zero. During the process,
the disperse-phase content is then enhanced to the required
value by pressing more disperse phase through the membrane
pores. This is realized by recirculation of the emulsion.
Recirculation, in turn, decreases energy efficiency. Alterna-
tively, preemulsions may be pressed through the membrane
pores [17], and phase inversion during membrane emulsifica-
tion may be induced [18]. Thus, high quantities of emulsions
(several tons per m2 of membrane surface and hour) having a
high disperse-phase content (e.g. 50±80 %) may be produced
by a single pass through the membrane. Membrane emulsi-
fication is reported to produce small droplets (< 3 lm) at a very
narrow droplet size distribution and low energy input [15,17].
Investment and manufacturing costs are less than for high-
pressure homogenizing [16]. By microchanneling [19], emul-
sions of extremely narrow droplet size distributions may be
produced. However, this process being under development is
far from industrialization until now.
4 Emulsion Property Functions
As already stated, property functions for emulsion products
highly depend on product recipe. Thus, they have to be
determined anew for each product recipe. Therefore, we will
restrict ourselves to one example. In Fig. 3, the viscosity of a
salad-cream type emulsion (60 % of vegetable oil in water) is
depicted as a function of the type and concentration of the
emulsifier, as well as the mean droplet size.
The addition of a high-molecular weight emulsifier (such as
Lacprodan-60, a spray-dried whey protein, mainly consisting
of b-lactoglobulin) increases the product viscosity consider-
0930-7516/03/0202-0069 $ 17.50+.50/0 69
 Full Paper

1.0 produced, hundreds of these diagrams have to be generated

Microfluidizer®
viscosity η / Pa s at 100 s-1

60 % vegetable oil in water before being able to optimize the process, and to scale it up.
0.8 c Lac = 60 g/l Process functions, however, should be independent of the
c = 0 g/l
Tween 80
1,2 g/l
equipment or process parameters.
0.6
3,5 g/l In order to evaluate the effectivity of droplet disruption, the
7,5 g/l volumetric specific energy Ev was proposed by Karbstein [6]
0.4 15 g/l
as an independent parameter to describe continuous emulsi-
35 g/l
0.2
fication processes for all machine types and process param-
c = 0 g/l
c
Lac
= 150 g/l eters1).
Tween 80
0.0
E P  V
tres
0 1 2 3 4 5 Stang, 1998
EV ˆ ˆ ˆ Dph ˆ P (1)
mean droplet diameter x3,2 / µm _
V V
Figure 3. Example of a property function for a food emulsion: Viscosity g of the Herein, the power density is defined as
emulsion as a function of its mean droplet size x3,2 (Sauter diameter) and the
emulsifier type and concentration, respectively [10]. P
Pv ˆ (2)
V
ably compared to a low-molecular weight surfactant (e.g.
Tween 80, a sorbitane fatty acid). For a given recipe, however, Herein, E is the energy applied by the machine, V the
viscosity decreases with increasing mean droplet diameter. volume of the emulsion, P the power applied by the machine,
This effect is even more marked for the high-molecular-weight V_ the volume flow rate of the emulsion, Dph the homogenizing
emulsifier. pressure (in case a high-pressure homogenizer is used), Pv the
The parameter droplet size' thus offers a good possibility to mean volume power density of the machine, and tres the mean
influence rheological and sensorial quality parameters, such as residence time of the emulsion within the dispersing zone of
viscosity, creaminess or mouth-feel, even if the recipe is the machine.
already given by raw material or legislative restrictions. Using The specific energy summarizes the influences of machine
the property function (as in Fig. 3), the mean droplet size may and process parameters into the energy produced by these
be determined that is responsible for a target viscosity. This processes and applied to a certain volume of emulsion. Thus,
mean droplet size characterizes the microstructure, that now only one diagram per recipe/formula is required (see Fig. 4b),
has to be realized by an emulsification process. The necessary giving one curve per formula [6]. In a double logarithmic
equipment has to be chosen and the process to be designed. diagram, this curve is a straight line.
The specific energy also depicts the influences of power
density and residence time on the disruption process. The
5 Process Functions in Emulsification influence of the residence time is always found in continuous
emulsification processes of short residence times (tres << 1 s)
5.1 Independent Parameter for Describing Emulsification [20]. Similar dependencies are also found in wet grinding
Processes processes of solids [e.g. 21].
In batch processes, however, the residence time of the
Emulsification results are usually given as a mean droplet product in the emulsification machine is in the order of
diameter in dependency of a chosen machine type, and one minutes. Here, both, the power density Pv [5], and the
or more selected process parameters (such as rotor speed), residence time tres influence the mean droplet size. Since the
and/or machine parameters (e.g. dispersing zone design). dependencies on power density and residence time follow
These diagrams are then given for a selected recipe/formula different mathematical functions, they cannot be summarized
(e.g. Fig. 4a). Even if only one product formula has to be in the energy density, as for continuous emulsification
processes. Only for very long residence times tres ! ¥, no
30 % rape seed oil in water, emulsifier: LEO-10 influence of the residence time on the final emulsification
toothed discs dispersing machine
result is found.
5 10
1 r/s
36 mm (a) (b)
4 2 r/s
36 mm 5.2 Theory of Droplet Disruption
x3,2 / µm

x3,2 / µm

3 1 r/s
48 mm 1
2 r/s
rotor
2 48 mm speed Droplets are disrupted, if they are deformed over a period of
1 number or
diameter of r/s-discs
time tdef that is longer than a critical deformation time tdef,cr,
0 0.1 and, if the deformation (described by the Weber number, We)
0 4.000 12.000 20.000 106 107 108 exceeds a critical value, the critical Weber number Wecr [5]:
rotor speed n / min-1 spec. energy Ev / (J/m3)

Figure 4. Mean droplet diameter x3,2 as a function of a process parameter (rotor

speed) and as a function of specific energy EV (r/s: rotor-stator system or ±
toothed-disk pair, respectively). 1) List of symbols at the end of the paper.

70 Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1618-0240/03/0101-0070 $ 17.50+.50/0 Eng. Life Sci. 3 (2003) 2

We > Wecr and tdef > tdef,cr
with
r time are in the same order of magnitude (0.25 ... 0.4), they may
We ˆ
pc
4
c
pc ˆ
x
gd
tdef;cr ˆ
r pc
Herein, r are the external tensions responsible for
deformation, pc the capillary pressure, c the interfacial
tension, x the droplet diameter, and gd the droplet viscosity.
Droplet disruption mechanisms are described in detail in
the literature [e.g. 5,6,10,22,]. They will be summarized here,
only. Droplet disruption due to laminar shear flow has been
widely investigated (e.g. [5,23]). However, it is restricted to a
narrow range of viscosity ratios (between the disperse and
continuous phases gd/gc for single droplet disruption, or
between the disperse phase and the emulsion gd/ge for
emulsions, respectively [24]). Pure laminar shear flow is
realized in special machine constructions, only, most of them
being on lab scale.
In turbulent flow, droplets are deformed and disrupted
mostly by inertial forces that are generated by energy-
dissipating small eddies. Due to internal viscous forces
droplets try to gain their initial form and size [25]. Two
dimensionless numbers, the turbulent Weber number Weturb
and the Ohnsorge number Oh, characterize the tensions
working on droplets in deformation and break-up:
2=3
C2
Pv
rc 1=3
x5=3
Weturb ˆ
c
gd
Oh ˆ  1=2
c
rd
x
Herein, C is a constant, Pv is the volumetric power density, rc
the continuous phase density, and rd the droplet density,
respectively. For very long deformation times, limits can be given
for the droplet diameter after disruption in turbulent flow [26]:
For droplet phases of low viscosity:
2=5
xmax|Oh®0µrc 1=5 c3=5 PV
For highly viscous droplet phases:
1=4
xmax|Oh®¥µ rc 1=2 gd3=4 PV
Experiments depict that the droplet phase viscosity is
negligible for gd < 10 mPa s [6,25].
In continuous emulsification processes, residence time is in
the order of milliseconds to tenths of a second. Thus, an
additional dependency on residence time tres is found [20]:
Eng. Life Sci. 3 (2003) 2, Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper
0:3
(3) xµtres (11)
Since the dependencies on power density and residence
(4) be combined into a dependency on specific energy Ev [6]. This
results in:
(5) x µ EV
 0:25 ::: 0:4

gd 0 ::: 0:75
(12)
(6) The higher the specific energy applied during emulsifica-
tion, the smaller the droplets. In addition, smaller droplets are
found for lower viscosities of the droplets.
A similar dependency is also found for droplet disruption
due to sound-induced cavitation (ultrasonic emulsification)
[13,14]. This depicts that similar droplet break-up mechanisms
are effective in high turbulent flow and sound-induced
cavitation.
In laminar elongational flow, however, no dependency of
the mean droplet diameter on the droplet viscosity is found
[5,10,22,23]. The mean droplet diameter after disruption in
laminar elongational flow is given by [7,22]:
1
x µ EV
 (13)
In membrane emulsification, droplets are not disintegrated
but formed by pressing them through pores into the
continuous phase. Several forces (as e.g. interfacial forces,
resisting forces, impulse forces) are working on the droplets.
Process parameters determining the final droplet size are the
continuous-phase wall stress, the diameter of the membrane
pores, and the transmembrane pressure difference [16,27,28].
For a given membrane (i.e. pore sizes), the sum of
(7) transmembrane pressure difference and wall stress, being
proportional to the specific energy input, governs the
emulsification result. Thus, membrane emulsification may
(8) also be compared to other emulsification processes based on
the specific energy concept. No dependency on the disperse-
phase viscosity is found for droplet formation through
membrane emulsification (i.e. the exponent of gd is 0 in
Eq. (12)). However, the disperse-phase flux highly depends on
the disperse-phase viscosity.
In batch processes, the dependencies on power density
and residence time cannot be summarized in the energy density
[22]:
(9) x µ PV
 0:25 ::: 0:4

f…tres †
gd 0 ::: 0:75
(14)
In most industrial processes, f(tres) can be described by an
exponential function as t±a. However, a saturation function
(10) better describes the reality, such as:
btres
f…tres † ˆ k
e ‡
xt (15)
res !1
Herein, k and b are constants, and xtres ®¥ is the equilibrium
mean droplet diameter which is found for tres®¥.
0930-7516/03/0202-0071 $ 17.50+.50/0 71
 Full Paper
5.3 Process Functions for Droplet Disruption
Using the specific energy for continuous processes as well as
the power density for batch processes, process functions can
be defined for emulsification (Eqs. (12) to (15)). For each
formula, the proportional constants as well as the exact
exponents can then be determined using this diagram. The
required specific energy for a target mean droplet diameter
may be determined by the resulting equation. Process
parameters, such as homogenizing pressure, or power input
for a given throughput, can then be calculated.
Fig. 5 gives mean droplet diameters for emulsions of 30 % of
rapeseed oil in water as a function of specific energy. The
emulsions were stabilized using a very fast adsorbing
surfactant (a lauryl alcohol, LEO-10), that allows to have a
low re-coalescence rate. Thus, disruption governs the process
of emulsification. For food applications, this emulsifier would
have to be replaced by another small-molecular weight
emulsifier, which is permitted for food by law, e.g. Tween 20.
Using continuous rotor-stator machines or a high-pressure
homogenizer with a standard flat nozzle results in relatively
large mean droplet diameters. If smaller droplet diameters are
required, machines have to be used that induce other flow
patterns as well, e.g., laminar elongational flow or cavitation.
In commercial emulsification machines, this is realized in
high-pressure homogenizers with, for example, the opposing
jet nozzle, such as Microfluidizer, or the simple orifice valve,
or by sound-induced cavitation (ultrasonic emulsification). In
the simple orifice valve, for example, droplets are deformed
mainly by laminar elongational flow, destabilized, and thus
easy to disrupt by the following turbulent flow conditions
[7,29]. This nozzle design is very efficient in droplet disruption,
especially if highly viscous phases have to be finely dispersed
[10]. Similar results are found for the jet disperser [9,10].
Emulsification mainly due to cavitation (continuous ultra-
r/s-systems
ultrasonic em.
10
mean droplet diameter x 3,2 / µm
1 ϕ = 1 → 50 %
rape seed oil-in-water
0.1
104 105
spec. energy E v / (J/m )
Figure 5. Process functions for droplet disruption by different emulsification
processes (values from [6,10,16, 30]).
72 Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
sonic emulsification) results in droplet sizes comparable to
those of effective high-pressure homogenizers.
However, if ultrasonic waves are applied to an emulsion
continuously, there are some restrictions. Up to now, it is
almost impossible to avoid bimodal droplet size distributions,
even if the specific energy is increased (Fig. 6). Droplets of the
second peak are the remaining premix droplets. In a product to
be industrialized, these droplets will cause instabilities due to
sedimentation. Avoiding this instability problem requires
either recirculation of the emulsion or a design improvement
of the flow chamber [30]. Since cavitation causes droplet
disruption and high viscosities decrease cavitation, ultrasonic
emulsification is restricted to products of low viscosity.
In comparison to droplet disrupting machines, membrane
emulsification is very efficient, if small droplets (x < 3 lm)
have to be produced [15] (Fig. 5). However, the higher the
disperse phase content, the more specific energy has to be
applied, since recirculation of the emulsion is required to
concentrate the dispersed phase. If a pre-emulsion is pressed
through the membrane pores instead of the droplet phase, no
recirculation is required [17]. This does not decrease the
required energy input in total but simplifies the process for
industrialization. Emulsification by membranes results in very
steep droplet size distributions if the membrane surface is not
wetting the future continuous phase (small wetting angle). For
example, in order to produce oil droplets, the membrane
surface material has to be hydrophilic [16,31].
An improved effectivity in droplet disruption simplifies and
cheapens emulsification processes. Producing a mean droplet
diameter of, e.g., 1 lm by high-pressure homogenizing
requires a homogenizing pressure of 35 bars only if the simple
orifice nozzle is used for this recipe. However, about 2000 bars
would have to be applied using the standard homogenizing
nozzle. Using membranes instead of a high-pressure homo-
genizer would decrease the required specific energy by
an additional factor of 3±300 (Fig. 5). This is of interest
from both manufacturing costs and product quality point of
view (especially, if stress-sensitive products have to be
HPH: membrane em.
flat nozzle ϕ=1% produced).
orifice valve, ϕ = 50 %
opposing jet
volume frequency distribution q3(x) / µm-1
2
10 % water in
vegetable oil
1.5 premix
EV / 107 J/m3
0.9
1 3.6
7.0
10
0.5 Behrend, 2001
106 107 108 0
10 -2 10 -1 10 0 10 1 10 2
3
droplet size x / µm
Figure 6. Droplet size distributions of emulsions produced continuously by
ultrasonic emulsification [30].
1618-0240/03/0101-0072 $ 17.50+.50/0 Eng. Life Sci. 3 (2003) 2

5.4 Theory of Droplet Recoalescence
In emulsification, droplets are disrupted. However, they
rarely exist for long due to recoalescence. This takes place
within milliseconds after disruption and is found in practically
every emulsification process [32,33]. This is due to the fact that
new droplets are not sufficiently stabilized directly after their
disruption (either by adsorbing emulsifier molecules or by
hydrodynamic effects). Emulsification processes are thus
governed by both droplet disruption and recoalescence of
droplets [6]. Only if the droplet stabilization is very fast and if
the droplet concentration is very low, recoalescence rates are
low enough for droplet disruption to govern the emulsification
process.
The effect of recoalescence is well described in the literature
[summarized in 32]. A short insight will be given here, only.
The number of droplet collisions (collision frequency c)
depends on several parameters and may be calculated for
turbulent flow according to [34]:
c = K ´ u ´ x2 ´ n2 (16)
with
p
Kˆ 2 p=15 (17)
p
uˆ e=md
x (18)
Herein, K is the collision coefficient, u the droplet velocity, x
the droplet diameter, n the number of droplets per volume
emulsion, e the mass-specific power density, and md the
kinematic viscosity of the droplet phase. The coalescence
probability of colliding droplets usually is smaller than 1. An
easy experimental method based on image analysis allows to
determine the coalescence probability pcoal and the coales-
cence rate X for different emulsion formulas [32]:
X = pcoal
c (19)
Coalescence is found whenever droplets are disrupted. In
case of very low disperse-phase contents (< 1±3 %), the
recoalescence rate in emulsification processes is low [35]. In
industrial emulsification processes, the disperse-phase con-
tent is always higher. Thus, disruption and recoalescence are
found. In rotor-stator machines, droplets are disrupted at
relatively low power densities (Pv » 1 ... 109 W/m3) and long
residence times (tres » 0.1 ... 1 s). The number of disrupted
droplets is relatively low, and time is given to stabilization.
Thus, the coalescence probability is relatively low. In high-
pressure homogenizers or in continuously working ultrasonic
devices, however, the power density is higher (Pv » 108 ...
1012 W/m3), whereas the mean residence time is shorter
(tres » 0.001 ... 0.01 s). Many small droplets are formed in a time
being too short for stabilization. In addition, the forces in
droplet collision are higher, leading to a higher coalescence
Eng. Life Sci. 3 (2003) 2, Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper
probability. At a constant specific energy, the coalescence rate
therefore is higher than in rotor-stator machines.
If the recoalescence rate is high, the emulsification result in
total is changed [6,32]. The mean droplet diameter achievable
at a given specific energy increases. Using Eqs. (12) to (15),
that means, both the proportionality factor and the exponent
of Ev and Pv increase. The negative slope of the line in the
diagram x3,2 = f(Ev) decreases. For high recoalescence rates,
even an increase in mean droplet diameter is found for
increasing specific energy (see chapter 6.5).
5.5 Process Functions for Droplet Disruption and
Recoalescence
If a certain recipe requires an emulsion of a high content of
disperse phase or slow stabilization characteristics (as is often
found in food industry), recoalescence of newly formed
droplets determines the emulsification result (characterized
by the mean droplet diameter).
Fig. 7 gives an example of a typical food emulsion
(mayonnaise type: 80 % oil in water, emulsifier: egg yolk).
In this emulsion, the coalescence rate is high. A production in
rotor-stator machines results indeed in droplet diameters
larger than expected (compare to the results of a fast
adsorbing emulsifier, as e.g. LEO-10, in Fig. 7). However, if
a high-pressure homogenizer is used, no emulsion can be
formed at all. The coalescence rate is so high that the disperse
and continuous phase separate within minutes after produc-
tion. In production, this phenomenon is called ªover-
processingº.
In conclusion, rotor-stator machines are recommended at
higher coalescence probability [36]. Only at very low
coalescence rates, high-pressure homogenizers or ultrasonic
devices are to be chosen. In this case only, one can profit from
the advantage of high disruption effectivity [37,38]. Using a
stabilization zone may help in preventing recoalescence
during the production of food emulsions of high coalescence
rates (high oil fraction and slowly stabilizing emulsifiers)
[10,39].
100
RSM HPH - egg yolk
mean droplet diameter x3,2 / µm
egg yolk --> emulsions break
RSM
LEO-10 HPH:
tv = 0.001 - 0.01 s
10
HPH
RSM:
LEO-10
tv = 0.1 - 10 s
1
10 4 105 10 6 10 7 10 8
spec. energy Ev / (J/m3)
Figure 7. Process function for emulsification at different coalescence rates
(RSM: rotor-stator machine, HPH: high-pressure homogenizer).
0930-7516/03/0202-0073 $ 17.50+.50/0 73
 Full Paper

5.6 Scale-up in Emulsification Using Process Functions

100

mean droplet diameter x3,2 / µm

50 % rape seed oil in water
After having designed and optimized the process it has to be emulsifier: egg yolk

scaled up for industrialization. The process functions given in

chapters 6.3 and 6.5 based on specific energy may also be used
for scale-up. 10 x3,2
For rotor-stator machines, only one process function per
recipe exists. Fig. 8 gives an example of the production of a machine and
rotor diameter
dr = 50 mm
salad-cream-type emulsion, which is produced in a colloid process parameters
EV rotor diameter
may be varied
mill. The process function does not change by going from pilot dr = 110 mm
plant scale to production scale. In order to get the same 1
106 107 108
product quality (i.e., the same mean droplet diameter), the
spec. energy EV / (J/m3 )
specific energy has to be kept constant. One may even change
machine or process parameters, such as teeth design or rotor Figure 8. Scale-up using the specific energy for a salad-cream-type emulsion.

speed, as long as the specific energy is kept at a constant level.

In high-pressure homogenizers, the process function con-
stants are determined by the flow conditions within the nozzle.
During scaling up, the flow conditions thus may not be
changed. If this is not possible while scaling up, it is
recommended to keep the pilot nozzle design and enlarge
the number of nozzles according to the throughput. This is
easy, for example, for the simple orifice valve design.
In continuous ultrasonic emulsification, sound inducers of
high capacities are rarely found on the market. In addition, the
problem of designing a flow chamber will increase with
increasing throughput. Thus, it is recommended to work with
an increased number of parallel working ultrasonic devices for
scale-up [30].
In membrane emulsification scale-up is done by keeping the
membrane material and enlarging the membrane surface
according to the required throughput [16]. The specific energy,
i.e., the continuous-phase wall stress plus the transmembrane
pressure difference, has then to be kept constant.
6 Product Design
6.1 Target Product Properties
Product design will be depicted for three different food
emulsion examples: a spread, a dressing, as well as a drink. In
Tab. 1, target properties of these selected products are
presented.
The basic requirements are as fol-
Table 1. Defining the required product quality.
lows: all food products have to be of
high physical and microbiological
stability. A spread should be highly
quality parameter
viscous and high in fat. As, in general,
the spread consumption is low, the stability
should be of low viscosity. The example product also should be
low in fat. About 1.000 markets have to be delivered by one
third of a palette per day during the season in order to fulfill
the predicted consumption by marketing. Therefore, several
ten thousand tons per year have to be produced.
Besides these principal target properties, a lot of special
product or production features, as taste or color, are to be
defined, depending on marketing desires, local raw material
availability and company production facilities.
6.2 Property and Process Functions for a Spread
Based on the target product quality, the main recipe
features as well as the target microstructure are determined.
High viscosity, high fat fraction and high stability result in a
spread, that has to be of w/o-emulsion type with a low water
content (j < 30 %). The fat melting point has to be in the range
of body temperature to enable melting in the mouth, but not
on the bread sleeve. For high physical stability the droplet size
distribution has to be narrow. High microbiological stability
requires a maximum droplet size being smaller or in the order
of the size of one microorganism (» 3 lm). This prevents
microorganism growth without preservative addition. A fast
stabilizing emulsifier has to be chosen to ensure small droplets
and a narrow droplet size distribution.
product / property
spread dressing drink
high physical & microbiological

required tonnage also is very low viscosity high medium low

(several tons per year). A dressing is quantity low medium high
of medium viscosity, the chosen ex-
fat fraction high (> 70 %) medium (40-60 %) low (< 30 %)
ample should have a fat content in the
medium range (40±60 %). Higher ... ...
tonnages than for the spread are special product & depending on marketing desires, raw material availability, company production
production features faciilities, ...
required. Needless to say, the drink

74 Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1618-0240/03/0101-0074 $ 17.50+.50/0 Eng. Life Sci. 3 (2003) 2

Knowing the property functions, the process has to be
designed based on the process functions. Since the production
quantities are low, a batch process is chosen. During
emulsification the product has to be heated above its melting
point. The product is of high viscosity during production, thus
rotor-stator machines are able to realize the required high
specific energy for the target droplet diameter (xmax » 3 lm ).
As low-cost machines are a must for a product of low tonnage
and low price, a rotor-stator machine even is recommended.
The batch process allows heating and cooling to be combined
with the emulsification process. The process functions are
given in Eqs. (14) and (15). A pilot plant trial has to be
performed in order to determine the exponents of power
density, the residence time function, as well as the proportion-
ality constant for this special recipe. Scale-up will then be done
using the same process functions for the rotor-stator machine
(see chapter 5.6).
6.3 Property and Process Functions for a Dressing
The dressing should have a medium viscosity at a medium
fat content and high stability. This requires an o/w-emulsion
type with a fat fraction of 50±60 %. As emulsifier, Lacprodan-
60 has been chosen. A mean droplet diameter below 5 lm
ensures a high stability in this viscosity range. Higher tonnages
per year may be achieved in a semi-continuous process.
60 % vegetable oil in water, emulsifier: Lacprodan-60
1.0
viscosity η / Pa s at 100 s-1
(a)
0.8
0.6
0.4
ηe → x3,2
0.2
0.0
0 1 2 3
mean droplet diameter x3,2 / µm
Figure 9. Product design for a dressing based on the property function (left-hand
side) and the process function (right-hand side).
In Fig. 9 the determination of the property function for the
viscosity (a), as well as the process function for this process (b)
is depicted. At an oil content of 60 %, the target viscosity
of » 500 mPa s at 100 s±1 of this pseudoplastic emulsion is
achieved at a mean droplet size of » 4 lm. The viscosity of the
emulsion is high enough to work with a rotor-stator system
(such as a toothed colloid mill). This machine may be easily
used in a semi-continuous process. The process function
(Fig. 9b) has been determined in a pilot-plant trial. A specific
Eng. Life Sci. 3 (2003) 2,
Full Paper
energy of 6 to 7 ´ 106 J/m3 results in the target mean droplet
diameter of 4 lm. The required value of specific energy may be
realized by several combinations of process parameters, as
rotor speed, gap width or tooth design. For scale-up, the
process function in Fig. 9b can also be used.
6.4 Property and Process Functions for a Drink
A drink, of low viscosity, low fat content and high stability
requires an o/w-emulsion type with a fat fraction below < 10 %.
A fast stabilizing emulsifier, such as Tween 20, has to be used.
A narrow droplet size distribution and a mean droplet
diameter < 1 lm ensure a high stability for this system. High
tonnages per year are best achieved by a continuous process.
Since a small mean droplet diameter is required and the
disperse-phase content (and thus the coalescence rate) is low,
a high pressure homogenizer is recommended. Using the
simple orifice plate as homogenizing nozzle keeps investment
costs and energy consumption low. The process function is
given in Eq. (13) (Eq. (12), respectively for other homogeniz-
ing nozzles). As for the other products, a pilot-plant-scale trial
will give the proportionality constant and the exponents.
Having thus determined the required specific energy the
homogenizing pressure is easily calculated. For scale-up, this
pressure has to be kept constant. In order to simplify the scale-
up, the number of orifices should be multiplied without
changing geometry.
mean droplet diameter x3,2 / µm
100
7 Summary
(b)
Product design, in general, involves first the definition of the
10 toothed required product quality. On the basis of the quality
colloid mill
parameters as well as special marketing and production
features, the product recipe can be defined. Knowing the
x3,2 → Ev
aspired quality and the predefined recipe, the property
4 5
1
105 106 107 function has to be measured. Thus, the required product
spec. energy Ev / (J/m3) microstructure will be defined. In the case of emulsions, the
microstructure is characterized by the size of the droplets and
the structure built up by the droplets and high-molecular
additives. Process functions are then determined using the
specific energy for continuous processes or the power density
and the residence time for batch processes, respectively. In a
limited number of pilot-plant trials, the exponents and the
proportionality factor of the process functions have to be
determined. Using these functions, emulsification processes
are easily designed, optimized and scaled up. The determina-
tion of process functions in emulsification was depicted in
detail. Product design was finally shown for three food
emulsions of different properties.
Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/03/0202-0075 $ 17.50+.50/0 75
 Full Paper

Acknowledgements ge [Pa s] dynamic viscosity of the emulsion

md [m2 ´ s±1] kinematic droplet viscosity
The authors wish to thank the DFG (Deutsche Forschungs- rc [kg ´ m±3] continuous phase density
gemeinschaft) and the AIF/FEI (Arbeitsgemeinschaft Indus- rd [kg ´ m±3] droplet density
trieller Forschung/Forschungskreis der Ernährungsindustrie) r [Pa] external tension
for financial support of parts of this work. X [s±1 ´ m±3] coalescence rate
The content of this review was given as lecture at ECCE 3,
Food Technologies, Nuremberg, June 26±28, 2001 [40].
Received: July 10, 2002 [CET 1654]
References
[1] H. Rumpf, Staub-Reinhaltung der Luft 1967, 27 (1), 3.
Symbols used [2] E. J. Windhab, K. Feigl, M. Stranzinger, P. Fischer, Chem. Ing. Tech.
2001, 73 (6), 694.
[3] R. Polke, Chem. Ing. Tech. 1990, 62 (10), 813.
a [±] constant
[4] J. Krekel, R. Polke, Chem. Ing. Tech. 1992, 64 (6), 528.
b [±] constant [5] P. Walstra, in Encyclopedia of Emulsion Technology (Ed: P. Becher), vol.
C [±] constant 1, Marcel Dekker Inc., New York 1983, 57.
[6] H. Karbstein, Dissertation, Karlsruhe University 1994.
c [s±1.m±3] collision frequency [7] M. Stang, H. P. Schuchmann, H. Schubert, Eng. Life. Sci. 2001, 1, 151.
d [m] droplet diameter [8] E. J. Cook et al. US Patent 4 533 254, 1985 and No. 4 908 154, 1990.
E [J] energy [9] W. Hovestad, B. Klinksiek, M. Melchiors, Farbe & Lack 2000, 7, 40.
[10] M. Stang, Dissertation, Karlsruhe University 1998.
EV [J.m-3] volumetric specific energy [11] A. Treiber, Dissertation, Karlsruhe University 1979.
k [±] constant [12] B. Freudig, S. Tesch, H. Schubert, Chem. Ing. Tech. 2002, 74 (6), 875
(part I), 880 (part II).
K [±] collision coefficient
[13] O. Behrend, H. Schubert, Ultrasonics Sonochemistry 2001, 8, 271.
n [m±3] number of droplets per volume [14] S. Bechtel, N. Gilbert, H.G. Wagner, part 1, Chem. Ing. Tech. 1999, 71
Oh [±] Ohnsorge number (8), 810; part 2: Chem. Ing. Tech. 2000, 72 (5), 450.
[15] T. Nakashima, Nippon Shokuhin Kogyo Gakkaishi 1994, 41 (1), 70.
P [W] power [16] V. Schröder, Dissertation, Karlsruhe University 1999.
PV [W ´ m±3] volumetric power density [17] K. Suzuki, I. Fujiki, Y. Hagura, Food Sci. Technol. Int. Tokyo 1998, 4 (2),
pc [Pa] capillary pressure 164.
[18] K. Suzuki, K. Hayakawa, Y. Hagura, Food Sci. Technol. Res. 1999, 5 (2),
pcoal [±] coalescence probability 234.
Dph [Pa] homogenizing pressure [19] T. Kawakatsu, Y. Kikuchi, M. Nakajima, JAOCS 1997, 74 (3), 317.
[20] B. Koglin, J. Paweloski, H. Schönring, Chem. Ing. Tech. 1981, 53 (8), 641.
u [m ´ s±1] droplet velocity
[21] A. Kwade, Dissertation, Braunschweig University 1996.
V [m3] volume [22] P. Walstra, P. E. A. Smulders, in Food Colloids (Eds: E. Dickinson,
V_ [m3 ´ s±1] volume flow rate B. Bergenstahl), The Royal Society of Chemistry, Cambridge 1997.
[23] B. J. Bentley, L. G. Leal, J. Fluid. Mech. 1986, 176, 241.
tdef [s] deformatiom time [24] H. Armbruster, Dissertation, Karlsruhe University 1990.
tdef,cr [s] critical deformatiom time [25] K. Arai, M. Konno, Y. Matunaga, S. J. Saito, Chem. Eng. Jap. 1977, 10
tres [s] residence time (4), 325.
[26] J. T. Davies, Chem. Eng. Sci. 1985, 40 (5), 839.
We [±] Weber number [27] V. Schröder, H. Schubert, Coll. and Surf. A: Physicochemical and
Wecr [±] critical Weber number Engineering Aspects 1999, 152, 103.
[28] H. Schubert, V. Schröder, in Conf. Papers ªMinimal Processing of
Weturb [±] turbulent Weber number
Foodsº EFFoST (Eds: I. D. Morton, J. A. Spring), Cologne 1996.
x [m] droplet diameter [29] G. Kolb, Dissertation, University Halle-Wittenberg 2001.
x3,2 [m] mean droplet diameter (Sauter [30] O. Behrend, Dissertation, Karlsruhe University 2002.
[31] G. T. Vladisavljevic, S. Brösel, H. Schubert, Chem. Papers 2000, 54 (6a),
diameter) 282.
xmax [m] maximum droplet diameter [32] T. Danner, Dissertation, Karlsruhe University 2001.
xtres!1 [m] equivalent mean droplet diameter [33] T. Danner, H. Schubert, in Proc. Int. Congr. Particle Techn. PARTEC
2001, 06/080, Nuremberg 2001.
[34] A. K. Chesters, in Trans IchemE July 1991, 69, part A, 259.
[35] S. Brösel, H. Schubert, Chem. Eng. Proc. 1999, 38, 533.
[36] H. Karbstein, H. Schubert, in Food Ingredients Eur. Conf. Proc.,
Greek symbols
Frankfurt 1995, 93.
[37] M. Stang, H. Karbstein, H. Schubert, in FOOD COLLOIDS Proteins,
e [m2 ´ s±3] mass specific power density
Lipids and Polysaccharides (Eds: E. Dickinson, B. Bergenstahl), 1997,
c [N ´ m±1] interfacial tension 382.
j [±] disperse phase content [38] H. P. Schuchmann, H. Schuchmann, Lebensmitteltechnik 2002, 5, 50.
[39] T. Danner, H. Schubert, M. Stang, in Proc. 8th Int. Congr. on
gc [Pa s] dynamic viscosity of continuous phase Engineering and Food ICEF 8, vol. I, ISDN 1±56676±951±5, 628.
gd [Pa s] dynamic viscosity of disperse phase [40] H. P. Schuchmann, H. Schubert, Chem. Ing. Tech. 2001, 73 (6), 694.

_______________________

76 Ó 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1618-0240/03/0101-0076 $ 17.50+.50/0 Eng. Life Sci. 3 (2003) 2