International Dairy Journal 9 (1999) 817}829

A study of fat and air structures in ice cream

H.D. Go!*, E. Verespej, A.K. Smith
Department of Food Science, University of Guelph, Guelph, Ont., Canada N1G 2W1
Received 16 July 1999; accepted 6 November 1999

Abstract

Three ice cream mixes of conventional composition with varying emulsi"er content (no emulsi"er; 0.15% mono- and di-glycerides;
0.15% mono- and di-glycerides plus 0.06% polysorbate 80) were frozen using three di!erent freezing regimes (continuous freezer at
low and high back pressure and batch freezer) in order to prepare a series of ice cream samples with varying levels of fat destabilization
and foam structures. The hardened samples were viewed by thin-section transmission electron microscopy after freeze-substitution
and low-temperature embedding. The images were compared to those obtained by low-temperature scanning electron microscopy.
The structures created by increasing levels of fat destabilization were observed as an increasing concentration of discrete fat globules
at the air interface and increasing coalescence and clustering of fat globules both at the air interface and within the serum phase.
However, air interfaces at the highest levels of fat destabilization were not completely covered by fat globules, nor was there evidence
of a surface layer of free fat. Air interfaces from continuous and batch freezing were similar. ( 2000 Elsevier Science Ltd. All rights
reserved.

Keywords: Fat destabilization; Freeze-substitution; Partial coalescence; Scanning electron microscopy; Transmission electron microscopy

1. Introduction the origin of which might be non-crystallized triglyceride

Ice cream is a complex food colloid, containing fat
globules, air bubbles and ice crystals dispersed in
a freeze-concentrated dispersion/solution of proteins,
salts, polysaccharides and sugars. A review of the collo-
idal aspects of ice cream (Go!, 1997a) revealed that one
element of structural information that was less well
documented was the nature of fat structures within the
frozen system itself, and the interaction of these fat struc-
tures with air. While the process of fat emulsion destabil-
ization and its e!ects on ice cream quality are well known
(Berger, 1997; Campbell & Pelan, 1998; Go!, 1997a, b;
Tharp, Forrest, Swan, Dunning & Hilmoe, 1998; Thom-
sen & Holtsborg, 1998; Walstra & Jonkman, 1998), it is
not clear whether the appropriate schematic model for
destabilized fat networks is one of partially coalesced
(clustered) fat globules adsorbed to air bubbles, or fat
clusters primarily in the bulk, or both. The role of coales-
cence in fat destabilization is not clear. There has been
debate about the presence of liquid fat at the air interface,
* Corresponding author. Tel.: 519-824-4120; ext. 3878; fax: 519-824-
6631.
E-mail address: dgo!@uoguelph.ca (H.D. Go!)
escaping from the fat globule during rupture (Berger,
1997). It is also not clear what happens to the structure of
fat at air bubble interfaces during the process of expan-
sion of the bubbles, which must happen in the transition
from imposed back pressure to atmospheric pressure
upon drawing ice cream from the swept-surface freezer
(Carnell, 1991).
Much of the information on fat/air interactions in ice
cream is based on measurements in melted ice cream, e.g.,
turbidity (Go! & Jordan, 1989), particle size distribution
(Gelin, Poyen, Courthadon, Le Meste & Lorient, 1994),
solvent extraction of fat (Pelan, Watts, Campbell & Lips,
1997), and microscopy (Go!, 1997a), rather than on the
system in its frozen state. If the fat globules are very
unstable, the melting of ice crystals and resulting struc-
tural collapse may cause coalescence, in which case the
structure measurements may not actually represent con-
ditions seen in the frozen system. There has been some
uncertainty as to what is actually being measured by
these various fat destabilization techniques, particularly
with the solvent extraction method, and which measure-
ments represents best the actual frozen structure of ice
cream. Extrapolation has also been made between the
structure of whipped cream and that of ice cream (Go!,
1997b). These are both dairy emulsions that undergo fat

0958-6946/00/$ - see front matter ( 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 8 - 6 9 4 6 ( 9 9 ) 0 0 1 4 9 - 1
818 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829
destabilization in the presence of shear, and into which
an air phase is introduced and comminuted to small
bubbles. However, ice cream contains a smaller phase
volume of fat plus an ice phase, and air is incorporated in
a di!erent manner than it is in whipped cream, so al-
though the fat and air structures and interactions may be
similar, there may be also signi"cant di!erences.
Electron microscopy has been used extensively to
examine ice cream structure, but the need to maintain the
ice phase (at subzero temperatures) has introduced major
limitations to the approaches that can be applied. Vari-
ous techniques and their applications to dairy products
based on fat, including ice cream, were reviewed by
Kalab (1985). The "rst reported methods utilized thin-
section transmission electron microscopy (TEM) of high-
ly "xed samples of melted ice cream (Alsafar & Wood,
1966, 1968), or freeze-fracture, metal shadowing, and
examination of replicas of frozen ice cream with TEM
(Berger & White, 1971). The development of low-temper-
ature (cryo) scanning electron microscopy (LT-SEM) has
allowed further detailed studies of ice cream structure,
particularly the ice phase (Caldwell, Go! & Stanley,
1992a, b). However, both the metal replication TEM
technique and LT-SEM are limited to the features of
fractured surfaces, and in the latter case, magni"cation
and resolution are not high enough to discern interac-
tions between fat globules. Thin sectioning for TEM after
subambient temperature "xation has been used to view
ice cream mix (Go!, Libo!, Jordan & Kinsella, 1987),
melted ice cream (Go!, 1997a), and whipped cream sam-
ples (Brooker, Anderson & Andrews, 1986), however, the
use of aqueous "xatives has limited the use of thin-
section techniques to non-frozen samples. In this paper,
TEM thin-section methods have been used through ap-
plication of a freeze-substitution technique for subzero
sample "xation. This latter method has been applied to
the study of biological materials (Robards & Sleytr, 1985;
Humbel & Muller, 1986) and whipped cream (Smith,
Go! & Kakuda, 1999) but its application to study ice
cream structure has not been recorded in the literature.
There has been no reported use of cryo-sectioning of ice
cream for viewing with cryo-TEM, but this may be forth-
coming with the presence of new, sophisticated instru-
mentation, and may provide the ultimate in structural
information.
The objective of this work was to examine fat and air
structures in ice cream samples containing di!ering de-
grees of fat destabilization, using LT-SEM and freeze-
substitution TEM. Variations in fat destabilization were
achieved by processing ice cream with three di!erent
emulsi"er concentrations in the formulation (no emulsi-
"er; mono- and di-glycerides added; mono- and di-gly-
cerides plus polysorbate 80 added) and by three di!erent
freezing techniques (batch freezing; continuous freezing
at low back pressure; and continuous freezing at high
back pressure). Each of these variations was expected to
have an impact on the formation of air bubbles and the
degree of fat destabilization exhibited in the ice cream;
hence, on the fat and air structures created.
2. Materials and methods
2.1. Ice cream processing
The ice cream mix used to prepare the ice cream
consisted of 11% fat (from sweet butter, Gay Lea Foods,
Guelph, ON, Canada), 11% milk solid-non-fat (from
instant skim milk powder, Gay Lea Foods, Guelph, ON,
Canada), 10% sucrose (Redpath Sugar, Toronto, ON,
Canada), 5% 42 DE corn syrup solids (Casco Inc.,
Toronto, ON, Canada), 0.15% guar, and 0.02% car-
rageenan (Germantown Canada Inc., Scarborough, ON),
the balance being made up with water. Three emulsi"er
variations were incorporated as follows: no emulsi"er
added; 0.15% mono- and di-glycerides added (German-
town Canada Inc., Scarborough, ON), referred to as the
mdg mix; 0.15% mono- and di-glycerides plus 0.06%
polysorbate 80 added (Germantown Canada Inc., Scar-
borough, ON), referred to as the ps80 mix. The 40 kg
mixes were batch pasteurized at 753C for 15 min and
homogenized at 17.2 MPa "rst stage, 3.4 MPa second
stage (Superhomo, Cherry Burrell Corporation, Chicago,
IL), cooled to 43C and aged for 24 h. Vanilla (two-fold,
vanillin-vanilla, CSP Foods, Toronto, Ont., Canada) was
added to the mixes after aging and prior to freezing at
a rate of 3 mL per kg.
Batch freezing of 2 L aliquots of each mix was carried
out in a Taylor batch freezer (model B733-32, Tekni-
Craft, Rockton, IL). Ice cream was frozen to a draw
temperature of !53C, and whipping was continued for
a total of 15 min. The target overrun was 90%, which was
achieved with the mdg and ps80 mixes. However, the mix
with no added emulsi"er only whipped to an overrun of
60% in the 15 min residence time within the freezer. Ice
cream was drawn at !53C into 1 L cups and immediate-
ly placed into a hardening room at !253C for storage.
Continuous freezing was performed in a Vogt Freezer
(Model VA-80, Cherry-Burrell Corporation, Chicago,
IL) to!53C. Two di!erent extrusion back pressures were
used: low pressure at 172 kPa; and high pressure at
414 kPa. In the former case, ice cream achieved an over-
run of 65% from the mix with no added emulsi"er, and
80% from either mdg or ps80. In the latter case, ice
creams achieved an overrun of 100% in all three samples.
Ice cream was drawn at!53C into 1 L cups and immedi-
ately placed into a hardening room at!253C for storage.
2.2. Quantitative measurements of fat destabilization
Turbidity: To evaluate the degree of fat destabilization
by spectroturbidimetry (Go! & Jordan, 1989), a sample
 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829
of each of the nine hardened ice creams (three formula-
tions, three freezing methods each) was thawed and an
aliquot of 40 mL was taken for analysis. Samples were
diluted 1 : 500 with "ltered water (MiliQ, Millipore Ca-
nada, Mississauga, ON) and absorbance was measured
by a spectrophotometer (Shimadzu UV-1201, Kyoto, Ja-
pan) at 540 nm against MiliQ water as a blank. Lower
absorbance, as the solution becomes less turbid, relates
to higher levels of fat destabilization due to #occulation
or coalescence of the fat. Analyses were performed in
triplicate.
Particle size analysis: Particle size distribution of the
melted ice creams after hardening was measured by in-
tegrated light scattering, using a Mastersizer X (Malvern
Instruments Ltd., Malvern, Worcs., UK). The measure-
ments were carried out at room temperature. The dilu-
tion of the emulsion in the sample chamber was
approximately 1 : 1000. Mean particle size diameters
d (the volume-weighted diameter) and d (the surface
4,3 3,2
weighted diameter) were recorded, as well as the cumu-
lative percentage of the particles at 2.30 lm. Analyses
were performed in triplicate.
2.3. Scanning electron microscopy
Ice cream specimens (approximately 100 mm3) for
LT-SEM were taken from the inner bulk of the hardened
samples at !253C with a surgical blade and immediately
placed into liquid nitrogen (!1963C). Small pieces
(2 mm]3 mm) suiting the LT-SEM specimen holders
were then broken from the specimen with the surgical
blade under slight force. The holders were a double-screw
spring-loaded support style (Caldwell et al., 1992a) each
holding two specimens. The holder and specimen under
liquid nitrogen were transferred into the cryo-prepara-
tion unit (EMscope SP2000A, Ashford, Kent, UK). At
!1603C inside the unit, both specimens were fractured
revealing a fresh surface of the ice cream sample for
scanning. The specimen was sublimated at !803C for
25 min (Caldwell et al., 1992a) and then sputter coated
with 30 nm layer of gold at !1603C. The holder was
transferred under vacuum into the cold stage (!1403C)
of the Hitachi S-570 scanning electron microscope (To-
kyo, Japan), where samples were viewed and photo-
graphed at 10 kV accelerating voltage using di!erent
magni"cations. At least three pieces of each ice cream
sample from di!erent containers and 20 "elds per piece
were viewed.
2.4. Transmission electron microscopy
Ice cream specimens for TEM were collected as for
SEM and immediately placed into liquid nitrogen
(!1963C), where they were broken into (1.0 mm3
pieces as above. The freeze-substitution techniques of
Robards and Sleytr (1985) and Humbel and Muller
819
(1986) were adapted for use. Frozen specimens of ice
cream were transferred into vials that contained a mix of
"xatives consisting of 2.15% (w/v) uranyl acetate, 3.23%
(v/v) glutaraldehyde and 1.0% (w/v) osmium tetroxide in
absolute methanol, which had been kept at !1963C in
liquid nitrogen (the "xative mixture is solid at this tem-
perature). The vials were then placed in a !803C freezer
for 4 d, where after warming up from !196 to !803C
the "xative mixture melted and the gradual freeze substi-
tution of ice with methanol took place. Samples were
then transferred to !403C for 1 d where some "xation
with glutaraldehyde, osmium tetroxide and uranyl acet-
ate took place (there was evidence of staining after treat-
ment at !403C). Samples were then transferred to
!203C for 2 d where "xation proceeded at a faster rate
and further staining was accomplished. The "xative mix-
ture was replaced by washing the specimens in precooled
(!203C) 100% methanol followed by repeated washing
(3]) with absolute ethanol. Specimens were placed into
gelatin capsules and the in"ltration with low-temper-
ature embedding resin Lowicryl HM20 (Marivac Ltd.,
Halifax, NS, Canada) was performed as follows: one
wash with ethanol : Lowicryl 1 : 1 at !203C for 10 min,
one wash with ethanol : Lowicryl 1 : 3 at !203C for
10 min and three washes with 100% Lowicryl at !203C.
The resin-in"ltrated samples were polymerized under
360 nm UV light at !203C. Resin blocks were sectioned
at a thickness of 90 nm using an LKB ultramicrotome
(Leica Reichert Ultracut S, Vienna, Austria). Sections
were mounted immediately on formvar coated grids
(Marivac Ltd., Halifax, NS, Canada). No post-staining
was carried out, as the applied "xation process provided
a su$cient contrast for TEM imaging. The sections were
observed and photographed in a Hitachi H-7100 TEM
(Tokyo, Japan) at 75 kV. At least three blocks of each ice
cream sample from di!erent containers were sectioned,
and at least three sections per block and 20 "elds per
section were viewed.
3. Results and discussion
3.1. Ewect of ingredients and processing on fat
destabilization
Three formulations and three freezing methods that
were chosen were expected to provide variations in fat
and air structures. Table 1 shows the quantitative
measurements characterizing fat destabilization. The ex-
tent of fat destabilization (as either absorbance or par-
ticle size) generally increased from the mix with no added
emulsi"er to the mdg mix to the ps80 mix. Similar trends
have been reported by Gelin et al. (1994), Gelin, Poyen,
Rizzotti, Le Meste, Courthadon and Lorient (1996a, b),
Go! and Jordan (1989), Pelan et al. (1997), Tharp et al.
(1998) and Thomsen and Holtsborg (1998). Batch
820 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829

Table 1
Fat destabilization measurements (absorbance at 540 nm; volume weighted diameter d , surface weighted diameter d , and cumulative percentage
4,3 3,2
of distribution at 2.3 lm as determined by integrated light scattering), mean and standard deviation (n"3), for ice creams frozen from mixes with no
added emulsi"er (No-e), mono- and di-glycerides added (MDG), and mono- and di-glycerides and polysorbate 80 added (PS80), using continuous
freezing at low (L) and high (H) back pressures and batch (B) freezing

Mix formulation and Absorbance at Particle size distribution

freezing process 540 nm
d (lm) d (lm) Cumulative %
4,3 3,2
at 2.30 lm

No-e } mean 1.28 1.80 0.54 94.7

L
} st. dev. 0.02 0.03 0.01 0.2
No-e } mean 1.28 1.67 0.52 94.5
H
} st. dev. 0.01 0.24 0.02 0.8
No-e } mean 1.23 2.31 0.55 91.2
B
} st. dev. 0.04 0.26 0.02 0.7
MDG } mean 1.14 1.57 0.62 93.1
L
} st. dev. 0.01 0.40 0.01 0.3
MDG } mean 1.16 2.29 0.63 90.6
H
} st. dev. 0.03 0.45 0.01 1.2
MDG } mean 0.78 9.48 1.04 54.8
B
} st. dev. 0.01 1.1 0.03 0.9
PS80 } mean 0.97 4.69 0.92 67.6
L
} st. dev. 0.01 1.0 0.02 0.7
PS80 } mean 0.92 5.87 1.15 60.5
H
} st. dev. 0.01 0.30 0.10 1.2
PS80 } mean 0.79 8.67 1.06 49.5
B
} st. dev. 0.01 0.70 0.02 0.8

freezing produced higher levels of fat destabilization than

continuous freezing, under the conditions used. There
was a slight trend to increased fat destabilization from
low back pressure continuous freezing to high back pres-
sure continuous freezing, especially in the ps80 mixes.
This trend has been shown by Kokubo, Sakurai,
Hakamata, Tomita and Yoshida (1996) and Kokubo,
Sakurai, Iwaki, Tomita and Yoshida (1998). As fat de-
stabilization proceeded, the particle size distribution
became more bimodal (Fig. 1), representing either coales-
cence and/or partial coalescence of emulsion droplets
(Gelin et al., 1994, Boode & Walstra, 1993), which were
initially of sizes similar to those created by the homogen-
ization process (data not shown). Thus, the cumulative
percentage of the distribution at 2.30 lm (Table 1) was Fig. 1. Particle (fat globule and fat cluster) size distributions (lm) in ice
also taken as a measure of the extent of change from the creams showing low (solid line: mix with no emulsi"er added, continu-
original distribution to a destabilized emulsion. This ous freezer, low back pressure) and high (dashed line: mono- and
di-glycerides and polysorbate 80 added, continuous freezer, low back
value was chosen to represent the upper end of the pressure) levels of fat destabilization.
normal distribution in the mixes prior to freezing. These
trends are as expected, and indicate that varying degrees
of fat/fat, and presumably fat/air, structures were created
by the formulations and processes used. This provides clearly between them. Thus, we have used the following
assurance that the microscopic images of each sample nomenclature to describe the fat destabilization process:
represent structures from ice creams with signi"cantly coalescence, representing an irreversible increase in the
di!erent levels of fat destabilization. size of fat globules and a loss of identity of the coalescing
The complex fat destabilization process is made up of globules; yocculation, describing a reversible (with minor
several phenomena, and as we are attempting to discern energy input) agglomeration of fat globules; partial co-
their speci"c involvement, it is important to distinguish alescence, an irreversible agglomeration of fat globules,
 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829 821

Fig. 2. Scanning electron micrographs of ice cream frozen from the mix with no added emulsi"er processed with three di!erent freezing regimes (left
set, A and B, continuous freezer, low back pressure; center set, C and D, continuous freezer, high back pressure; right set, E and F, batch freezer) at two
magni"cations each (top set, A, C, and E, bar (shown in F)"15 lm; bottom set, B, D, and F, bar"5 lm). a"air bubble, f"fat globule, i"ice
crystal, s"serum phase.

held together by a combination of fat crystals and liquid
fat; and a retention of identity of individual globules as
long as the fat crystal structure is maintained (i.e., a tem-
perature-dependent phenomenon); a cluster describes
the agglomerate of partially coalesced globules (Go!,
1997a, b).
3.2. Ewect of ingredients and processing on microstructure
There were slight di!erences in the air phase volume
among the samples as indicated in the methods section,
resulting from varying composition and the capabilities
of the swept surface freezers used. However, di!erences
were small and thus were not expected to have a!ected
the air interface structure formed. Figs. 2 and 3 are
representative LT-SEM and TEM images, respectively,
selected from all "elds viewed of the mix with no added
emulsi"er frozen with the three di!erent processes. Fat
globules are seen by LT-SEM in both the bulk phase and,
in low concentration, at the air interface. However, they
are very discrete showing minimal interaction with each
other. There appears to be little di!erence in fat or air
structure from any of the three freezing regimes. This
similarity in structure is in consent with the results in
Table 1, which showed little development of fat destabil-
ization in the mix with no emulsi"er added. It is worthy
of note that, while there is minimal coverage of fat at the
air interface, there is a strong interface present that dem-
onstrated continuity even when ice was removed by sub-
limation (i.e., the air interface is not in direct contact with
ice). This illustrates the important functional role of pro-
tein in creating the air interface.
Freeze-substitution thin-section TEM (Fig. 3) provides
a new structural look at frozen ice cream. Some chatter
can be seen as a result of the ultramicrotome knife pas-
sing across the regions of di!ering density due to high
and low resin substitution, however, the structure has
been remarkably well preserved in this technique. Both
the fat and the ice have been substituted with resin, so
di!erentiation between them was only possible on the
basis of conformation and interface. In Fig. 3A, an ice
crystal can be seen, the shape (corner) of which is much
di!erent than the air bubble seen in Fig. 3C. Likewise, the
interface of air and ice can be di!erentiated in Fig. 3E, in
which fat and protein can be seen adsorbed or residing at
the air interface whereas the ice interface is presented as
merely a disruption in the continuity of the unfrozen
phase (i.e., little evidence of an actual interfacial layer).
Air and ice interface di!erentiation became easier as the
extent of fat destabilization increased with the mdg and
822 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829

Fig. 3. Transmission electron micrographs of ice cream frozen from the mix with no added emulsi"er processed with three di!erent freezing regimes
(top set, A and B, continuous freezer, low back pressure; center set, C and D, continuous freezer, high back pressure; bottom set, E and F, batch freezer)
at two magni"cations each (left set, A, C, and E, bar (shown in F)"5 lm; right set, B, D, and F, bar"1 lm). a"air bubble, c"casein micelle, f"fat
globule, i"ice crystal, s"serum phase. Arrowheads in C and D point to chatter during thin-sectioning as the ultramicrotome knife passed through
areas of di!ering density. Arrowhead in F points to a casein micelle adsorbed and spread at the fat globule interface.

ps80 mixes. Fat globules can be seen as discrete particles
of the order of 1 lm or less, and the interfacial adsorption
of casein micelles to fat globules can clearly be seen (Fig.
3B, D and F). There is little if any direct fat/air contact
seen, but the important role of protein at the air interface
is evident in Fig. 3C. The serum (unfrozen) phase appears
rather uniform, being characterized by fat globules and
casein micelles. Dissolved solutes and suspended macro-
molecules must reside in the milieu surrounding the other
discrete structural elements.
Figs. 4 and 5 are representative LT-SEM and TEM
images, respectively, of the mdg mix frozen with the three
di!erent processes. Reference to Table 1 suggests that
structural di!erences should become apparent compar-
ing the mix frozen in the batch freezer to that frozen at
either pressure in the continuous freezer or to the mix
 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829 823

Fig. 4. Scanning electron micrographs of ice cream frozen from mix with mono- and di-glycerides added processed with three di!erent freezing regimes
(left set, A and B, continuous freezer, low back pressure; center set, C and D, continuous freezer, high back pressure; right set, E and F, batch freezer) at
two magni"cations each (top set, A, C, and E, bar (shown in F)"15 lm; bottom set, B, D, and F, bar"5 lm). a"air bubble, f"fat globule, i"ice
crystal, s"serum phase. Arrowhead in F points to clustered fat at the air interface.

with no added emulsi"er under any freezing regime.
Although slight di!erences in fat destabilization existed
in the mdg mix from the continuous freezer, they were
similar in magnitude to the mix with no added emulsi"er.
In comparing Fig. 4 with Fig. 2, it appears under LT-
SEM that more complete coverage of fat at the air
interface existed in the mdg mix than in the mix with no
added emulsi"er, and following batch freezing (Fig. 4E
and F) compared to continuous freezing (Fig. 4A}D). It
also appears that the fat globules in the mdg mix follow-
ing batch freezing were more connected (in closer contact
with each other, hence probably displaying either #oc-
culation or partial coalescence) than in the sample fol-
lowing continuous freezing or in the mix with no
emulsi"er added. This is also evident in Fig. 5 as discern-
ed by TEM. In general, more fat is seen at the air
interface from the mdg mix (Fig. 5) than from the mix
with no emulsi"er added (Fig. 3). Evidence of fat clusters,
both at and extending away from the fat interface, is seen
in Fig. 5E and F; as well, fat clusters existing in the bulk
phase independent of the air interface can be seen. Al-
though fat clustering exists at the air interface, there is no
evidence of fat spreading. Casein micelles can also be seen
occupying an important role at the air interface between
adsorbed fat globules.
Figs. 6 and 7 are representative LT-SEM and TEM
images, respectively, of the ps80 mix (mono- and di-
glycerides plus polysorbate 80 added) frozen with the
three di!erent processes. The indices of fat destabiliz-
ation (Table 1), suggest that the ps80 mix behaved di!er-
ently than the mix with no added emulsi"er or the mdg
mix, and batch freezing produced much higher evidence
of fat destabilization than continuous freezing. The LT-
SEM images in Fig. 6, when compared to Figs. 2 and 4,
con"rm the trend suggested previously toward higher
concentrations of fat globules at the air interface with
increasing fat destabilization. There appears to be more
fat/fat interaction (especially evident in Fig. 6B and 6F)
than previously, both at the air interface and in the bulk,
and there was evidence of both coalescence (Fig. 6D)
and partial coalescence (Fig. 6F). However, even with
the highest levels of fat destabilization (ps80 mix,
batch frozen, Fig. 6E and F), there is still a considerable
amount of air interface not covered by fat, which is
presumably covered by protein and/or surfactant, similar
to the air interface observed in Fig. 2 with no added
emulsi"er.
Fat/air interactions both at the air interface and in the
serum phase are easily seen in Fig. 7. At the air interface,
there are examples of both bare fat protruding into the
824 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829

Fig. 5. Transmission electron micrographs of ice cream frozen from mix with mono- and di-glycerides added processed with three di!erent freezing
regimes (top set, A and B, continuous freezer, low back pressure; center set, C and D, continuous freezer, high back pressure; bottom set, E and F, batch
freezer) at two magni"cations each (left set, A, C, and E, bar (shown in F)"5 lm; right set, B, D, and F, bar"1 lm). a"air bubble, c"casein micelle,
f"fat globule, i"ice crystal, s"serum phase. Arrowhead in F points to clustered fat at and extending away from the air interface.

bubble, and fat still coated by a thin (stained) membrane,
suggesting that desorption of the membrane at the air
interface, although energetically favoured, may be kineti-
cally hindered by the low temperatures. Chains and clus-
ters of fat globules, with varying extent of interaction
between them, can be seen extending away from the air
interface into the bulk phase (e.g., Fig. 7F). The incom-
plete coverage of fat at the air interface with the highest
level of fat destabilization can be seen in Fig. 7E and F,
re-emphasizing the importance of non-fat emulsifying
materials (proteins, surfactants) at the air interface.
A similar structure, with incomplete coverage of the air
interface by fat, and partially coalesced fat extending
away from the air interface into the serum phase, was
presented by Anderson and Brooker (1988) in their TEM
examination of whipped cream. There does not seem to
 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829 825

Fig. 6. Scanning electron micrographs of ice cream frozen from mix with mono- and di-glycerides and polysorbate 80 added processed with three
di!erent freezing regimes (left set, A and B, continuous freezer, low back pressure; center set, C and D, continuous freezer, high back pressure; right set,
E and F, batch freezer) at two magni"cations each (top set, A, C, and E, bar (shown in F)"15 lm; bottom set, B, D, and F, bar"5 lm). a"air
bubble, f"fat globule, i"ice crystal, s"serum phase. Arrowheads in B and F indicate clustered fat globules at the air interface while arrowhead in
D indicates example of coalesced fat.

be a distinct di!erence at the air interface created by the
freezing and air incorporation regime used, when com-
paring the high and low back pressures in the continuous
freezer, or comparing these to the batch frozen samples.
This suggests that air expansion on extrusion of ice
cream from the high pressure environment of the freezer
is not responsible for any major disruption of fat net-
works at the air interface.
The concentration of casein micelles in the unfrozen
phase is more noticeable with higher levels of fat destabil-
ization as the fat becomes more clustered (comparing the
serum phase in Fig. 7 with Fig. 3). Go! (1997a) presented
a TEM thin-section micrograph of melted ice cream
showing distinct regions of concentrated casein micelles
within a coalesced fat matrix. It appears that this region
of concentrated micelles originates from the frozen struc-
ture and was not produced during the melting of the
sample for the previous TEM analysis. Walstra and
Jonkman (1998) concluded that there was very little
change to casein micelle conformation or structure as
a result of ice cream manufacture and freezing. However,
the concentration of casein micelles seen in Fig. 7 is
striking. Based on an average value of casein in the milk
solids-not-fat, these mixes contained about 3.2% casein.
At !253C, the temperature at which the structure was
"rst quench-frozen, the ice content was approximately
80% of the total moisture, or about 50% by weight (or
volume excluding the air). Thus the casein micelles were
dispersed in approximately 12.5% (by wt.) unfrozen
water. Assuming a voluminosity of the micelles of
2}4 mL g~1 (Walstra & Jonkman, 1998), it is therefore
not surprising that a very large proportion of the serum
phase volume is occupied by casein micelles. The process
of partial coalescence further concentrates the casein
micelles into discrete areas.
Fig. 8 shows a high magni"cation comparison of the
fat interface in the serum phase of ice cream from mixes
with no added emulsi"er and with mono- and di-glycer-
ides and polysorbate 80 added. Similar to the structure of
the fat globules in mixes prior to freezing (Go! et al.,
1987), casein micelles can be seen adsorbed and spread at
the fat interface in large numbers in ice cream with no
added emulsi"er (Fig. 8A), and mostly absent from the fat
interface in ice cream with emulsi"er present (Fig. 8B).
The result is that the fat in Fig. 8A is seen as very discrete
and globular, while the fat in Fig. 8B is seen as clustered
through partial coalescence. Fat and casein are more
evenly distributed with no surfactant (Fig. 8A), while the
interacting fat globules in Fig. 8B have given rise to
localized regions of high casein content.
Another striking microstructural feature was the het-
erogeneity present in the serum phase, increasing as the
826 H.D. Gow et al. / International Dairy Journal 9 (1999) 817}829

Fig. 7. Transmission electron micrographs of ice cream frozen from mix with mono- and di-glycerides and polysorbate 80 added processed with three
di!erent freezing regimes (top set, A and B, continuous freezer, low back pressure; center set, C and D, continuous freezer, high back pressure; bottom
set, E and F, batch freezer) at two magni"cations each (left set, A, C, and E, bar (shown in F)"5 lm; right set, B, D, and F, bar"1 lm). a"air bubble,
c"casein micelle, f"fat globule, s"serum phase. Arrowhead in D points to thin membrane between fat globule and air while arrowhead in E points
to clustered fat globules in the serum phase.

extent of fat destabilization in the samples increased. It
was noted above that as the fat became more clustered,
regions of high casein content also became more appar-
ent. However, the sample heterogeneity in the serum
phase manifested itself in regions that were nearly devoid
of both casein and fat. Fig. 9 demonstrates such regions
as seen by both LT-SEM (Fig. 9A) and TEM (Fig. 9B).
These regions may have been comprised solely of the
serum phase solution (sugars and water), or they may
have represented localized regions of high concentration
of the polysaccharides.
Several comparisons have been made in the past be-
tween the fat network established in ice cream and that
established in whipped cream (Go!, 1997b, Campbell