Joumul of Food Engineering 33 (1997) 305-320

0 1997 Elsevier Science Limited

All rights reserved. Printed in Great 13ritain
PII: SO260-8774(97)00030-7 0260-8774197 $17.00 +O.(H)
ELSEVIER

Viscoelastic Properties of Edible Lipids

T. H. Shellhammer,“‘f T. R. Rumsey” & J. M. Krochta”Th*

“Department of Biological and Agricultural Engineering, University of California, Davis.
CA 95616, USA
“Department of Food Science and Technology, University of California, Davis,
CA 95616, USA

(Received 22 April 1996; revised 12 March 1997; accepted 17 March 1997)

ABSTRACT

The viscoelastic natures of beeswax, candelilla wax, camauba wax and a high-
melting milkfat fraction were compared by estimating their relaxation
parameters from stress relaxation data. A generalized Maxwell model consisting
of one spring element and two Maxwell elements best descn’bed the stress
relaxation of all lipids tested. The stress responses over both the compressive
and relaxation portions of a stress relaxation test were considered. Cone
penetration tests were used as a comparison to the stress relaxation tests.
Candelilla wax and camauba wax behaved similarly as hard and elastic
materials which resisted deformation, Beeswax and the milkfat fraction, on the
other hand, were significantly more viscous, less elastic, and more easily
deformed. 0 1997 Elsevier Science Limited. All rights reserved

NOTATION

E Elastic modulus (Pa)

~3, Elastic modulus for a first spring element of the generalized Maxwell
model (Pa)
Ei Elastic modulus for an ith spring element (Pa)
fI Number of elements in the generalized Maxwell model
t Time (s)
t’ Dummy variable for time
[I Elapsed time to reach E()(s)

*To whom correspondence should be addressed. Tel.: (916) 752-2164, Email: jmkrochta@
ucdavisedu
tcurrently with Departments of Food Science and Technology and Food; Agricultural and
Biological Engineering; The Ohio State University, Columbia, OH 43210.
305
306 T H. Shellhammer et al.

E Strain (unitless)
Eo Strain at the end of the compression phase of the stress relaxation pro-
cedure (unitless)
4(t) Relaxation modulus (Pa)
Vi Viscosity of the fluid in the ith dashpot of the generalized Maxwell model
(Pa s)
0 Stress (Pa)
z Relaxation time (s)

INTRODUCTION

Beeswax, candelilla wax, carnauba wax, and high-melting milkfat fractions are
natural lipid materials which have been examined for use alone, or in conjunction
with other materials, for use as edible barriers (Greener, 1992; McHugh & Krochta,
1994; Shellhammer et al., 1993). The chemical composition of these materials is
quite diverse (Table 1). Anhydrous milkfat is composed chiefly of triglycerides of
various fatty acid length and degree of saturation. Over 60% (w/w) of the fatty acids
in milkfat are saturated, approximately 30% are monounsaturated, and 4% are
polyunsaturated (Baer, 1991). It is this range of triglyceride size and degree of
saturation that gives milkfat a broad melting range, from -20 to 37°C (Banks,
1991). At room temperature, it is a mixture of oil, semi-hard fat, and hard fat. The
ratio of liquid to solid fat determines milkfat’s rheological properties.
The melting point, and in turn the mechanical properties of milkfat, can be
altered by separation into high and low temperature-melting fractions. This can be
accomplished by a number of methods, but the most popular industrial technique is
selective crystallization (Mortensen, 1983). Fractionation by selective crystallization
from melted milkfat concentrates the higher-melting, long-chained saturated tri-
glycerides in the higher-melting fraction while concentrating the short-chained

TABLE 1
The Chemical Composition and Melting Point of Selected Lipids
Milkfat Beeswaxh Candelilla Camauba
fraction” WfLVh WUXh

Wax acid esters (%) - 71 28.5 84.5

Wax acids (free) (%) - 14 8 3.0
Fatty alcohols (%) - 1 13
Lactides (%) - ;::
Fatty acid esters (%) - T -
Hydrocarbons (%) - 12 50
Resins (%) 98 - ::tl
Triacylglycerides (%)
Mono- and diglycerides (%) 2 -
Moisture (%) 0 T 0.5 OS
Melting point (“C) _ 45’ 62-64 66-71 83-86
aBanks (1991).
‘Bennett (1975).
‘Kaylegian (1993).
 Viscoelasticproperties of edible lipids 307

triglycerides in the lower-melting fractions. This results in the higher-melting frac-

tions having significantly higher solid-fat : liquid-fat ratios than the original milkfat
(deMan & Finoro, 1980). These fractions are therefore harder in terms of mechani-
cal properties than the non-fractionated milkfat.
Waxes are naturally occurring esters of long-chained carboxylic acids (C,, or
greater) with long-chained alcohols (C,, or greater) (Streitwieser & Heathcock,
1985). Beeswax is considered an animal wax and is produced by honey bees for
building honeycomb cells. It is a mixture of wax esters, wax acids and hydrocarbons.
Hydrolysis of these esters yields mainly CZ6 and Cl8 carboxylic acids and C30 and C&
straight-chained primary alcohols (Streitwieser & Heathcock, 1985). Candelilla wax
is a vegetable wax that is produced by a reed-like plant (Euphorbia antisiphilitica,
Euphorbia cerifera, and Pedilanthus pavonis) which grows wild in northwestern Mex-
ico and southern Texas (Bennett, 1975). It is composed in large part of
hydrocarbons with some wax acid esters and alcohols. Carnauba wax is a plant
exudate from the Brazilian ‘tree of life’ (Copemica cerifera). It is the hardest,
highest-melting, natural commercial wax (Bennett, 1975) and is composed almost
entirely of esters of CZ4 and CZ8 carboxylic acids and C32 and C34 straight-chained
primary alcohols (Streitwieser & Heathcock, 1985).
Instrumental measurement of solid and semi-solid lipid rheological properties has
been accomplished with a wide variety of methods. Penetrometers are most com-
monly used, while extrusion instruments, parallel plastometers and wire cutting
devices are used less frequently (deMan, 1976). Penetrometers are classified as
constant-weight or constant-speed. In the former test, the weight of the penet-
rometer drives the plunger into the sample until it comes to rest due to the lipid’s
resistive force. Constant-speed tests involve driving a penetrometer into a lipid
sample at a constant speed and measuring either the distance of penetration to
reach a specified resistive force or the magnitude of resistance at a specified pene-
tration distance. A standard, constant-weight, cone penetration method (AOCS,
1960) has been used for determining consistency of fats (Davey, 1989; Hayakawa &
deMan, 1982). Comparisons of cone angles and test speeds were performed by
Tanaka et al. (1971) on solid and semi-solid foods including several fats. The
viscoelastic nature of lipids, in a few instances, have been studied by examining their
creep compliance (deMan et al., 1985; Shama & Sherman, 1970). Although the
experimental procedure and data analysis are more complex with viscoelastic meas-
urements than with the cone penetrometer, the results obtained give more
information about the structure of the lipid materials (Shama & Sherman, 1970).
The objectives of this study were to determine the viscoelastic properties of edible
lipid film-forming materials using stress relaxation testing, and to compare these
results with constant-speed, cone penetrometer measurements to determine whether
these two tests provided unique or redundant measures of lipid hardness.

THEORETICAL BACKGROUND

The stress-strain relationship for linear viscoelastic behavior is often given in

hereditary integral form. Hereditary integrals describe the stress response given a
strain input plus any strain history (Fltigge, 1975). For uniaxial stress relaxation, this
integral is written as
308 T H. Shellhammer et al.

a(t)= b&t-t’) Eat

The time-dependent relaxation modulus, 4(t), describes the relationship between
the strain input and stress response. Models for the relaxation modulus are often
determined from mechanical models, that is combinations of springs and dashpots in
series or parallel. A spring and dashpot in series is called a Maxwell model (Mohse-
nin & Mittal, 1977) or Maxwell material (Fhigge, 1975), and a number of Maxwell
bodies in parallel with each other and with a spring is called a generalized Maxwell
model (Mohsenin, 1986). The relaxation modulus for the generalized Maxwell
model can be shown to be

$(t) = Eo+ jg, Eje- $’ (2)

In the present study, a static, constant-strain test was used to observe stress relaxa-
tion. Viscoelastic materials behave in a linear manner generally under very small
strains. Findely et al. (1976) suggest that strain be no greater than one to two
percent; Mohsenin and Mittal (1977) found that fresh fruit exhibit linear viscoelastic
behavior if they are strained less than 1.5 to 3.0%; and Skinner and Rao (1986)
reported that frankfurters should be strained no more than 3.8%. Ideally, a stress
relaxation test involves an instantaneous strain to a fixed value (co) followed by
observation of the material stress decaying over time. In reality, instantaneous
loading to a fixed strain is impossible to achieve because of the mechanical limita-
tions of the testing equipment. Depending upon the viscous nature of the material,
the material will begin to relax as it is deformed to reach co. Rather than ignore the
initial loading portion of the experiment and assume that the material was instanta-
neously strained, the relaxation which occurs during the loading of the sample can
be included in the mathematical analysis using eqn 1. Mathematically, these condi-
tions are

a&
Compression: - = Z!L mo<t<t,
at t,

i3E
Relaxation: - =O@ t>t, (4)
at

Substituting eqn 2 in place of 4 in eqn 1 and using the strain rate stated in eqn 3,
the stress response during compression (0 <t < t,) can be described as

(5)

Using the same substitution of 4 but evaluating a(t) over the relaxation portion of
the test (t > t,), the stress response is
 Viscoelasticproperties of edible lipids 309

(6)

A similar development was performed by Chen & Fridley (1972) yielding nearly
identical equations. Their development differed in that they used a generalized
Maxwell model which did not have a spring element in parallel with the series of
Maxwell bodies.
A relaxation time (r) is frequently used rather than the quantity vi/E, so that eqn
2 would read

In the case of a Maxwell body, when t = z the stress will have decayed to 37% of the
total stress relaxation. This describes why z is called the ‘relaxation time’; it is a
measure of how fast the stress decays. A material that has a short relaxation time
infers that the stress imposed by a strain input dissipates quickly and is therefore
more viscous than elastic in nature. The converse is true for long relaxation times.
One means of determining z is to measure it as the time when 63% of the stress
dissipation has occurred; however, it can be calculated by fitting the relaxation
modulus to stress relaxation data using non-linear regression. Equally, the coeffi-
cients ye and E can be calculated using non-linear regression and then 7 is simply
IIIE.
The objective of stress relaxation testing is to identify the elastic (spring) and
viscous (dashpot) constants, E, and vi. This is achieved by testing the materials
under a set of given conditions and then fitting eqn 5 and 6 to the loading and
relaxation data by adjusting the viscoelastic constants. However, difficulties can
occur in gathering and interpreting force-deformation data because of interactions
at the sample-platen interface. Bonding between the sample surface and the cross-
head platen can result in the cylindrical sample assuming a barrel shape, whose
surface is a paraboloid of revolution (Prentice, 1992). If the material can be bonded
to the platen using an adhesive, results may be more consistent than without
adhesion (Bagley & Christianson, 1987). Alternatively, the friction between the
sample and the platen can be eliminated by using a thin coating of paraffin or
silicone oil on a platen surface which is made of Teflon (Bagley et al., 1985). The
latter method was used in this study. For lubricated compression, stress on the
sample is calculated as recommended by Casiraghi et al. (1985).

MATERIALS AND METHODS

Stress relaxation sample preparation

Carnauba wax (Aldrich Chemical Company, Milwaukee, WI), candelilla wax (Strahl
& Pitsch Inc., West Babylon, NY), beeswax (Fisher Scientific Inc., Fair Lawn, NJ),
and a high-melting industrial milkfat fraction (VH-66, Center for Dairy Research,
University of Wisconsin, Madison) were melted in glass beakers immersed in a hot
310 lTH. Shellhammer et al.

water bath at 100°C. The molten lipids were poured into 10.0 mm i.d. by 10.2 mm
height cylindrical copper molds and allowed to cool until solidified. The beeswax
and the milkfat fraction were cooled to room temperature on a level granite slab
under ambient conditions. The carnauba and candelilla waxes had high melting
points: 82-85°C and 68-72°C respectively. Because of their volume change during
cooling and the large temperature difference between their melting points and the
room temperature, these materials would solidify too rapidly, producing irregularly
shaped cylinders under ambient conditions. To alleviate this problem, the molten
cylinders of wax were cooled slowly (approximately 0.5”C per minute) on a pro-
grammable hot plate (Dataplate@ Digital Hot Plate/Stirrer, Series 720, PMC
Industries, Inc., San Diego, CA). The slow cooling allowed the waxes to shrink
uniformly as they cooled. Once the lipids had solidified but not completely cooled,
the excess material was shaved from the top of the mold to produce uniform, right
cylinders which had smooth and parallel ends. The wax material slipped from the
copper mold once they had cooled to room temperature because of thermal shrink-
age. The milkfat fraction, which exhibited the least amount of shrinkage, was
removed by gently heating the copper sleeve between one’s fingers until the cylinder
slipped out into an ice bath. The lipid cylinders were stored at room temperature for
2 days prior to testing. The dimensions of the cylinders were measured with a
caliper micrometer after storage but prior to testing.

Stress relaxation data collection

At least 15 min prior to testing, individual lipid cylinders were placed in a water
bath held constant at 25fO.l”C. Tempered cylinders were removed from the water
bath and brushed dry with’ tissue paper. Each sample was placed between two
parallel polymethylmethacrylate plates mounted on a TA-XT2 Texture Analyzer
(Texture Technologies Corp., Scarsdale, New York). The surfaces of the plates were
lubricated with a thin film of silicone oil to prevent friction or bonding between the
lipid sample and the plate faces. The lipid cylinders were strained to 2% at a
crosshead speed of 1.0 mm per second, and the stress decay data (force, distance,
and time) were gathered for 140 s at a rate of 200 points per second. The plate faces
were cleaned and relubricated with silicone oil between each run. The room tem-
perature was monitored during testing to ensure that the environmental conditions
did not deviate significantly from the 25°C setpoint. During testing, the room varied
in temperature from a low of 23.5”C to a high of 26.2”C. Four individual samples of
each material were tested in a completely randomized fashion; however, in certain
cases samples failed during testing either due to cracking or incomplete straining
(~0.5%) and these data were eliminated. Incomplete straining was the result of
background vibrations in the TA-XT2 which triggered the commencement of data
collection prior to the crosshead actually contacting the sample surface. Additional
replicates of the failed sample were tested so that four replicates were tested for
each material.

Stress relaxation data reduction and analyses

The high data collection rate over the 140 s relaxation time produced nearly 30000
data points. To facilitate further analyses, a C-based data-reduction program was
written to reduce each data set to 106 points. Since much of the change in resistive
 Viscoelasticproperties of edible lipids 311

force during relaxation can occur shortly after compression, more data was retrieved
from the beginning of the relaxation than from the end. From each data set, the first
60 points were gathered over the first 0.3 s of the test without reduction and
included the compression (0.2 s) and initial relaxation (0.1 s) of the sample. The
next 17 points were gathered one point every 0.1 s and lasted from t = 0.3 to
t = 2.0 s. Another 17 points were gathered one point every 1.0 s from t = 2.0 to
t = 20 s. Finally, the last 13 points were gathered one point every 10 s from t = 20 s
until the completion of the test. This method of reduction provided a more manage-
able data set while minimizing the loss of information from the relaxation history.
The objective of the data analysis stage was to fit the relaxation modulus to the
stress relaxation data by adjusting the modulus coefficients. This was accomplished
by using a derivative-free, non-linear regression statistical package, BMDP AR
(BMDP, 1985). Initial estimates of the coefficients were required to use BMDP AR,
and these came as approximations derived using the method of successive residuals.
The method of successive residuals is a graphical technique used to approximate the
coefficient terms of an exponential equation (Mohsenin, 1986). A spreadsheet was
prepared (Microsoft Excel 4.0 for the Macintosh@) to calculate the exponential
terms’ coefficients using this technique. In addition to the estimates of the relaxation
modulus coefficients, a compressive elastic modulus was recorded from the loading
portion of the stress relaxation curves. A linear regression was fitted to stress versus
strain data and the slope of this regression was used as an estimate of the elastic
modulus.
In nearly all cases, the compression of the lipid cylinders did not immediately
produce a steady increase of force with time. In these instances, the force-time data
exhibited an initial toe prior to the linear increase in force from the compression of
the lipid cylinder (Fig. 1). The toe was likely due to compression of non-uniformities
on the lipid cylinder ends as the plunger face came into contact with the entire lipid
cylinder face. Once the two surfaces mated, the entire cylinder was strained. The

/ 0 Immediate deformation
o Delayed deformhon
I

5 5 tstrain t2
Time (set)
Fig. 1. Delayed deformation resulting from surface non-uniformities on the lipid cylinders.
312 7: H. Shellhammer et al.

TA-XT2 calculated the distance to strain each sample by measuring the sample’s
height at the moment the plunger came into contact and determined the correct
displacement based on a preset strain value (2% in this study). In the case of the
delayed deformation, the instrument thought it was straining the sample to 2%,
when in fact it was straining the sample to something less than 2%. The true strain
was calculated based on the time over which the linear increase in force occurred.
In Fig. 1, this time is the difference between tstrain and to for the data in which the
deformation begins the moment the plunger meets the lipid cylinder. In the case
where the data exhibited a toe (as does the delayed deformation data in Fig. l),
actual deformation of the entire cylinder was calculated as the distance the plunger
traveled during the time t1 to tz (i.e., tstrain, multiplied by the crosshead speed). The
time axis of the force-time graph was adjusted in these instances such that the
beginning of the stress relaxation test occurred at t, and not to.
Once the initial estimates of the relaxation modulus coefficients had been deter-
mined via the method of successive residuals, they were used as initial values for the
non-linear regression to fit eqn 5 and 6 to the stress relaxation data. These two
equations represent the stress response over different periods of the test. Eqn 5,
representing the compression phase, was fitted to the compression data, while eqn 6
was fitted to the subsequent relaxation portion of the same data. The coefficients of
the relaxation modulus are the same for both of these equations since they are
material specific, The non-linear regression package returned the ‘fitted’ coefficients
along with residual statistics such as mean squared error and a pseudo-?. The
experimental design for stress relaxation tests on individual lipid samples was a
single factor completely randomized experiment. Using these coefficients returned
from the non-linear regression, statistical differences among the four lipids were
determined using a single factor analysis of variance via Statistical Analysis Software
(SAS, 1989). Means comparisons were performed in the cases which were signifi-
cantly different using a Duncan’s multiple range test and a Tukey’s studentized
range test.

Cone penetration of lipih materials

Empirical measurements of fat ‘hardness’ have been standardized in a number of

cases (AOCS, 1960; ASTM, 1986, 1987, 1988, 1991). In the absence of the specific
cone penetrometer required by the ASTM methods, the TA-XT2 Texture Analyzer
was employed to measure resistive force to a 3 mm penetration by a 60” cone.
Samples were prepared by heating the lipid materials in a boiling water bath for at
least 15 min beyond melting. The molten lipid was poured into 5.0 cm i.d. by 1.0 cm
cylindrical cups and allowed to temper at room temperature for 24 h. Once the lipid
had crystallized during the cooling process, the exposed lipid surface was shaved
level with the edge of the cup so that a smooth, uniform surface was produced. An
hour prior to testing, the cups containing the lipids were placed in a water bath held
constant at 25°C.
Testing consisted of first removing the cup from the water bath and gently
brushing the surface dry with tissue paper. A 60” cone (2.6 cm height and 3 cm
maximum diameter), lubricated with a light coating of silicone oil, penetrated the
lipid sample at a rate of 0.2 mm/s to a distance of 3 mm. Force readings were
sampled at a rate of 400 points per second, and the maximum force at, or near,
3 mm penetration was recorded. The sample was placed back in the water bath prior
 Viscoelastic properties of edible lipids 31.3

to repetitive testing. Five replicate penetrations on the exposed surface of the lipid
in each cup were performed in a completed randomized fashion. The experimental
design for the penetration tests was similar to that of the stress relaxation study, a
completely randomized design. Data were analyzed using a single factor analysis of
variance and a multiple correlation via Statistical Analysis Software (SAS, 1989).
The high-melting waxes were brittle enough that a 3 mm penetration caused
cracking in some instances. A second test was performed in which the cone pene-
trated to a depth of only 2 mm. Force readings were recorded at 1 mm and at 2 mm.
As with the first test, five replicates of each material were tested in a completely
randomized fashion.

RESULTS AND DISCUSSION

Model selection for the relaxation modulus

The mathematical description of a viscoelastic material’s stress response to strain

input, eqn 5 and 6, was derived using a generalized Maxwell model as the relaxation
modulus. Before eqn 5 and 6 could be fitted to stress relaxation data, a specific
model (rather than the generalized Maxwell model) had to be chosen for the
relaxation modulus. The question was how many Maxwell bodies to include and
whether to include the spring element. Peleg (1976) lists several requirements for
constructing a rheological model of a food material. The model must be able to
predict real material behavior, it should be able to respond to both positive and
negative forces, and changes in the material’s behavior must be explainable in terms
of the model parameters. These requirements suggested reaching a balance between
choosing a model with as few terms as possible with one that has many. For the sake
of simplicity, as well as being able to explain changes in the material’s behavior, a
model should have few terms. On the other hand, a model with many terms will
yield predictable results with high accuracy.
Initial tests of the viscoelastic properties of the lipid materials using small strains
pointed towards a five element model (Fig. 2) because none of the lipid materials
relaxed completely such that the stress went to zero. The elastic spring element in
parallel with the Maxwell bodies accounts for this residual stress in the lipid, thus it
was kept as part of the relaxation modulus. Determining how many Maxwell bodies
to be included was based on a degree-of-fit of predicted stress response to the actual
stress response. A model with one spring element and one Maxwell element in
parallel yielded r’ values from the non-linear regression which ranged from 0.926 to
0.974 depending on the lipid type. Adding another Maxwell body, as shown in Fig.
2, yielded significantly improved 2 values which ranged from 0.991 to 0.998. Models
with more elements yielded slightly higher ? values, as would be expected since the
number of parameters in the model increased. Yet, the authors felt that a five
parameter model would be optimal by achieving the previously mentioned balance
of model size. A similar model was used by Sato and Nakayama (1970) to describe
the mechanical behavior of white chicken meat. The five-parameter relaxation mod-
ulus is written as
6, F.-
I
(b(t) = Eo+EIe T’+E2e ‘I, (8)
314 T H. Shellhammer et al.

Fig. 2. Model used for the relaxation modulus for the lipid materials. Stress, 0 (kPa); strain,
F.(unitless); spring constants, Ej (kPa); and dashpot viscosities, vi (kPa s).

Adjustments to strain

Achieving a 2% strain in all samples was difficult because of the inability to achieve
a true right cylinder in each of the samples and because of irregularities in the
samples’ surfaces. The actual distance each sample was strained was calculated as
described in the methodology section. These values ranged from 1.52% to 1.93%.
Averaged among the four replicates of each material, the true strain ranged from a
low of 1.6% for carnauba and candelilla waxes to a high of 1.8% for the milkfat
fraction.

Stress relaxation of the lipid materials

The parameters in eqn 8, EiS and qis, for each replicate tested were averaged to
produce Table 2 and Fig. 3. The individual parameters were compared using a
single-factor analysis of variance and means testing, the results of which are
included in Table 2. Large differences in the relaxation modulus coefficients were
observed among the four lipid types. Carnauba and candelilla wax behaved relatively
similarly as hard, non-deformable materials following the 2% strain input (Fig. 3
and Table 2). In comparison to the candelilla wax, the carnauba wax displayed a

TABLE 2
Averaged Actual Strain and Relaxation Modulus Coefficients of Selected Lipids
Material Corrected EI ‘II vr
strain Pa) (Pas) (Pa s) $ 2
Milkfat fraction 0.018 13517” 3722” 4831” 2092” 201420” 0.565” 40.69”
Beeswax 0.017 21347b 7589’ 5507” 1206h 78 987’ 0.161’ 14.31h
Candelilla wax 0.016 34830’ 2094’ 1917h 706’ 3961 lb 0.352’ 20.81h
Carmaiba wax 0.016 34 270’ 3880” 1647’ 282d 10459h 0.072h 6.71h
Means (n = 4) with different superscripts are significantly different at a = 0.05.
 Viscoelastic properties of edible lipids 315

rapid relaxation immediately upon loading (Fig. 4). Analysis of variance indicated
that these two materials were not significantly different in terms of residual stress
(E,,) and f2 (Table 2) while they were significantly different with respect to z,. It is
also clear from Fig. 4 that the peak stress upon loading was very different among
the four lipids. The large differences in the degree to which these waxes resisted
deformation made comparisons of their relaxation behavior difficult. To improve the
contrast in relaxation behavior, the stress data were normalized by dividing each
stress data point by the maximum stress during the test. In this manner, the normal-
ized stress values ranged from zero to one for all of the materials (Fig. 5).
The four materials appear as two similar groups in Fig. 5. Carnauba and cande-
lilla wax were hard materials resisting deformation. They had very high peak stresses
(Figs 3 and 4) and dissipated less than 9% of this stress. The beeswax and milkfat
fraction, on the other hand, had much lower peak stress, and they yielded much
more of this stress during relaxation. It is clear from Figs 4 and 5 that carnauba wax
exhibited the fastest relaxation times along with a very small degree of relaxation,
thus it was the most elastic material of the four. The candelilla wax relaxed over a
much longer period of time than the carnauba wax, as indicated by a significantly

I Beeswax I
1
l

i
T
T Milkfat Fraction
I a

lb0
Time (set)

Fig. 3. Average stress relaxation response following a 2% compressive strain of the bulk lipid
materials (n = 4). Solid lines represent fitted values based on eqn 5 and 6 using average
coefficients from Table 2. Due to the time scale, the loading portion of the stress history has
been removed.
316 T. H. Shellhammer et al.

larger z1 and a considerably larger r2. These results indicate that candelilla wax had
more viscous behavior than the carnauba wax, yet both behaved more as elastic
materials as compared to the beeswax and the milkfat fraction.
The other group in Fig. 5 included beeswax and the milkfat fraction. One would
expect the milkfat fraction to be less elastic and more viscous than the high-melting
waxes. The milkfat fraction is comprised of triglycerides which are made in large
part by saturated fatty acids, principally palmitic and stearic fatty acids. The milkfat
fraction also contains unsaturated fatty acids, principally oleic acid (Jenness &
Patton, 1959); thus a portion of the milkfat fraction is liquid at room temperature.
Beeswax, on the other hand, is comprised of esters of rather long fatty acid and fatty
alcohol, hence its elevated melting point. It resisted deformation but also exhibited
a large degree of stress relaxation. The presence of a significant amount of free fatty
acids contributed to its viscous qualities.

Penetrometer results

Lipid type had a significant effect (p <O.OOl) on resistance to penetration with a 60

cone (Fig. 6). At any of the three depths, the beeswax and milkfat fraction had

I , I

0.25 0.5 0.75

Time (set)
Fig. 4. Compression and initial stress relaxation following a 2% compressive strain of the
bulk lipid materials (n = 4). Solid lines represent fitted values based on eqn 5 and 6 using
average coefficients from Table 2.
 Viscoelasticproperties of edible lipids 317

resistive forces which were not significantly different, while carnauba wax and cande-
lilla wax were significantly different from each other and from the beeswax and
milkfat fraction group (p < 0.05). Differences that might have stemmed from crack-
ing of the lipid material during penetration did not appear outstanding, since the
trend in hardness was similar at all three depths. The reason that cracking may not
have been as large a problem as initially expected lies with the experimental pro-
cedure. The lipid materials were tested in the aluminum cups in which they were
cast. When the high-melting waxes cracked during deep penetration, the various
segments of the cracked lipid disk were contained by the mold and continued to
offer resistance as the cone penetrated further into the newly formed crack.
One concern with gathering viscoelastic property data is whether the information
is of more value than data from simpler tests such as cone penetration. Usually, a
cone test gives a composite parameter and not a true property of the material being
tested. With regard to viscoelastic data, a stress relaxation test yields more informa-
tion than a penetrometer (Shama & Sherman, 1970). As for discriminating
differences among lipids, this may not always be the case. For example, hardness as
measured by the cone penetrometer might correlate well with degree of elastic
behavior. To examine whether this was the case in this experiment, a multiple

100

Time (set)
Fig. 5. Normalized stress relaxation response following a 2% compressive strain of the bulk
lipid materials (n = 4). The loading portion of the stress history has been removed from this
figure.
318 7: H. Shellhammer et al.

correlation was performed on the hardness and viscoelastic data. The force readings
from the penetrometer did not correlate well with most of the relaxation modulus
coefficients. The best correlations with the penetrometer readings and the relaxation
coefficients were with E0 (r = 0.9372) and E2 (r = -0.9547). The magnitude of E2
values was inversely related to measures of hardness from the penetrometer; but the
results for E2 fell into two groups (Table 2), while the results from the penetrometer
fell into three separate groups. Based on the correlation analysis, it is safe to state
that the results from penetrometer testing were different from the stress relaxation
analyses, and that these two tests were not redundant measures of the rheological
properties of the four materials tested in this study.

SUMMARY AND CONCLUSIONS

A five-element model, one spring and two Maxwell bodies in parallel, was chosen
for the relaxation modulus of viscoelastic lipid film materials. Parameters for this
relaxation modulus were identified by fitting equations which described the stress
response to a strain input over the loading and relaxation portions of the data.
Results of the stress relaxation analyses indicate that carnauba wax and candelilla
wax were very elastic, non-deformable materials. Penetrometer results indicated that
carnauba wax was the hardest wax of the four tested. Beeswax and the milkfat

q Imm penetration

250 - q Zmm penetration

n 3mmpenetration

200-

iz 150-
0
2
IE
loo-

50-

o-
Milkfat Fractiona Beeswax a Candelilla wax’ Camauba waxE
Fig. 6. Resistive force of lipid materials to cone penetration (n = 5). Lipids with different
superscripts are significantly different at tl = 0.05 at all three depths of penetration.
 Viscoelasticproperties of edible lipids 319

fraction were of similar hardness and exhibited similar relaxation behavior to each
other, but were significantly softer and more deformable than either the carnauba or
candelilla waxes. The milkfat fraction was the most viscoelastic material of the four
lipids tested as characterized by its long relaxation times, and the smallest amount
of residual stress remaining in the sample following relaxation. These later two
materials could be characterized as soft, viscous lipids which offer much less resist-
ance to deformation than the high-melting waxes.
The performance of these materials as edible lipid coatings would be dramatically
affected by their mechanical properties. The greater viscoelasticity of the beeswax
and milkfat fraction means these materials would be more flexible and less likely to
crack than the more elastic materials. Whether beeswax or milkfat would be better
films from this standpoint remains unknown. Comparative studies of the barrier and
mechanical properties of multicomponent films containing these lipids are needed to
verify that these latter two materials are more suitable as film forming materials
than carnauba wax or candelilla wax.

ACKNOWLEDGEMENTS

This work was supported by the California Milk Advisory Board and Dairy Manage-
ment, Inc., through the California Dairy Research Foundation and the California
Dairy Foods Research Center.

REFERENCES

AOCS (1960). Oficial and Tentative Methods: Additions and Revision, Method Cc. 16-60.
American Oil Chemist’s Society (reapproved 1973).
ASTM (1986). Designation 1321: standard test methods for needle penetration of petroleum
waxes. In Annual Book of ASTM Standards. American Society for Testing and Materials,
Philadelphia, PA.
ASTM (1987). Designation 937: standard test methods for cone penetration of petrolatum.
In Annual Book of ASTM Standards. American Society for Testing and Materials, Phila-
delphia, PA.
ASTM (1988). Designation 217: standard test methods for cone penetration of lubricating
grease. In Annual Book of ASTM Standards. American Society for Testing and Materials,
Philadelphia, PA.
ASTM (1991). Designation 1403: standard test methods for cone penetration of lubricating
grease using one-quarter and one-half scale cone equipment. In Annual Book of ASTM
Standards. American Society for Testing and Materials, Philadelphia, PA.
Baer, R. J. (1991). Alteration of the fatty acid content of milk fat. J. Food Protect., 54, 5
383-386.
Bagley, E. B. & Christianson, D. D. (1987). Stress relaxation of chemically leavened dough:
data reduction using the BKZ elastic fluid theory. J. Rheol., 31, 404-413.
Bagley, E. B., Wolf, W. J. & Christianson, D. D. (1985). Effect of sample dimensions,
lubrication and deformation rate on uniaxial compression of gelation gels. Rheol. Acta, 24,
265-271.
Banks, W. (1991). Milk fat. J. Sot. Dairy Technol., 44, 2 31-32.
Bennett, H. (1975). Industrial Waxes. Chemical Publishing Company, New York.
BMDP StatisticalSoftware (1985). University of California Press, Berkeley, CA.
320 7:H. Shellhammer et al.

Casiraghi, E. M., Bagley, E. B. & Christianson, D. D. (1985). Behavior of mozzarella,

cheddar and processed cheese spread in lubricated and bonded uniaxial compression. J.
Texture Studies, 16, 281-301.
Chen, P. & Fridley, R. B. (1972). Analytical method for determining viscoelastic constants of
agricultural materials. Trans. ASAE, 15,6 1103-l 106.
Davey, K. R. (1989). Temperature dependence of flow for plastic fats from penetration data.
J. Texture Studies, 19, 4 397-405.
deMan, J. M. (1976). Texture of fats and fat products. In Rheology and Tenture in Food
Quality, eds J. M. deMan, P. W. Voisey, V. F. Rasper and D. W. Stanley. AVI Publishing
Company, Westport, CT.
deMan, J. M. & Finoro, M. (1980). Characteristics of milk fat fractionated by crystallization
from the melt. Can. Inst. Food Technol. J., 13, 4 167-173.
deMan, J. M., Gupta, S., Kloek, M. & Timbers, G. E. (1985). Viscoelastic properties of
plastic fat products, J. Am. Oil Chemists Sot., 62, 12 1672-1675.
Findely, W. N., Lai, J. S. & Onaran, K. (1976). Creep and Relaxation of Nonlinear Viscoelastic
Materials. North-Holland, Amsterdam.
Fltigge, W. (1975). Viscoelasticity,2nd rev. ed. Springer-Verlag, Berlin.
Greener, I. K. (1992). Physical properties of edible films and their components. Ph.D. thesis,
University of Wisconsin, Madison.
Hayakawa, M. & deMan, J. M. (1982). Interpretation of cone penetrometer consistency
measurement of fats. J. Tature Studies, 13, 2 201-210.
Jenness, R. & Patton, S. (1959). Principles of Dairy Chemistry. Wiley, New York.
Kaylegian, K. E. (1993). Experimental Milqat Fraction Technical Data Sheet: Milkfat Fraction
W-66,42,45. Center for Dairy Research, University of Wisconsin, Madison.
McHugh, T. H. & Krochta, J. M. (1994). Water vapor permeability properties of edible whey
protein-lipid emulsion films. J. Am. Oil Chemists Sot., 71, 3 307-312.
Mohsenin, N. N. (1986). Physical Properties of Plant and Animal Materials. Gordon & Breach,
New York.
Mohsenin, N. N. & Mittal, J. P. (1977). Use of rheological terms and correlation of compat-
ible measurements in food texture research. J. Texture Studies, 8, 395-408.
Mortensen, B. K. (1983). Physical properties and modification of milk fat. In Developments in
Dairy Chemist9 2: Lipids, ed. P. F. Fox. Applied Science Publishers, London, pp. 159-194.
Peleg, M. (1976). Considerations of a general rheological model for the mechanical behavior
of viscoelastic solid food materials. J. Texture Studies, 7, 243-255.
Prentice, J. H. (1992). Dairy Rheology: A Concise Guide. VCH Publishers, New York.
SAS Institute Inc. (1989) SASISTAP Users Guide, Version 6, 4th ed. SAS Institute, Cary,
NC.
Sato, Y. & Nakayama, T. (1970). Discussion of the binding quality of minced meats based on
their rheological properties before and after heating. J. Texture Studies, 1, 309-326.
Shama, F. & Sherman, P. (1970). The influence of work softening on the viscoelastic proper-
ties of butter and margarine. J. Texture Studies, 1, 1 196-205.
Shellhammer, T. S., Fairley, P., German, J. B. & Krochta, J. M. (1993). Water vapor barrier
properties of simulated milkfat fractions. Annual Meeting of the Institute of Food Tech-
nologists, Chicago, IL, lo-14 July.
Skinner, G. E. & Rao, V. N. M. (1986). Linear viscoelastic behavior of frankfurters. J. Tatwe
Studies, 17, 421-432.
Streitwieser, A. & Heathcock, C. H. (1985). Zntroduction to Organic Chemistry, 3rd ed.
Macmillan, New York.
Tanaka, M., DeMan, J. M. & Voisey, P. W. (1971). Measurement of textural properties of
foods with a constant speed cone penetrometer. J. Texture Studies, 2, 1 306-315.