Time-dependent viscosity of stirred yogurt.

Part I: couette ow
H.J. O'Donnell, F. Butler
*
Department of Agricultural and Food Engineering, University College Dublin, Earlsfort Terrace, Dublin 2, Ireland
Received 8 September 2000; accepted 19 March 2001
Abstract
Viscosity data at 5C are presented for commercial yogurt for a shear rate range 5700 s
1
. Data were obtained using a con-
ventional rotational rheometer. When subjected to a range of constant shear rates, yogurt viscosity demonstrated time-dependent and
shear-dependent behaviour. The equilibrium structural parameter employed in the characterisation of the time-dependent nature of
the yogurt was found to vary over the shear rate range investigated. At both the initial and equilibriumconditions, the shear rate/shear
stress data were tted to an Ostwald power law model of the form s = K _ c
n
with good correlation (average r
2
= 0:97). Experimental
shear stress/time data at constant shear rate were modelled using a structural parameter approach and using the Weltmann model. The
experimental shear stress data was best described by the Weltmann model. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Structural parameter; Yogurt; Time-dependent viscosity
1. Introduction
Yogurt is produced by a fermentation process during
which a weak protein gel develops due to a decrease in
the pH of the milk. The pH of the milk is decreased due
to the conversion of lactose to lactic acid by the fer-
mentation culture bacteria. In liquid milk, casein mi-
celles are present as individual units. As the pH
approaches pH 5.0, the casein micelles are partially de-
stabilised and become linked to each other in the form
of aggregates and chains which form part of a three-
dimensional protein matrix in which the liquid phase of
the milk is immobilised. This gel structure contributes
substantially to the overall texture and organoleptic
properties of yogurt and gives rise to shear and time-
dependent viscosity. A large number of studies have
been performed to characterise the viscosity of yogurt
(Steventon, Parkinson, Fryer, & Bottomley, 1990; Ra-
maswamy & Basak, 1991a,b; Rohm, 1992; Benezech and
Maingonnat, 1993; Skriver, Roemer, & Qvist, 1993; De
Lorenzi, Pricl, & Torriano, 1995; Chan Man Fong,
Turcotte, & De Kee, 1996).
A number of authors have characterised the time-de-
pendent viscosity of food products (Tiu & Boger, 1974;
De Kee, Code, & Turcotte, 1983; Ramaswamy & Basak,
1991b; Benezech & Maingonnat, 1993; Alonso, Larrode,
& Zapico, 1995; Chan Man Fong et al., 1996). Tiu and
Boger (1974) employed a structural approach developed
by Cheng and Evans (1965) which included a structural
parameter k (0 Tk T1) which is an index of the relative
structural integrity of the sample. The assumption that
the value of k at equilibrium, k
e
, is constant and inde-
pendent of shear rate is central to that model. Another
approach, applied by Ramaswamy and Basak (1991b),
employed a model developed by Weltmann (1943) which
described time-dependent viscosity at constant shear rate
in terms of a logarithmic time model. Other models have
also been reported which describe the time dependency of
yogurt (De Kee et al., 1983; Chan Man Fong et al., 1996).
In this study, the time-dependent viscosity of yogurt
when subjected to a range of constant shear rates at 5C
was investigated. The objective of this work was to in-
vestigate the suitability of a structural model approach
(Tiu & Boger, 1974) and the Weltmann model (Welt-
mann, 1943) to characterise the time-dependent beha-
viour of yogurt. The aim was to use the viscosity data to
model the tube ow of yogurt. This work is described in
the second paper.
2. Methods and materials
Batches of stirred natural yogurt in 500 g retail
containers were purchased directly from the manu-
Journal of Food Engineering 51 (2002) 249254
www.elsevier.com/locate/jfoodeng
*
Corresponding author. Tel.: +353-1-716-7473; fax: +353-1-475-
2119.
E-mail address: f.butler@ucd.ie (F. Butler).
0260-8774/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S0260- 8774( 01) 00064- 4
facturer. Each batch consisted of approximately 150
containers of yogurt. All batches of yogurt were tested
within a week of manufacture and at least two weeks
before their maximum shelf-life date. The experiments
were performed over a period of three weeks with
three product batches used in total. All viscosity ex-
periments were repeated six times (3 batches 2).
Prior to testing, yogurt samples were stored at 4C in
a cold room until required. The pH of the yogurt was
monitored during storage using a Unicam pH meter
(model 9450, Unicam Analytical Systems, Cambridge,
UK). A total of 8 pH measurements were performed
on each batch of yogurt. A fresh container of yogurt
was used for each measurement. Protein content was
determined by the Kjeldahl method (IDF Standard
20B, 1993) (four measurements per batch). Fat con-
tent was determined by the Gerber method (AOAC,
1980) using 11.3 g of yogurt (3 per batch). Total
solids content (IDF Standard 151, 1991) (3 per batch)
was also determined.
The rotational rheometer used to measure viscosity
was a Physica Systems, Rheolab MC 100 (Physica
Metechnik, Stuttgart, Germany). The test geometry
was a concentric cylinder system conforming to
DIN53019, with a bob radius of 22.5 mm and a cup
radius of 24.4 mm. The bob length was 67.5 mm. Each
test required approximately 120 ml of yogurt. All ex-
periments were performed at 5C and a fresh container
of yogurt was used for each experiment. To minimise
damage to the yogurt structure prior to shearing,
samples were carefully poured into the cup before
lowering the bob. Samples were allowed to stand for 15
min prior to shearing. In order to determine the time-
dependent viscosity of yogurt, the samples were then
subjected to a constant shear rate for 50 min at shear
rates of 5, 8, 10, 15, 50, 100, 250, 500, and 700 s
1
.
Viscosity readings were recorded every 4 s for the rst
400 s of each run and every 26 s thereafter. The order
of the constant shear rate tests was randomised for
each replicate.
3. Theory
Cheng and Evans (1965) developed a structural the-
ory to describe thixotropy which stated that viscosity
was a function of both shear rate and a time-dependent
structural parameter (k). Tiu and Boger (1974) simpli-
ed this structural theory by using a Hershel Bulkey
model modied to include k. Their equation of state was
s = ks
0
: (1)
The shear stress at zero time of shear, s
0
was given by a
Hershel Bulkley model
s
0
= s
y
K _ c
n
(2)
Eq. (2) includes a yield stress term. Barnes and Walters
(1985) suggested that, except in a few limited circum-
stances, yield stress does not exist, i.e., viscosity is al-
ways nite. While it is recognised that yield stress is a
useful parameter in some practical applications, it is not
included in the analysis of the present data. Therefore,
Eq. (2) can be replaced by the Ostwald power law
model:
s
0
= K _ c
n
: (3)
Combining Eqs. (1) and (3) yields
s = kK _ c
n
: (4)
For their rate equation Tiu and Boger employed a sec-
ond-order kinetic equation developed by Petrellis and
Flumerfelt (1973)
dk
dt
= k
1
(k k
e
)
2
for k > k
e
; (5)
where the structural parameter, k, ranged from an initial
value of unity at zero shear time to an equilibrium value
of k
e
that is less than unity. The rate constant, k
1
is a
function of shear rate and has to be determined exper-
imentally. To determine k
1
experimentally, Eq. (5) can
be integrated analytically under conditions of constant
shear to yield
Nomenclature
a constant in exponential decay equation used to determine
equilibrium viscosity (Pa s)
A intercept in Weltmann model (Pa)
a
1
rate constant given by the slope of the line of best t
for a plot of 1=(g g
e
) versus t at a given shear
rate (Pa
1
s
2
)
b reciprocal time constant in exponential decay equation used
to determine equilibrium viscosity (s
1
)
B slope in Weltmann model (Pa)
K consistency index in power law model (Pa s
n
)
k
1
rate constant for the decay of the structural parameter with
time (s
1
)
n ow behaviour index in power law model
t time (s)
t
m
time at which maximum shear stress is measured (s)
_ c shear rate (s
1
)
g viscosity (Pa s)
g
e
viscosity at equilibrium conditions (Pa s)
g
0
viscosity at zero time of shear (Pa s)
k structural parameter
k
e
structural parameter at equilibrium conditions
k
0
structural parameter at zero time of shear
s shear stress (Pa)
s
0
shear stress at zero time of shear (Pa)
s
e
shear stress at equilibrium conditions (Pa)
s
y
yield stress (Pa)
250 H.J. O'Donnell, F. Butler / Journal of Food Engineering 51 (2002) 249254
1
k k
e
=
1
k
0
k
e
k
1
t: (6)
The instantaneous apparent viscosity for any uid may
be dened by the equation
g =
s
c
v
: (7)
Combining Eqs. (1) and (7) yields
k =
gc
v
s
0
: (8)
Eq. (8) is also valid at initial and equilibrium conditions
in which case k and g are replaced by k
0
and g
0
and k
e
and g
e
, respectively.
Substituting Eq. (8) into Eq. (6) yields
1
g g
e
=
1
g
0
g
e
a
1
t; (9)
where
a
1
=
k
1
c
v
s
0
: (10)
Therefore, for a given shear rate, a plot of 1=(g g
e
)
versus t should yield a straight line with a slope equal to
a
1
. Repeating the same procedure at other shear rates
will establish the relationship between a
1
and _ c, and
hence k
1
and _ c from Eq. (10).
4. Results and discussion
The average protein content of the yogurt was 4.4%
(S.D. 0.08), fat content 3.0% (S.D. 0.1), solids content
13.8% (S.D. 0.1), pH 3.94 (S.D. 0.05). The results were
typical for commercial yogurt (Tamime & Robinson,
1985) and there was no appreciable dierence in com-
position between batches.
The average viscosity as a function of time of shear at
5C for each shear rate employed is shown in Fig. 1. The
data clearly show the shear-dependent and time-depen-
dent viscosity of yogurt. Time dependence was particu-
larly signicant during the initial stages of shearing. As
the time of shearing approached 50 min, the rate at
which viscosity decreased as a function of time had re-
duced to a low level. However, even after 50 min of
shearing, an equilibrium viscosity had still not been
achieved. Both Ramaswamy and Basak (1991b) and
Butler and McNulty (1995) have reported similar nd-
ings after one hour of shearing for stirred yogurt and
buttermilk, respectively. Benezech and Maingonnat
(1993) assumed an equilibrium value at approximately
600 s for yogurt whereas Chan Man Fong et al. (1996)
reported equilibrium values for yogurt at times ranging
from a few seconds at high shear rates to several minutes
at low shear rates. Tiu and Boger (1974) reported that
the apparent viscosity of mayonnaise no longer ap-
peared to vary with time of shearing for values of time
exceeding 40 min.
Initial viscosity (t = 4 s), g
0
, was obtained from the
viscosity versus time data shown in Fig. 1. The equi-
librium viscosity, g
e
, was determined as outlined in
Butler and McNulty (1995) whereby the latter portions
of the viscosity curves shown in Fig. 1 were tted to an
exponential decay curve of the form
g = g
e
ae
bt
; (11)
where g
e
was determined by choosing a trial value of g
e
and calculating the line of best t between log(g g
e
)
and t. The value of g
e
that gave the best correlation was
selected. For the shear rate range investigated, the val-
ues of r
2
were all greater than 0.999. In this investiga-
tion, the time at which the rst viscosity measurement
was taken (t = 4 s) was considered to be zero time (t
0
).
This choice of initial time value was similar to the initial
time values reported by other workers (Ramaswamy &
Basak, 1991b; Benezech & Maingonnat, 1993; Butler &
McNulty, 1995).
Values of a
1
and k
1
were determined as outlined in
Butler and McNulty (1995) using the rst 25 min of
data. Both a
1
and k
1
were found to increase exponen-
tially with shear rate. The following exponential power
law models were found to describe the dependence of a
1
and k
1
on shear rate:
a
1
= 9:5 10
5
_ c
0:9
; r
2
= 0:99; (12)
k
1
= 2:7 10
3
_ c
0:19
; r
2
= 0:98: (13)
Previous workers (Tiu & Boger, 1974; De Kee et al.,
1983; Benezech & Maingonnat, 1993) have also shown
that for yogurt and mayonnaise, a
1
and k
1
could be re-
lated to shear rate using power law models. However, as
the variation in the value of k
1
over the shear rate range
investigated was small, these authors suggested that k
1
Fig. 1. Viscosity of stirred natural yogurt at 5C as a function of time
of exposure to constant shear. Data were recorded every 4 s for the rst
400 s and every 26 s for the remaining 2600 s.
H.J. O'Donnell, F. Butler / Journal of Food Engineering 51 (2002) 249254 251
was almost independent of shear rate. Butler and
McNulty (1995) found k
1
to be almost independent of
shear rate for buttermilk and employed a constant value
of k
1
in Eq. (5), the rate equation. In this investigation,
the value of k
1
increased by a factor of 2.4 over the shear
rate range investigated and this was considered sucient
change to eliminate the possibility of considering k
1
to
be a constant.
Shear stress values corresponding to viscosity values
at t = 4 s and t were determined using Eq. (7). A
plot of shear stress versus shear rate is shown in Fig. 2.
An examination of Fig. 2 shows that the plots for s
0
and
s
e
are not parallel as would be expected from the Tiu
and Boger model. A statistical procedure (Mead, Cur-
now, & Hasted, 1993) for tting parallel lines demon-
strated that the probability that the slopes were dierent
was highly signicant (P < 0:001). As the slopes were
not parallel, the ratio of s
e
: s
0
was not a constant and
hence, from Eq. (1), k
e
was not constant. Two power law
models were found to describe the relationship between
shear rate and shear stress at the initial (t = 4 s) and
equilibrium conditions:
s
0
= 28:39 _ c
0:285
; r
2
= 0:99; (14)
s
e
= 14:92 _ c
0:126
; r
2
= 0:95: (15)
The value of the structural parameter at the equilibrium
viscosity values, k
e
, was determined at each shear rate
employed using Eq. (1). Previous workers (Tiu & Boger,
1974; De Kee et al., 1983; Benezech & Maingonnat,
1993; Butler & McNulty, 1995) have employed a single
average value of k
e
. However, in this work, the value of
k
e
was found to decrease (range 0.450.19) with in-
creasing shear rate. A power law model where
k
e
= 0:526 _ c
0:16
; r
2
= 0:96 (16)
was found to describe the relationship between k
e
and _ c.
A plot of k
e
as a function of shear rate together with the
corresponding power law approximation is shown in
Fig. 3.
Although Eq. (5), the rate equation, cannot be inte-
grated analytically for varying shear rate conditions,
progress can still be made in predicting shear stress at
any given constant shear rate. Two approaches for
modelling shear stress/time data were considered, a
structural parameter model and the Weltmann model.
In order to derive an equation that would predict the
structural parameter, k, for a constant shear rate after a
given time, Eq. (6) was rearranged to yield
k = k
e

1
1=(1 k
e
) k
1
t
: (17)
Combining Eqs. (1), (14), (16) and (17) yielded an
equation that predicted values of shear stress after a
given time of shearing at a constant shear rate. A
comparison of experimental shear stresses and predicted
shear stresses is presented in Figs. 46 for shear rates of
5, 50 and 500 s
1
.
For stirred yogurt, Benezech and Maingonnat (1993)
reported a value of 0.3 for k
e
which is similar to the
average value of 0.22 which was determined in this work
by integrating equation (16) and calculating an average
value for k
e
over the range of shear rates investigated.
Predicted values of shear stress using the structural ap-
proach with average values of k
e
(0.22) and k
1
(0.0079)
are presented in Figs. 46. Using this average value of
k
e
, the experimentally measured values of shear stress at
shear rates less than 100 s
1
would be underestimated by
2550%. The error increased with decreasing shear rate
and was particularly evident at shear rates between 5
and 15 s
1
. The dierence in shear rate ranges employed
may provide an explanation for the contrast between
this work and previous work regarding the variation of
k
e
with shear rate. Benezech and Maingonnat (1993)
investigated a more limited range of shear rates
Fig. 2. Shear stress versus shear rate at t = 4 s (j) and t (d) at
5C determined from the viscosity versus time of shear data presented
in Fig. 1 together with power law approximations.
Fig. 3. Equilibrium structural parameter versus shear rate at 5C
together with the power law approximation.
252 H.J. O'Donnell, F. Butler / Journal of Food Engineering 51 (2002) 249254
(18280 s
1
) than was employed in the work presented
here (5700 s
1
).
The second approach to model shear stress/time data
employed was the logarithmic time model proposed by
Weltmann (1943) and previously used to describe the
stress decay behaviour of time-dependent foods (Ra-
maswamy & Basak, 1991b; Alonso et al., 1995) where
s = A B[ln(t=t
m
)[: (18)
This model was found to give good correlations when
applied to the data (average r
2
= 0:97). For stirred yo-
gurt, Ramaswamy and Basak (1991b) used linear mod-
els to characterise variations in A and B as a function of
shear rate for the range of shear rates employed (100
500 s
1
). The results presented here agree with this
nding only in the range 100700 s
1
(A : r
2
=
0:98; B : r
2
= 0:97). However, the following power law
and logarithmic models were found to best describe the
behaviour of A and B, respectively, over the full range of
shear rates used, i.e., 5700 s
1
:
A = 37:18 _ c
0:19
; r
2
= 0:99; (19)
B = 2:01 ln( _ c) 0:549; r
2
= 0:98: (20)
Using these values of A and B in Eq. (18), predicted
values of shear stress as a function of time of shearing
were determined and are shown in Figs. 46.
Analysis of residuals can be used to compare the
accuracy of values predicted by models relative to the
experimental values (Harr od, 1989). The sum of squared
residuals (SSRs) for predicted values of shear stress
compared to experimentally measured values of shear
stress in the shear rate range 5700 s
1
was calculated
for the three models used. The models predicted the
experimental values with varying degrees of success. The
greatest errors occurred using the structural model
where average values of k
1
and k
e
were employed (SSR
373 10
3
Pa
2
). Using the structural model where k
e
and
k
1
were functions of shear rate improved the t (SSR
260 10
3
Pa
2
) but the error was still large. The SSR for
the Weltmann model (SSR 17 10
3
Pa
2
) was substan-
tially lower than those for the structural models indi-
cating that this model was more suitable for tting the
measured data.
Fig. 5. Shear stress versus time at a constant shear rate of 50 s
1
: (full
line) measured experimentally; modelled (x) using structural param-
eter approach where k
e
and k
1
are functions of shear rate; modelled (d)
using the structural parameter approach where average values of k
e
and k
1
are employed; modelled () using the Weltmann approach. For
clarity, not all predicted values are plotted.
Fig. 6. Shear stress versus time at a constant shear rate of 500 s
1
: (full
line) measured experimentally; modelled (x) using structural param-
eter approach where k
e
and k
1
are functions of shear rate; modelled (d)
using the structural parameter approach where average values of k
e
and k
1
are employed; modelled () using the Weltmann approach. For
clarity, not all predicted values are plotted.
Fig. 4. Shear stress versus time at a constant shear rate of 5 s
1
: (full
line) measured experimentally; modelled (x) using structural param-
eter approach where k
e
and k
1
are functions of shear rate; modelled (d)
using the structural parameter approach where average values of k
e
and k
1
are employed; modelled () using the Weltmann approach. For
clarity, not all predicted values are plotted.
H.J. O'Donnell, F. Butler / Journal of Food Engineering 51 (2002) 249254 253
5. Conclusions
The Weltmann model gave good predictions of the
experimentally measured shear stress for controlled
shear conditions. In the case of the structural model, for
the shear rate range investigated, the equilibrium
structural parameter, k
e
, and the rate constant, k
1
, were
both found to be power functions of shear rate. Hence,
Eq. (5), the rate equation, cannot be easily integrated to
solve for practical ow conditions where shear rate is
varying.
Acknowledgements
This research has been part-funded under the Food
Sub-Programme of the Operational Programme for In-
dustrial Development (Administered by the Irish De-
partment of Agriculture and Food and supported by
Irish and EU funds).
References
Alonso, M. L., Larrode, O., & Zapico, J. (1995). Rheological
behaviour of infant foods. Journal of Texture Studies, 26, 193202.
Association of ocial analytical chemists (AOAC) (1980). Ocial
methods of analysis (13th ed.). Arlington, Virginia: AOAC.
Barnes, H. A., & Walters, K. (1985). The yield stress myth? Rheologica
Acta., 24, 323326.
Benezech, T., & Maingonnat, J. F. (1993). Flow properties of stirred
yogurt: structural parameter approach in describing time-depen-
dency. Journal of Texture Studies, 24, 455473.
Butler, F., & McNulty, K. (1995). Time dependent rheological
characterisation of buttermilk at 5C. Journal of Food Engineering,
25, 569580.
Chan Man Fong, C. F., Turcotte, G., & De Kee, D. (1996). Modelling
steady and transient rheological properties. Journal of Food
Engineering, 27, 6370.
Cheng, D. C. H., & Evans, F. (1965). Phenomenological characteri-
sation of the rheological behaviour of inelastic reversible thixo-
tropic and antithixotropic uids. British Journal of Applied Physics,
16, 15991617.
De Kee, D., Code, R. K., & Turcotte, G. (1983). Flow properties of
time-dependent foodstus. Journal of Rheology, 27(6), 581604.
De Lorenzi, D., Pricl, S., & Torriano, S. (1995). Rheological behaviour
of low-fat and full-fat stirred yogurt. International Dairy Journal, 5,
661667.
Harr od, M. (1989). Modelling the ow properties of starch pastes
prepared by dierent procedures. Journal of Food Process Engi-
neering, 11, 257275.
International Dairy Federation Standard 151 (1991). (Provisional).
Determination of total solids content.
International Dairy Federation Standard 20B (1993). (Provisional).
Determination of nitrogen content.
Mead, R., Curnow, R. N., & Hasted, A. M. (1993). Statistical methods
in agriculture and experimental biology (pp. 227233). London:
Chapman & Hall.
Petrellis, N. C., & Flumerfelt, R. W. (1973). Rheological behaviour of
shear degradable oils: kinetic and equilibrium properties. Canadian
Journal of Chemical Engineering, 51, 291301.
Ramaswamy, H. S., & Basak, S. (1991a). Rheology of stirred yogurts.
Journal of Texture Studies, 22, 231241.
Ramaswamy, H. S., & Basak, S. (1991b). Time dependent stress decay
rheology of stirred yogurt. International Dairy Journal, 1, 1731.
Rohm, H. (1992). Viscosity determination of stirred yogurt. Food
Science and Technology, 25(3), 297301.
Skriver, A., Roemer, H., & Qvist, K. B. (1993). Rheological charac-
terisation of stirred yogurt: viscometry. Journal of Texture Studies,
24, 185189.
Steventon, A. J., Parkinson, C. J., Fryer, P. J., & Bottomley, R. C.
(1990). The rheology of yogurt. In R. E. Carter (Ed.), Rheology of
food, pharmaceutical and biological materials with general rheology
(pp. 196210). Amsterdam: Elsevier.
Tamime, A. Y., & Robinson, R. K. (1985). Yogurt science and
technology. Oxford: Pergamon Press.
Tiu, C., & Boger, D. V. (1974). Complete rheological characterisation
of time dependent food products. Journal of Texture Studies, 5,
329338.
Weltmann, R. N. (1943). Breakdown of thixotropic structure as
function of time. Journal of Applied Physics, 14, 343350.
254 H.J. O'Donnell, F. Butler / Journal of Food Engineering 51 (2002) 249254