See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/252256912

A Viscosity Function for Viscoplastic Liquids

Article · January 2004

CITATIONS READS
4 428

2 authors:

Paulo R. De Souza Mendes Eduardo S. S. Dutra

Pontifícia Universidade Católica do Rio de Janeiro Petróleo Brasileiro S.A.
118 PUBLICATIONS   1,218 CITATIONS    10 PUBLICATIONS   100 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Rheological behavior of time-dependent materials View project

Taylo-Couette Instabilities inpartially-obstructed annular space with inner cylinder rotation View project

All content following this page was uploaded by Paulo R. De Souza Mendes on 18 March 2015.

The user has requested enhancement of the downloaded file.

Applied Rheology Vol.14/6.qxd 27.12.2004 8:33 Uhr Seite 296

Viscosity Function for Yield-Stress Liquids

Paulo R. Souza Mendes* and Eduardo S. S. Dutra

Department of Mechanical Engineering

Pontifícia Universidade Católica - RJ Rua Marquês de São Vicente 225
Rio de Janeiro, RJ 22453-900, Brazil

*Email: pmendes@mec.puc-rio.br
Fax: x55.11.3114.1165
Received: 15.5.2004, Final version: 20.9.2004

Abstract:
A viscosity function for highly-shear-thinning or yield-stress liquids such as pastes and slurries is proposed. This
function is continuous and presents a low shear-rate viscosity plateau, followed by a sharp viscosity drop at a
threshold shear stress value (yield stress), and a subsequent power-law region. The equation was fitted to data
for Carbopol aqueous solutions at two different concentrations, a drilling fluid, an water/oil emulsion, a com-
mercial mayonnaise, and a paper coating formulation. The quality of the fittings was generally good.

Zusammenfassung:
Eine Viskositätsfunktion für stark scherverdünnende Flüssigkeiten oder Flüssigkeiten mit Fliessgrenze wie Pas-
ten und Schlämme wird vorgeschlagen. Diese Funktion ist stetig und besitzt bei kleinen Scherraten ein Viskosität-
splateau, an das sich ein abrupter Viskositätsabfall bei einem Grenzwert der Schubspannung (Fliessgrenze)
anschliesst, und von einem nachfolgenden Potenzgesetzbereich abgeschlossen wird. Die Gleichung wurde an
Daten von wässrigen Carbopollösungen, Bohrlochspühlung, Wasser/Öl Emulsion, kommerziell vertriebener
Mayonnaise und einer Formulierung für Papierbeschichtung gefittet. Die Qualität der Fits war allgemein gut.

Résumé:
Une fonction pour décrire la viscosité est proposée pour des fluides fortement rhéo-fluidifiant ou à seuil de con-
trainte d’écoulement comme des boues et des pâtes. Cette fonction est continue et caractérisée par un plateau
aux taux de cisaillements faibles, suivi d’une abrupte diminution de la viscosité au seuil de contrainte d’é-
coulement et un régime final en loi de puissance. La fonction est appliquée aux solutions aqueuses de Carbopol
à deux concentrations, à un fluide de forage, une émulsion eau/huile, une mayonnaise commerciale et à une
formulation pour le revêtement du papier. La qualité de la description est généralement très bonne.

Key words: Viscosity function, viscoplastic liquid, yield stress measurement

1 INTRODUCTION
Viscoplastic or yield-stress liquids are materials
that have an yield stress below which they
behave as a high viscosity liquid, and above
which they behave as a shear-thinning liquid. At
the yield stress, an often dramatic drop of the vis-
cosity level is observed.
Most of the viscoplastic materials that
appear in processes of interest are viscoplastic
liquids. A few examples of viscoplastic liquids
are: cement slurries, drilling muds and heavy oils
in the petroleum industry; mayonnaise, butter,
creams, pastes and many dairy products in the
food and cosmetics industries; clay, mud and
other concentrated suspensions in nature.
With the technological advancement in
rheometry, high (but finite) Newtonian viscosity

© Appl. Rheol. 14 (2004) 296-302

Applied Rheology
296 Volume 14 · Issue 6
Applied Rheology Vol.14/6.qxd 27.12.2004 8:33 Uhr Seite 297

Figure 1: The qualitative

behavior of the Carreau and
Cross functions for
viscoplastic liquids.

plateaus at low shear rates have been observed in

most viscoplastic materials of interest. A compre-
hensive discussion on this subject is found in [2].

2 VISCOSITY FUNCTIONS FOR

VISCOPLASTIC LIQUIDS

2.1 OVERVIEW OF THE AVAILABLE VISCOSITY η0 − η∞

η = η∞ +
1 + ( λγ )
FUNCTIONS . n−1
(3)
Perhaps the viscosity function that is the most
often employed to fit viscosity data of viscoplas- and the Carreau model (e.g. [5])
tic materials is the Herschel-Bulkley viscosity η0 − η∞
function [5]. The shear stress corresponding to η = η∞ + 1−n
this function is given by ⎡ 1 + ( λγ. )2 ⎤ 2
⎣ ⎦ (4)

.n
⎪⎧τ = τ o + K γ if τ > o
⎨.
⎪⎩γ = 0 otherwise
(1)
where t is the shear stress, t0 is the yield stress,
g· is the shear rate, K is the consistency index, and
n is the behavior (or power-law) index. When n =
1, it reduces to the traditional Bingham plastic
function
.
⎧⎪τ = τ o + µpγ if τ > τ o
⎨.
⎪⎩γ = 0 otherwise
(2)
where mp is the plastic viscosity. Both equations
predict an infinite viscosity in the limit of zero shear
rate. This behavior is not compatible with the con-
servation equations that govern many complex
flows [6]. Moreover, the prediction of an infinite
vicosity yields rather poor curve fittings to data per-
taining to viscoplastic liquids. One alternative
when the low shear rate range is not of interest is
to discard the data in the plateau region, but then
the dilemma of which data points to exclude comes
into play. The quality of the fitting and, more impor-
tantly, the value obtained for t0 are typically
strongly affected by this subjective decision. It is
rather frustrating to observe such an arbitrariness
in determining a parameter such as t0, which has
a clear physical meaning. Possible choices when no
shear-thinning is observed beyond the yield stress
are the Cross model (e.g. [3])
where h ∫ t/g· is the viscosity function, and l is a
curve-fitting parameter with dimension of time.
The two viscosity functions above are common-
ly used for pseudoplastic liquids with a zero-
shear-rate plateau, h0, a power-law region that
begins at g· > 1⁄l and an infinite-shear-rate
plateau, h0. In the limit of a very steep power-law
region (n ô 0), these two equations yield the
behavior illustrated in Fig. 1. Beyond the yield
stress there is just the infinite-shear-rate
plateau, because the capability of mimicking a
power-law behavior is used to obtain the viscos-
ity drop at the yield stress.
A regularized Bingham function was
proposed by Papanastasiou [7] for usage in finite-
element flow simulations, namely,
( .
)
τ = 1 − exp( −aγ ) τ 0 + µ pγ
.
(5)
where a is the regularizing parameter. As a ô
•, Papanastasiou's function approaches the
Bingham function, with the important advan-
tage of holding uniformly in yielded and unyield-
ed regions. The usual extension of Papanasta-
siou's idea for shear-thinning viscoplastic liquids,
that is, viscoplastic liquids that shear-thin at
shear stresses larger than the yield stress, is the
following modified Papanastasiou function:
( .
)
τ = 1 − exp( −aγ ) τ 0 + K γ n
.
(6)

Applied Rheology
Volume 14 · Issue 6 297
Applied Rheology Vol.14/6.qxd
Figure 2 (left):
The qualitative behavior of
the modified Papanastasiou
viscosity function and of the
modified bi-viscosity
function.
Figure 3 (right): Shear stress
as a function of shear rate
as predicted by Eq. 8.
298
27.12.2004 8:33 Uhr Seite 298
As a ô •, the modified Papanastasiou’s
function approaches the Herschel-Bulkley func-
tion. The shear stress as given by Eq. 6 results in
the viscosity function plotted in Fig. 2 as a
function of the shear stress. This figure illustrates
that the modified Papanastasiou function is
unable to predict a finite viscosity plateau in the
limit of zero shear rate for shear-thinning
viscoplastic liquids.
An interesting alternative that has a
more suitable qualitative behavior for viscoplas-
tic liquids is the so-called modified bi-viscosity
function [4] [10], which equation for shear stress
is given by
. . .
⎧⎪τ + K γ n if γ 0 γ
τ =⎨ 0 .
⎪⎩η0γ otherwise
(7)
where g· 0 = t0/(h0 - Kg0n-1) > t0/h0 is
the yield shear rate, usually very low. Figure 2
shows the plot of the corresponding viscosity
function as a function of the shear stress. This vis-
cosity function involves two different expres-
sions, each one applicable to a different shear-
rate range. These ranges are delimited by the
yield shear rate, which is to be determined in the
curve-fitting procedure itself. These characteris-
tics, together with its discontinuous derivative
(Fig. 2), cause severe practical problems that typ-
ically prevent good-quality fittings of Eq. 7 to vis-
cosity data of viscoplastic liquids.
2.2 THE PROPOSED VISCOSITY FUNCTION
We now propose and briefly discuss the proper-
ties of the following function:
( .
)(
τ = 1 − exp( −η0γ / τ 0 ) τ 0 + K γ n
.
) (8)
Applied Rheology
Volume 14 · Issue 6
This function is plotted in Fig. 3 for illustration
purposes. The physical meaning of the parame-
ters that appear in Eq. 8, namely, h0, t0, K, n are
also illustrated in this figure.
The zero-shear rate viscosity is just equal
to the ratio t/g· provided t is smaller enough than
t0 to ensure that g· is within the zero-shear rate
plateau region. The yield stress becomes evident
in the t â g· plot, because of the plateau that
occurs at t0. The behavior index n is the slope of
the power-law region in the log-log plot of t â g· .
The intercept of the extrapolated power-law-region
straight line with the vertical line at g· =1 s-1 occurs
at t=K. This information can be used to generate
good initial guesses for the curve-fitting para-
meter values in least-squares fitting procedures.
In Fig. 4 we can observe the qualitative
behavior of the viscosity function corresponding
to Eq. 8. The zero-shear rate plateau is followed by
a sharp drop at t = t0, and then it follows a power-
law region. This behavior is quite similar to the one
presented by the bi-viscosity function (Fig. 2),
except that there is no discontinuity in the curve
derivative at t = t0. A good feature of Eq. 8 is that
it predicts a viscosity function such that
η ( 0) = lim
( 1 − exp( −η γ / τ ))
.
0 0
(τ .
+ Kγ n )
. .
γ →0 γ 0
(η / τ 0 )exp( −η0γ / τ 0 )
.
= lim
.
γ →0
0
1
(τ 0
.
)
+ K γ n = η0
(9)
in contrast to the modified Papanastasiou func-
tion, which predicts an unbounded viscosity
function in the limit of zero shear rate (see
Fig. 2).
Often the data do not present a viscosi-
ty drop as sharp as predicted by Eq. 8. For these
cases, the exact yield stress value is not so evi-
dent, and one possible procedure to decrease
Applied Rheology Vol.14/6.qxd 27.12.2004 8:34 Uhr Seite 299

Figure 4 (left above):

Viscosity as a function of
shear stress for the function
given by Eq. 8.

Figure 5 (right above):

Derivative of the logarithm
of the shear stress with
respect to the logarithm of
the shear rate as a function
of shear stress, for the
function given by Eq. 8.

Figure 6 (left below): Flow

curve of a Carbopol 0.12%
wt aqueous solution.

Figure 7 (right below):

Derivative of the logarithm
of the shear stress with
respect to the logarithm of
the shear rate as a function
of shear stress, for the data
presented in Figure 6.

subjectivity is to define t0 as the shear stress cor- that the derivative given by Eq. 10 when t is given
responding to the minimum derivative of the by Eq. 8 tends to 1 at low shear rates (or shear
logarithm of the shear stress with respect to the stresses), is minimum at the yield stress, and
logarithm of the shear rate, as illustrated in tends to the behavior index n in the power-law
Fig. 5. This derivative for a given set of data can region, as illustrated in Fig. 5.
be easily evaluated by central differences and
then plotted as a function of t:
3 FITTINGS TO SOME REPRESENTATIVE
. .
DATA
d lnτ γ i +1 2 dτ γ i +1 2 ⎡ τ (γ. i +1 ) − τ (γ. i ) ⎤
. = . ⎢ . . ⎥
d lnγ τ i +1 2 dγ τ i +1 2 ⎢⎣ γ i +1 − γ i ⎥⎦ We now illustrate with the aid of Figs. 6, 8, 10, 12,
i +1 2 i +1 2
(10) 14, and 16 the fitting capability of Eq. 8 to data of
a few viscoplastic liquids that appear in industrial
where applications. It is worth emphasizing that the goal
here is not to discuss measurement techniques or
1 . .
γ i + 1 2 = (γ i + γ i + 1 )
.
the rheological behavior of these materials, but
2
rather to demonstrate the general fitting capabil-
1
τ i +1 2 = ⎡⎣τ (γ i ) + τ+ (γ i )⎤⎦
. . ity of Eq. 8 to viscoplastic liquid data. All data pre-
1
2 (11) sented in this paper were obtained with the aid of
a commercial rotational rheometer (ARES, Rheo-
and t(g· i) is the measured shear stress at g· = g· i, metric Scientific). The tests were conducted in
while g· i = 1, 2, ..., m is the monotonically increas- strain-rate-controlled mode. The geometry
ing series of m shear rate values at which mea- employed was a home-made roughened-surface
surements were made. It is not difficult to see Couette geometry to avoid slip effects [1-3, 8, 9].

Applied Rheology
Volume 14 · Issue 6 299
Applied Rheology Vol.14/6.qxd 27.12.2004 8:34 Uhr Seite 300

Figure 8 (left above):

Flow curve of a Carbopol
0.20% wt aqueous solution.

Figure 9 (right above):

Derivative of the logarithm
of the shear stress with
respect to the logarithm of
the shear rate as a function
of shear stress, for the data
presented in Figure 8.

Figure 10 (left middle):

Flow curve of a drilling mud.

Figure 11 (right middle):

Derivative of the logarithm
of the shear stress with
respect to the logarithm of
the shear rate as a function
of shear stress, for the data
presented in Figure 10.

Figure 12 (left below):

Flow curve of an oil/water
emulsion with surfactant.

Figure 13 (right below):

Derivative of the logarithm
of the shear stress with
respect to the logarithm of
the shear rate as a function
of shear stress, for the data
presented in Figure 12.

In these figures we can observe that the
fittings are generally of good quality, because the
qualitative behavior of Eq. 8 throughout the
whole range of shear rate is essentially the same
as the one of the data. Therefore, there is no need
to discard data pertaining to the low-shear-rate
range, which is undesirable because often the
parameter values obtained in the curve-fitting
procedure depend strongly on the choice of the
data to be discarded.
The procedure for evaluating the yield
stress t0 described in Sec. 2.2 was applied to the
data shown in Figs. 6, 8, 10, 12, 14, and 16. The cor-
responding plots of the derivative of the loga-

Applied Rheology
300 Volume 14 · Issue 6
Applied Rheology Vol.14/6.qxd 27.12.2004 8:34 Uhr Seite 301

Figure 14 (left above):

Flow curve of mayonnaise.

Figure 15 (right above):

Derivative of the logarithm
of the shear stress with
respect to the logarithm of
the shear rate as a function
of shear stress, for the data
presented in Figure 14.

Figure 16 (left below):

Flow curve of a paper
coating formulation.

Figure 17 (right below):

Derivative of the logarithm
of the shear stress with
respect to the logarithm of
the shear rate as a function
of shear stress, for the data
presented in Figure 16.

rithm of the shear stress with respect to the log-
arithm of the shear rate as a function of shear
stress are shown in Figs. 7, 9, 11, 13 , 15, and 17,
respectively. These plots illustrate, for different
materials, the success of the method for obtain-
ing an unambiguous value for the yield stress. In
some cases the t â g· plot presents some waviness,
which leads to more than one local minimum of
the derivative dlnt/dlng· , as seen in Figs. 9, 11, and
13. In these cases the lowest stress at which a min-
imum occurs should be taken as the yield stress.
It is interesting that this first minimum was the
lowest for all viscoplastic materials examined in
this research. Lastly, it is worth commenting the
remarkable agreement between the t0 values
obtained in the two different ways, namely, via
curve fitting and via the derivative method.
tions and curve-fitting procedures. Its qualitative
behavior is the same as the one observed for
most viscoplastic liquids of interest, i.e. a high-
viscosity plateau at low shear rates, followed by
a sharp drop of the viscosity level, and then a
power-law region. A simple method to deter-
mine the yield stress is also proposed, which is
independent of the viscosity model chosen.
ACKNOWLEDGMENTS
This research was possible due to the financial
support of Petrobras, the Brazilian Petroleum
Company, FINEP (Financiadora de Estudos e Pro-
jetos), and CNPq (Conselho Nacional de Desen-
volvimento Científico e Tecnológico).

CONCLUSION REFERENCES

A viscosity function for yield-stress liquids is pro- [1] Alderman NJ, Heywood NI: Improving slurry vis-
posed. It is continuous and has continuous deriv- cosity and flow curve measurements, AICHE
atives, as it is convenient for numerical simula- Chem. Eng. Prog. 101 (2004) 27–32.

Applied Rheology
Volume 14 · Issue 6 301
 Applied Rheology Vol.14/6.qxd 27.12.2004 8:34 Uhr Seite 302

[2] Barnes BA: The yield stress – a review, J. Non-
Newtonian Fluid Mech. 81 (1999) 133–178.
[3] Barnes BA: A handbook of elementary rheology.
University of Wales, 2000.
[4] Beverly CR, Tanner RI: Numerical analysis of
three-dimensional Bingham plastic flow, J. Non-
Newt. Fluid Mech., 42:85–115, 1992.
[5] Bird RB, Armstrong RC, Hassager O: Dynamics of
Polymeric Liquids, John Wiley & Sons, New York,
1987.
[6] Lipscomb GG, Denn MM: Flow of Bingham fluids
in complex geometries, J. Non-Newtonian Fluid
Mech. 14 (1984) 337–346.
[7] Papanastasiou TC: Flow of materials with yield,
J. Rheology 31 (1987) 385–404.
[8] C. Picart. Réomètrie de cisaillement et
microstructure du sang, Ph.D. thesis, Université
Joseph Fourier, Grenoble, France, 1997.
[9] Picart C, Piau JM, Galliard H: Human blood shear
yield stress and its hematocrit dependence, J.
Rheology 42 (1998) 1–12.
[10] Soares M, Naccache MF, Souza Mendes PR: Heat
transfer to viscoplastic materials flowing lami-
narly in the entrance region of tubes, Interna-
tional Journal of Heat and Fluid Flow 20 (1999)
60–67.

Applied Rheology
302 Volume 14 · Issue 6

View publication stats