thixotropic elasto-viscoplastic model for structured fluids
TRANSCRIPT
Dynamic Article LinksC<Soft Matter
Cite this: Soft Matter, 2011, 7, 2471
www.rsc.org/softmatter PAPER
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online / Journal Homepage / Table of Contents for this issue
Thixotropic elasto-viscoplastic model for structured fluids
Paulo R. de Souza Mendes*
Received 19th September 2010, Accepted 20th December 2010
DOI: 10.1039/c0sm01021a
A constitutive model for structured fluids is presented. Its predictive capability includes thixotropy,
viscoelasticity and yielding behavior. It is composed by two differential equations, one for the stress and
the other for the structure parameter—a scalar quantity that represents the structuring level of the fluid.
The equation for stress is obtained in accordance with a simple mechanical analog composed by
a structuring-level-dependent Maxwell element in parallel with a Newtonian element, leading to an
equation of the same form as the Jeffreys (or Oldroyd-B) equation. The relaxation and retardation
times that arise are functions of the structure parameter. The ideas found in de Souza Mendes, J. Non-
Newtonian Fluid Mech., 2009, 164, 66 are employed for the structure parameter equation as well as for
the dependencies on the structure parameter of the structural viscosity and structural shear modulus.
The model is employed in constant-rate, constant-stress, and oscillatory shear flows, and its predictive
capability is shown to be excellent for all cases.
1. Introduction
Structured fluids are present in everyday life. Most suspensions,
emulsions, and foams are structured fluids. Many foods,
personal care products, paints, inks, cements, adhesives, greases,
natural muds, drilling muds, crude oils, gels, and various slurries
are some examples of structured fluids.
The mechanical behavior of most structured fluids is highly
non-Newtonian. At small stress levels, their behavior is visco-
elastic. Beyond a certain stress threshold, usually called yield
stress, a major microstructure collapse occurs which causes
dramatic drops in viscosity and elasticity.
The microstructure of a structured fluid usually acquires
a stable configuration if exposed for a long time to a constant
stress or shear rate. This steady state is the result of the equi-
librium between the microstructure buildup and breakdown
rates. If a new equilibrium is not achieved instantaneously after
a step change to a new stress or rate, then the structured fluid is
said to be time-dependent. Note that yield-stress as well as other
structured fluids can in principle be either time-dependent or
time-independent, but in most cases this classification itself is
a function of the combination and range of parameters under
study.
A time-dependent fluid is said to be thixotropic if its steady-
state viscosity decreases with the shear rate (i.e. if it is shear-
thinning), and if in addition the viscosity changes are reversible.
On the other hand, a time-dependent fluid is said to be antith-
ixotropic if its steady-state viscosity increases with the shear rate
Department of Mechanical Engineering, Pontif�ıcia UniversidadeCat�olica-RJ, Rua Marques de Sao Vicente 225, Rio de Janeiro, RJ22453-900, Brazil. E-mail: [email protected]
This journal is ª The Royal Society of Chemistry 2011
(i.e. if it is shear-thickening), and if in addition the viscosity
changes are reversible.
Time dependency is observed in most structured fluids, espe-
cially in the small stress range.1 Irreversibility of the micro-
structure changes is also common, which render rather difficult
the tasks of modeling and characterizing such fluids.2,3
A number of useful reviews on thixotropy are available in the
literature, among which the ones published by Barnes4 and, more
recently, Mewis and Wagner.3 Mujumdar et al.5 also gives
a thorough discussion of the thixotropy literature, including
a quite complete comparison between the various structural
kinetics models then found in the literature, and de Souza
Mendes6 discusses the main aspects of these literature reviews.
The more recent thixotropy models found in the literature
include elasticity effects.7,8,9,10 Some models consider elasticity
and yielding, but no thixotropy effects e.g. ref. 11,12. One
shortcoming of these thixotropy models that include elasticity is
their incapability of predicting what experimentally distinguishes
thixotropy from viscoelasticity, namely an instantaneous drop in
shear stress when the shear rate is suddenly decreased.3
In summary, so far most of the available thixotropy models
rely heavily on ad hoc assumptions that most often lack justifi-
cations other than simplicity’s sake. As a result, their predictive
capability is often inadequate and, in addition, restricted to
a narrow range of applications. It is not unusual that the
obtained predictions are qualitatively wrong even for the steady-
state shear viscosity.
More recently, de Souza Mendes6 proposed a model for elasto-
viscoplastic thixotropic fluids that introduced a number of novel
ideas. The stress equation was developed strictly in accordance
with a Maxwell-like mechanical analog and well founded phys-
ical arguments. The proposed evolution equation for the
Soft Matter, 2011, 7, 2471–2483 | 2471
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
structure parameter was also developed in a conceptually
different manner, viz., the instantaneous stress and the steady-
state viscosity data were included in the breakdown term
according to a clear physical basis. The resulting formulation is
more straightforward than the previous ones found in the liter-
ature, and the drawbacks presented by the previously published
viscoelasticity-based thixotropy models were removed alto-
gether. Consequently, the predictions of the model proposed by
de Souza Mendes6 were shown to be faithful to experimental
observation for a much wider range of rheological flows.
The present paper follows the lines of ref. 6 to propose a new
model for structured fluids whose differential equation for stress
is developed consistently with a mechanical analog composed by
a strucuring-level-dependent Maxwell-like element in parallel
with a Newtonian element. The presence of the Newtonian
element is shown to enhance significantly the quality of the
predictions. Results obtained with the model for different rheo-
logical test flows demonstrate a far superior predictive capability.
2. The model
This section describes the assumptions and gives the equations
that compose the proposed constitutive model for thixotropic
materials. One of the key assumptions is the existence of
a microstructure whose state can be described by a single scalar
parameter. Let l be this parameter that expresses the state of the
structure. By definition, it ranges from 0 to 1, 0 corresponding to
a completely unstructured state and 1 corresponding to
a completely structured state.
2.1. Equation for stress
The differential equation for the shear stress s used in the present
model is now derived with basis on the mechanical analog shown
in Fig. 1. In this figure, Gs(l) is the shear modulus of the
microstructure; hs(l) is the structural viscosity, a function that
describes the purely viscous response of the microstructure; hN is
the viscosity corresponding to the completely unstructured state
(i.e. to l ¼ 0); ge is the elastic shear strain of the microstructure
when it is submitted to the shear stress s; gv is the viscous shear
strain; and g is the total shear strain. This analog corresponds to
the Jeffreys (or Oldroyd-B) viscoelastic constitutive model,
except that here both Gs and hs are assumed to be functions of the
structure parameter l. It is clear that, according to this analog,
a very large value of hs combined with a finite value of Gs implies
Fig. 1 The mechanical analog.
2472 | Soft Matter, 2011, 7, 2471–2483
the behavior of a viscoelastic solid (with a retardation time equal
to hN/Gs); conversely, a very large value of Gs combined with
a finite value of hs implies the behavior of a purely viscous fluid
(whose viscosity is hv ^ hs + hN).
It is easy to see from the model represented in Fig. 1 that
ge + gv ¼ g 0 _ge + _gv ¼ _g (1)
s ¼ s1 þ s20_s ¼ _s1 þ _s2 (2)
where the dot on top of the variables denotes differentiation with
respect to time t. s1 is the structural stress and s2 is the structure-
independent stress. It is clear that
s2 ¼ hN _g (3)
Thus,
s1 ¼ s� hN _g0_s1 ¼ _s� hN€g (4)
Moreover, the following two expressions also hold for s1:
s1 ¼ hs _gv (5)
s1 ¼ Gs(ge � ge,n) (6)
where ge and ge,n are deformations measured from an arbitrary
fixed reference configuration of the microstructure. ge is the
elastic deformation corresponding to the current configuration,
and ge,n is the deformation corresponding to the natural or
neutral configuration, i.e. the configuration assumed when s ¼ 0.
The neutral configuration is a characteristic of the microstruc-
ture, and hence it is expected to change if (and only if) the
microstructure changes. Consequently, ge,n is expected to be
a sole function of the structure parameter l. Differentiation of
eqn (6) with respect to time gives
_s1 ¼ _Gs
�ge � ge;n
�þ Gs
�_ge � _ge;n
�(7)
The hypothesis that changes in ge,n are solely due to changes in
the microstructure (i.e. k_¼ 0 0 _Gs ¼ 0, _ge,n ¼ 0) leads to the
conclusion that changes in ge are solely due to changes in s1, i.e.
_ge ¼ 05_s1 ¼ 0.
In other words, eqn (7) can be decomposed into two inde-
pendent expressions:
0 ¼ _Gs
�ge � ge;n
�� Gs _ge;n or _ge;n ¼
_Gs
Gs
�ge � ge;n
�(8)
and
_s1 ¼ Gs _ge (9)
The above considerations grant the desired behavior of the
model. For example, it is worth examining the case in which the
material microstructure is initially completely destroyed due to
a long exposure to high shear stresses, and then at a given instant
of time the stress is set to zero so that the microstructure starts
building up. In this case, the model predicts that ge,n ¼ ge (see
eqn (6)), _ge ¼ 0 (see eqn (9)), and thus _ge,n ¼ 0, despite the fact
that _Gs s 0 (see eqn (8)). In other words, in this case the model
predicts that the building up of the microstructure will render the
material elastic while keeping it in its relaxed configuration ge,n,
as it should.
This journal is ª The Royal Society of Chemistry 2011
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
Now eqn (1) is multiplied by hs and then eqn (5) and (9) are
plugged into it. The result is
s1 þhs
Gs
_s1 ¼ hs _g or s1 ¼ �q1 _s1 þ hs _g (10)
where q1 ¼hs
Gs.
Now eqn (2), (4), and (10) are combined, yielding
s ¼ s1 þ s2 ¼ �q1 _s1 þ ðhs þ hNÞ _g¼ �q1
�_s� hN€g
�þ ðhs þ hNÞ _g (11)
or
sþ q1 _s ¼ ðhs þ hNÞ _gþ q1hN€g (12)
Rearranging, a Jeffreys-like equation for stress is finally
obtained:
sþ q1 _s ¼ hv
�_gþ q2€g
�(13)
In this equation,
hv ¼ hs þ hN and q2 ¼hshN
Gsðhs þ hNÞ¼ hshN
Gshv
¼ hN
hv
q1 (14)
Thus, q2 < q1.
Because the characteristic times q1 and q2 vary continuously
with the structuring level, eqn (13) actually represents a family of
Jeffreys materials parametrized by the structuring level. This
feature lends a remarkable predictive capability to eqn (13), as it
will be illustrated in Sec. 3 below.
2.2. Relaxation and retardation times
It is clear that q1(l) is a relaxation time, while q2(l) is a retarda-
tion time. These quantities can be written in the following form:
q1 ¼�
1� hN
hvðlÞ
�hvðlÞGsðlÞ
(15)
q2 ¼�
1� hN
hvðlÞ
�hN
GsðlÞ(16)
When the material is highly structured (l / 1, and hence
hv [ hN), then q1zhvðlÞGsðlÞ
, and q2zhN
GsðlÞ. In this case s/hv is
very small, and eqn (13) approaches the one pertaining to the
Kelvin–Voigt viscoelastic solid model, namely
_gþ hN
Gs
€g z_s
Gs
(17)
Thus, when the material is highly structured, the behavior as
predicted by eqn (13) is essentially independent of the structural
viscosity hs.
On the other extreme, when the material becomes unstructured
(l / 0) both q1 and q2 go to zero, and eqn (13) reduces to the
Newtonian fluid model (s ¼ hN _g).
2.3. Viscosity function h
The viscosity function is defined as
This journal is ª The Royal Society of Chemistry 2011
hhs_g
(18)
It can be written as a product of two functions, one carrying
the elasticity information and the other the purely viscous
character. To this end, eqn (1) is rewritten as:
1� _ge
_g¼ _gv
_g¼ hv _gv
hv _g¼ hs _gv þ hN _gv
hv _g
¼ hs _gv þ hN _g� hN _ge
hv _g
z}|{¼ s
¼ s_ghv
� hN
hv
z}|{¼q2=q1
_ge
_g¼ h
hv
� q2
q1
_ge
_g(19)
Thus,
h ¼�
1��
1� q2
q1
�_ge
_g
�hv (20)
where the term between brackets is the elastic contribution to the
viscosity function h.
2.4. Generalization of the stress equation
A tridimensional, frame-indifferent version of eqn (13) for
general flows is readily obtained by replacing (i) the shear stress sby the extra-stress tensor s ^ T + p1 (T is the total stress tensor
field, p is the pressure field, and 1 is the unit tensor), (ii) the shear
rate _g by the rate-of-deformation tensor _g ^ Vv + VvT (v is the
velocity vector field), and (iii) the time derivative by the upper-
convected time derivative (for example):
sþ q1ðlÞsV ¼ hvðlÞ
�_gþ q2ðlÞ _g
V �(21)
This paper restricts the discussion to homogeneous shear
flows, for which eqn (21) yields13 s22 ¼ s23 ¼ s32 ¼ s33 ¼ 0, and�sþ q1ðlÞ _s ¼ hvðlÞð _gþ q2ðlÞ€gÞ ð21� componentÞ
s11 þ q1ðlÞ _s11 ¼ 2sq1ðlÞ _g� 2hvðlÞq2ðlÞ _g2 ð11� componentÞ(22)
where 1 is the flow direction, 2 is the gradient direction, and 3 is
the neutral direction. Eqn (22) show that the 21–component of
eqn (21) coincides with eqn (13) for homogeneous shear flows.
Moreover, a non-zero first normal stress difference s11 � s22 is
predicted by eqn (21). This predicted normal stress difference
depends of course on the structure parameter l, and it tends to
zero as l / 0, as expected for an unstructured fluid.
Eqn (22) also show that the shear stress s as predicted by the
present model is independent of the normal stresses for homo-
geneous shear flows.
2.5. The structural shear modulus function Gs(l)
A physically reasonable function Gs(l) should ensure that both q1
and q2 are monotonically decreasing functions of l (see eqn (15)
and (16)). This criterion is satisfied, for example, if (but not only
if) (i) Gs(l)¼ constant, and (ii) Gs(l) is smallest when the material
Soft Matter, 2011, 7, 2471–2483 | 2473
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
is fully structured (l ¼ 1), and increases monotonically as l is
decreased.
One possible choice for Gs(l) that is consistent with the above
discussion is6
Gs ¼Go
lm (23)
where Go is the structural shear modulus of the completely
structured material and m is a dimensionless positive constant.
2.6. The structural viscosity function hs(l)
Eqn (20) shows that, when _ge ¼ 0, the viscosity function reduces
to h ¼ hv ¼ hs + hN. In particular, for steady-state flows,
hss( _g) ¼ hv(lss( _g)) (24)
where hss is the steady-state viscosity function and lss the steady-
state structure parameter. Therefore, the steady-state viscosity
function carries useful information regarding the dependence of
h and hv on the structure parameter l.6 For example, it is clear
that both hss and hv vary from hN to ho, where ho is the steady-
state viscosity of the completely structured material (l ¼ 1).
Thus, the function hv(l) should map the range [0,1] into the range
[hN, ho].
The above considerations suggest the following form for
hv(l):6
hvðlÞ ¼�
ho
hN
�l
hN (25)
Eqn (25) can be seen as the definition of the structure
parameter. It can be solved for l to yield
l�
_g; t�¼�
ln hvð _g; tÞ � ln hN
ln ho � ln hN
�(26)
In particular, the above equation reduces to the following
expression for the steady-state structure parameter lss:
lss
�_g�¼�
ln hssð _gÞ � ln hN
ln ho � ln hN
�(27)
Therefore, once the flow curve is determined experimentally,
the steady-state structure parameter is also determined from eqn
(27)6.
2.7. The evolution equation for the structure parameter
The structure parameter l is now assumed to obey the following
evolution equation:6
dl
dt¼ 1
teq
�ð1� lÞa�f ðsÞlb
�(28)
where teq is a characteristic time of change (of l), and a and b are
positive dimensionless constants. In this equation, the first term
on the right-hand side is a structure buildup term, while the
second one is a breakdown term.
As discussed in ref. 6, in the literature the function f appearing
in the breakdown term of eqn (28) is always taken to depend on
the shear rate _g, such that it is zero at _g ¼ 0 and increases
monotonically as _g is increased e.g. see ref. 4,5,14. However, it
2474 | Soft Matter, 2011, 7, 2471–2483
seems more adequate that the breakdown term be a function of
the shear stress s instead, such that it is zero at s¼ 0 and increases
monotonically as s is increased. This is so because what breaks
the microstructure is the level of stress, and not the level of
deformation rate. For example, if a completely structured
material, initially at rest, is suddenly submitted to a constant
shear rate, the shear stress, initially null, will increase linearly
with time. It is reasonable to expect that (i) the breakdown rate is
small at early times, when the microstructure is nearly unde-
formed and under small stresses, and (ii) as time elapses and the
stress builds up, the breakdown rate increases. However, note
that if a dependence of f on _g is assumed, a non-physical response
is obtained. In this situation, f does not change with time. As
a consequence, the breakdown rate, lbf, is maximum at the onset
of the flow (when the stress is zero), and decreases as the
microstructure breaks and l decreases.
In steady state (dl/dt ¼ 0), the function f(s) should reduce to
f ðsssÞ ¼ f�hv
�lss
�_g��
_g�¼ ½1� lssð _gÞ�a
lssð _gÞbin steady state (29)
In view of the above discussion, the following form for f(s) has
the adequate properties:
f ðsÞ ¼ ½1� lssð _gÞ�a
lssð _gÞb�
shvðlÞ _g
�c
(30)
where c is a dimensionless positive constant. Note that, in steady
state, eqn (30) reduces to eqn (29) as it should. The terms
hvðlÞ _gis
of course equal to the term between brackets of eqn (20), and it
introduces elasticity and stress level effects in the breakdown
term.
Thus the model uses the steady-state experimental information
as an input for the evolution equation, because the function
lss( _g) used is determined from the observed material behavior
(see eqn (27)).
Combination of eqn (28) and (30) gives
dl
dt¼ 1
teq
�ð1� lÞa�ð1� lssÞa
�l
lss
�b� shvðlÞ _g
�c�(31)
In the related literature, the function f appearing in eqn (28) is
often assumed to be of the form k _gd, where k is a fitting constant
and d is in most cases taken to be equal to 1 for simplicity (e.g.
ref. 5,14,15. This assumption defines the form of the steady-state
structure parameter lss( _g) (see eqn (29)), which is not necessarily
compatible with the experimentally observed steady-state
viscosity function hss( _g) e.g. see ref. 5,15.
In this work a reverse approach is adopted,6 viz. a representative
form for hss( _g) is chosen, and then it is used in eqn (27) to obtain
lss. This approach yields a more elaborate function f, which is
consistent with the observed behavior of the material while
flowing in steady state. An immediate benefit of this procedure is
the certainty of excellent predictions of steady shear flow.
Eqn (31) can be easily adapted for usage in complex,
non-homogeneous flows, by interpreting (i)dl
dtas the material
derivative of l, i.e.dl
dt¼ vl
vtþ v,Vl, (ii) _g as the second invariant
This journal is ª The Royal Society of Chemistry 2011
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
of the rate of deformation tensor, i.e. _g ¼ffiffiffiffiffiffiffiffiffiffi1
2tr _g
r2, and (iii) s as
the second invariant of the extra-stress tensor, i.e. s ¼ffiffiffiffiffiffiffiffiffiffiffi1
2trs2
r.
2.8. The steady-state viscosity function
Many of the steady-state viscosity functions available in the
literature can be used in this model, provided it fits well to the
steady-state data. In this work the form proposed by de Souza
Mendes6 is chosen:
hss
�_g�¼�
1� exp
�� ho _g
so
���so � sod
_ge� _g= _god þ sod
_gþ K _gn�1
þ hN
(32)
In this equation, so is the static yield stress, sod the dynamic
yield stress, _god a shear rate that marks the transition in stress
from so to sod, K the consistency index, and n the power-law
index. Eqn (32) is capable of predicting all the features observed
in steady-state data for viscoplastic materials.
2.8.1. Features of the steady-state viscosity function. For
completeness, the main features of the steady-state viscosity
function given by eqn (32) are now discussed, with the aid of
Fig. 2. It is clear that, consistently with the assumptions and
definitions above, eqn (32) obeys the limiting values ho and hN,
as _g / 0 and _g / N respectively.
In Fig. 2 it can be seen that the stress changes from the
static yield stress so to the dynamic yield stress sod in the
vicinity of _g ¼ _god. As discussed in ref. 6, the reason for
including the static yield stress in the flow curve stems from
the fact that the static yield stress must eventually be attained
in steady state, provided the shear rate is low enough,
Fig. 2 The steady-state shear stress, viscosity, and structure parameter
as a function of the shear rate.
This journal is ª The Royal Society of Chemistry 2011
otherwise the structuring/unstructuring process would not be
reversible as required by the definition of a thixotropic
material. In the viscosity function given by eqn (32), this
sufficiently low shear rate is roughly _god.
This figure illustrates that, in addition to _god, three other shear
rates mark important transitions in the flow curve, namely _g0, _g1,
and _g2. These shear rates are given by:6
_go ¼so
ho
; _g1 ¼�sod
K
�1=n
; _g2 ¼�hN
K
�1=n�1
(33)
The steady-state viscosity of the completely structured mate-
rial (l ¼ 1), ho, corresponds to the Newtonian plateau observed
in Fig. 2 in the shear rate range _g < _go. Therefore, _go is roughly
the maximum shear rate at which the material structure is
unaffected.
Another transition shear rate is _g1, which marks the beginning
of the power-law region that follows the sharp viscosity decrease
observed at s ¼ sod. _g1 can be conveniently used as a character-
istic shear rate to non-dimensionalize viscoplastic fluid
mechanics problems.16
The highest transition shear rate, _g2, is the one at which the
transition from power-law behavior to Newtonian behavior
occurs, in the high shear rate region (see Fig. 2) where the
material structure is completely destroyed.
Fig. 2 illustrates that the flow curve predicted by eqn (32) is
non-monotonic within the stress range sod < s < so, as observed
in flow curves of many thixotropic materials. This feature is
known to be directly related to interesting phenomena such as
shear banding.17
2.9. Parameters of the model
In summary, eqn (13), (23), (25), (15), (16), (27), (31), and (32)
compose the thixotropy model proposed in this paper. These
equations are gathered below:
sþ q1 _s ¼ hv
�_gþ q2€g
�(13)
Gs ¼Go
lm (23)
hvðlÞ ¼�
ho
hN
�l
hN (25)
q1 ¼�
1� hN
hvðlÞ
�hvðlÞGsðlÞ
(15)
q2 ¼�
1� hN
hvðlÞ
�hN
GsðlÞ(16)
lss
�_g�¼�
lnhssð _gÞ � lnhN
lnho � lnhN
�(27)
Soft Matter, 2011, 7, 2471–2483 | 2475
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
dl 1�
a a
�l�b� s
�c�
dt¼teq
ð1� lÞ �ð1� lssÞlss hvðlÞ _g
(31)
hss
�_g�¼�
1� exp
�� ho _g
so
���so � sod
_ge� _g= _god þ sod
_gþ K _gn�1
þ hN
(32)
Thus, comparing the present model with the one proposed by
de Souza Mendes,6 it is seen that the parameters that appear in
the two models are exactly the same, i.e. no additional parameter
was needed for the inclusion of the retardation time in the present
model.
Specifically, the parameters are: ho, hN, so, sod, _god, K, n, Go,
m, teq, a, b, and c.
The flow curve parameters sod, K, and n can be determined
from a least-squares fitting to the flow curve, while the remaining
parameters can in principle be obtained via fittings to data per-
taining to transient flows. Some transient flows are discussed in
Sec. 3, where possible procedures for determination of parame-
ters are suggested. However, a detailed discussion on the meth-
odology to determine the model parameters is beyond the
intended scope of this text.
3. Results and discussion
To demonstrate the predictive capability of the model, this
section presents solutions of eqn (13), (23), (25), (15), (16), (27),
(31), and (32) for some selected flows. Simultaneous integration
of the differential equations involved was performed numerically
with a fourth-order Runge–Kutta discretization in time.
The results are presented in dimensionless form, following the
ideas described elsewhere.16 The characteristic shear rate
employed in the scaling is _g1 (defined in eqn (33), see also Fig. 2),
whereas the characteristic shear stress is sod.
All the results below pertain to the following set of parameter
values: ho _g1/sod ¼ 107; n ¼ 0.5; so/sod ¼ 2; _god/ _g1 ¼ 10�4; hN _g1/
sod ¼ 10�2; Go/sod ¼ 1; m ¼ 0.1; _g1teq ¼ 10; a ¼ 1; b ¼ 1; c ¼ 0.1.
3.1. Steady-state predictions
As discussed earlier, the steady-state viscosity function predicted
by the present model is, by construction, exactly the same as the
one used to compose the evolution equation for the structure
parameter, eqn (31) (see eqn (27) and (30)). Thus, in the present
case, the predicted steady-state viscosity function is given by eqn
(32) and plotted in Fig. 2.
Fig. 3 Stress evolution for breakdown or rejuvenation experiments: the
material is at rest for a long time and, at t ¼ 0, a constant shear rate is
imposed.
3.2. Constant shear rate flows
The predictions of the model for the flows characterized by a step
change in shear rate are now examined. The flow is initially
steady at a given shear rate value, _gi, and, at time t¼ 0, the shear
rate is abruptly changed to another value, _gf. That is,
_g(t) ¼ _gi + ( _gf � _gi)H(t) (34)
where H(t) is the Heaviside step function.
2476 | Soft Matter, 2011, 7, 2471–2483
For the simple case in which both Gs and hv are constant,
eqn (13) can be integrated to give the following solution for s:
sðtÞ ¼ _gf
�hv � ðhv � hNÞ
�1� _gi
_gf
�exp
�� Gs
hv � hN
t
�(35)
This equation is useful in the analysis of the results for
constant shear rate flows presented below.
3.2.1. Breakdown or rejuvenation experiments. In this flow,
the material is initially at rest for a long time, and hence
completely structured ( _gi¼ 0; li¼ 1). At time t¼ 0, the shear rate
is suddenly changed from _gi ¼ 0 to _gf, and kept constant until
steady state is attained.
The evolution of the shear stress s and structure parameter l
are given in Fig. 3 and 4, respectively, for different values of _gf. In
this flow, it is easy to see from eqn (35) and also from the
mechanical analog in Fig. 1 that s ¼ hN _gf at time t ¼ 0, whereas
s / hss( _gf) _gf as t / N (steady state), for all values of _gf.
However, Fig. 3 shows that the transient response of s depends
strongly on _gf.
The time needed for steady state to be reached increases
dramatically as _gf is decreased. When _gf is very small so that
lss z 1 ( _gf < _go), then the solution given by eqn (35) is valid. In
this case, using the fact that Gs x Go and hv x ho [ hN, it
reduces to
s ¼ _gf
�hN þ ho
�1� exp
�� Go
ho
t
��(36)
Therefore, in this case very large times (of the order of ho/Go)
are needed for the stress to reach its steady-state value, namely
s ¼ ho _gf. Moreover, while t � ho/Go the material behaves elas-
tically, i.e. the stress increases linearly with time:
This journal is ª The Royal Society of Chemistry 2011
Fig. 4 Structure parameter evolution for breakdown or rejuvenation
experiments: the material is at rest for a long time and, at t¼ 0, a constant
shear rate is imposed.
Fig. 5 Stress evolution for buildup or aging experiments: the material is
previously pre-sheared and, at t ¼ 0, a different constant shear rate is
imposed.
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
s z hN _gf + Go _gft (37)
Eqn (37) is just the asymptote of eqn (36) for small values of
Go
ho
t, and it could of course be inferred directly from the expected
behavior of the mechanical analog (Fig. 1) for this case.
Even in the range _go < _gf� _g1, when _gf is still small enough so
that q1(l) _gf (a Deborah number) is very large at all times during
the transient flow (see curves for _gf/ _g1 ¼ 10�3 and 10�5 in Fig. 3),
the stress increases linearly with time and then levels off as steady
state is approached. In these cases, the microstructure parameter
is always large and changes mildly during the flow. The transient
flow is thus dominated by elastic strain, and steady state occurs
at t � O(q1(lss)), when the microstructure ceases to deform
elastically and hence the stress levels off at so or sod, depending
on the value of _gf (see Fig. 2).
For larger values of _gf, the solution given in eqn (36) is valid
only up to the time when no appreciable breakdown has
occurred. Large values of _gf imply a fast stress growth, and thus
an early onset of microstructure breakdown. Because micro-
structure breakdown is not instantaneous (Fig. 4), the stress
keeps growing during the earlier stages of the breakdown
process, until the sharp breakdown related to yielding takes
place. This is the cause of the stress overshoot observed in Fig. 3,
in the curves for _gf/ _g1 ¼ 10�1 and larger. It is thus clear that the
stress overshoot is of elastic nature, and the maximum transient
stress attained at a given shear rate is sometimes referred to in the
literature as the ‘‘elastic yield stress.’’ In contrast to the dynamic
and static yield stresses, which are clearly steady-state properties,
the elastic yield stress is a shear-rate dependent transient quan-
tity.
Finally it is seen that, as the shear rate is increased, the ratio of
the maximum (elastic yield) stress to the steady-state stress
This journal is ª The Royal Society of Chemistry 2011
initially increases and then decreases. This trend is related to the
fact that the breakdown rate of the microstructure increases as
the stress is increased, which constitutes a regulating mechanism
that controls stress growth.
All the trends observed in Fig. 3 are plentifully corroborated
by data found in the literature e.g. see ref. 14,18.
3.2.2. Aging or buildup experiments. In this flow, the material
is initially flowing in steady state at a relatively high shear rate
( _gi/ _g1 ¼ 100 in the present example), and hence its structuring
level is low (li ¼ 0.12 in the present example). At time t ¼ 0, the
shear rate is suddenly changed from _gi ¼ 100 _g1 to _gf, and kept
constant until steady state is attained.
Fig. 5 and 6 show the model predictions for the buildup
experiments. The evolution of the shear stress s and structure
parameter l are given in these figures for different values of _gf.
For the buildup experiments, from eqn (35) it is clear that the
shear stress at t ¼ 0 is
s(0) ¼ hN _gf + (hss( _gi) � hN) _gi (38)
where the first term on the right hand side is the hN—dashpot
contribution, while the second term is the contribution of the
Maxwell element (see Fig. 1). Eqn (35) also gives that s /
hss( _gf) _gf as t / N, for all values of _gf.
It is first noted in Fig. 5 and 6 that the curves for _gf/ _g1 ¼ 103
and 105 actually correspond to breakdown experiments, inas-
much as _gf > _gi in these cases. They appear in these figures just to
illustrate the case of further micrustructure breakdown of a pre-
sheared material. It is seen that mild stress overshoots occur in
these cases despite the low structuring level at t ¼ 0, because
q1(li) _gf is still quite large in these cases (q1(li) _gf¼ 89.17and 8917,
respectively).
Soft Matter, 2011, 7, 2471–2483 | 2477
Fig. 6 Structure parameter evolution for buildup or aging experiments:
the material is previously pre-sheared and, at t ¼ 0, a different constant
shear rate is imposed.
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
For the microstructure buildup experiments ( _gf < _gi) relative to
the curves for _gf/ _g1¼ 10�5, 10�3, 10�1, 1, and 10, it is observed that
a steep l increase occurs in the time interval from _g1t � 0.1 up to
t� O(teq) (Fig. 6). Subsequently, either its steady-state value lss is
achieved ( _gf/ _g1¼ 1 and 10) or it mildly overshoots and then levels
off to lss ( _gf/ _g1¼ 10�5, 10�3, and 10�1). The l overshoot occurs at
lower shear rates because in these cases, during the elastic defor-
mation the stress increases very slowly and there is plenty of time
for microstructure buildup before the stress becomes large enough
to start causing appreciable breakdown.
Because the structure parameter does not change significantly
in the early stages of the flow ( _g1t ( 0.5), all the curves for the
buildup experiments nearly coincide in this range, displaying
a stress relaxation that obeys eqn (35):
sðtÞ ¼ hss
�_gi
�_gf þ
�hss
�_gi
�� hN
��_gi � _gf
�exp
�� GsðlssÞ
hssð _giÞ � hN
t
�(39)
Thus, this relaxation at early times occurs within times of the
order of (hss( _gi) � hN)/Gs(lss). It is clear that this relaxation time
would be zero if the material were initially completely unstruc-
tured (lss ¼ 0).
The steep l increase observed in the time interval from _g1t �0.1 up to t � O(teq) (Fig. 6) causes the stress increase observed in
Fig. 5. Due to the fast structuring, important changes in struc-
tural shear modulus (decrease) and structural viscosity (increase)
occur in this interval, causing the interesting transition observed
in the stress curves in this time interval.
Fig. 7 Shear rate evolution for viscosity bifurcation experiments: the
material is at rest for a long time and, at t ¼ 0, a constant shear stress is
imposed.
3.3. Constant shear stress flows
The predictions of the model for two different constant shear
stress flows are now examined.
2478 | Soft Matter, 2011, 7, 2471–2483
The first flow is characterized by a step change in shear stress,
i.e. it is initially steady at a given shear stress value, si ¼ hss( _gi) _gi,
and, at time t¼ 0, the shear stress is abruptly changed to another
value, sf. That is,
s(t) ¼ si + (sf � si)H(t) (40)
For the simple case in which both Gs and hv are constant, eqn
(13) can be integrated to give the following solution for _g:
_gðtÞ ¼ sf
�1
hv
þ�
1
hN
� 1
hv
��1� si
sf
�exp
�� hv
hv � hN
Gs
hN
t
�(41)
This equation is useful in the analysis of the results presented
in Sec. 3.3.1, where an example of this flow is discussed.
The second constant shear stress flow examined is the one
followed by pre-shear. The material is in steady-state flow at
a relatively high shear rate value, and then it is brought to rest
( _g ¼ 0) for some period of time after which a constant stress is
imposed until steady state is achieved. This flow is discussed in
Sec. 3.3.2.
3.3.1. Viscosity bifurcation when the material is initially at
rest. Fig. 7, 8, and 9 show results for the so-called viscosity
bifurcation experiments. The material is at rest and thus
completely structured (li¼ 1), and at time t¼ 0 the shear stress is
suddenly changed from zero to sf, and kept constant until steady
state is attained. The evolution of the viscosity h, shear rate _g,
structure parameter l are given in Fig. 7, 8, and 9, respectively,
for different values of sf.
When sf is very small so that lss z 1, then Gs x Go and hv xho [ hN, and hence eqn (41) reduces to
This journal is ª The Royal Society of Chemistry 2011
Fig. 8 Viscosity evolution for viscosity bifurcation experiments: the
material is at rest for a long time and, at t ¼ 0, a constant shear stress is
imposed.
Fig. 9 Structure parameter evolution for viscosity bifurcation experi-
ments: the material is at rest for a long time and, at t¼ 0, a constant shear
stress is imposed.
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
_gðtÞ ¼ sf
�1
ho
þ�
1
hN
� 1
ho
�exp
�� Go
hN
t
�(42)
Therefore, when sf is not large enough to cause notable
microstructure breakdown, the shear rate falls due to the
increasing elastic resistance of the deformed microstructure, and
a time of about 30hN/Go is needed for the shear rate to reach its
steady-state value, namely _g ¼ sf
ho
(thus very small). Moreover,
This journal is ª The Royal Society of Chemistry 2011
when the microstructure equilibrium time teq is significantly
larger than 30hN/Go, it happens that eqn (42) is always the
solution of eqn (13) for t < teq, even for large values of sf, because
significant microstructure breakdown starts at t � O(teq) only.
Thus, a pseudo-steady state is observed where _gxsf
ho
, roughly in
the time range 30hN/Go < t < teq (provided 30hN/Go < teq). Fig. 7,
8, and 9 illustrate this interesting behavior.
These figures also illustrate that, in the cases pertaining to sf/
sod ¼ 0.01, 0.5, 1, 1.5, and 1.98, steady state is achieved without
much breaking of the microstructure. However, the same is not
true for the curves pertaining to sf/sod¼ 1.99, 2, 3, 5, and 7, where
the steady-state viscosities are all several orders of magnitude
lower than ho, indicating a dramatic breakdown of the micro-
structure in these cases.
The viscosity bifurcation that occurs between sf/sod¼ 1.98 and
1.99 is related to the local maximum slmax (1.98 < slmax/sod <
1.99) that exists in the flow curve in the range of small shear rates
(see Fig. 2). Indeed, from eqn (32) it is easy to see that this local
maximum is slmax/sod ¼ 1.98547, and occurs at _glmax/ _g1 ¼1.3921 � 10�6. This bifurcation obviously occurs as a conse-
quence of the non-monotonic nature of the flow curve. When the
material is initially at rest and the stress is set to a value below
slmax, no important microstructure breakdown occurs, and the
steady state is achieved at a shear rate value smaller than _glmax.
However, when the stress is set to a value above slmax, no matter
how close to it, the microstructure eventually collapses, a shear
rate jump occurs, and the steady state is achieved at a viscosity
several orders of magnitude lower than ho. It is worth noting that
slmax is essentially equal to the static yield stress so, and so in
principle the flow corresponding to Fig. 7, 8, and 9 can be used to
determine experimentally the static yield stress.
It is especially remarkable that, for stress levels around the
yield stress (e.g. curve for sf/sod ¼ 1.99 in Fig. 7), the shear rate
seems to be leveling off to a negligibly low value, and, after a very
large delay time (O(100teq) in this example), it suddenly increases
steeply and then levels off to a much higher shear rate value. In
addition, this delay time decreases as the shear stress is increased.
This predicted behavior agrees with experimental evidence
available in the literature.3,17,19
3.3.2. Viscosity bifurcation when the material is pre-sheared.
Another type of flow at constant stress is now examined. The
material is pre-sheared to bring about a low structuring level in
steady state (li ¼ 0.12 in this example). Subsequently, the shear
rate is set to zero, and the stress is allowed to start relaxing. At
the moment t ¼ 0 when the relaxing stress reaches sf, it is kept
constant until steady state is achieved. It is assumed that no
significant changes in the structuring level occur during the
resting period (time interval between the cessation of
pre-shearing and t ¼ 0). Therefore, the initial shear rate is null
( _g(0) ¼ 0).
It is worth noting that negative values of _g(0) (recoil) would
occur if the flow started immediately after pre-shearing and the
pre-shearing stress were larger thanhvðliÞ
hvðliÞ � hN
sf (see eqn (41)).
On the other hand, if the experiment started after the stress
generated during pre-shearing had relaxed to zero, it is possible
that significant structuring would have occurred, which implies
Soft Matter, 2011, 7, 2471–2483 | 2479
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
an uncertain initial condition. For these reasons, it seems more
representative of laboratory practice to impose the _g(0) ¼0 initial condition described in the previous paragraph. It is clear
that each value of sf implies a different resting period. If instead
a fixed resting period were imposed, each imposed value of sf
would cause a different shear rate jump at t ¼ 0.
Fig. 10, 11, and 12 show the evolution of the shear rate _g,
viscosity h, and structure parameter l for this flow, for different
Fig. 10 Shear rate evolution for viscosity bifurcation experiments: the
material is pre-sheared, then brought to rest, and the stress is allowed to
relax up to s ¼ sf. From this moment (t ¼ 0) on, it is kept constant at sf.
Fig. 11 Viscosity evolution for viscosity bifurcation experiments: the
material is pre-sheared, then brought to rest, and the stress is allowed to
relax up to s ¼ sf. From this moment (t ¼ 0) on, it is kept constant at sf.
Fig. 12 Structure parameter evolution for viscosity bifurcation experi-
ments: the material is pre-sheared, then brought to rest, and the stress is
allowed to relax up to s ¼ sf. From this moment (t ¼ 0) on, it is kept
constant at sf.
2480 | Soft Matter, 2011, 7, 2471–2483
values of sf. In these figures, all the imposed stresses up to sf/sod¼10, correspond to steady-state structure parameter values lss
which are higher than li, which means that some microstructure
buildup is expected for these cases. The remaining two cases
shown, namely, sf/sod ¼ 50 and 100, correspond to breakdown
experiments, inasmuch as lss < li for these cases, and were
included for completeness to illustrate cases of further micro-
structure breakdown. The following discussion, however, focuses
on the cases pertaining to lss > li, which carry more interesting
features.
Fig. 12 shows that the structure parameter does not change
significantly until t� O(0.01teq), when it starts increasing to reach
a steady state at t � O(teq) or later, depending upon the level of
the stress sf. Thus it is not difficult to see that, while t < O(0.01teq)
the shear rate should obey
_gðtÞ ¼ sf
hvðliÞ
�1� exp
�� hvðliÞ
hvðliÞ � hN
GsðliÞhN
t
�(43)
Thus, for any shear stress value, the shear rate departs from
zero, increases monotonically and, at t � O((hv(li) � hN)hN/
(hv(li)Gs(li))), reaches a pseudo-steady state value, namely
_g ¼ sf
hssðliÞ, which lasts up to t� O(0.01teq) (Fig. 10). In this same
range of early times, for all shear stress values the viscosity
function departs from infinity at t ¼ 0, then decreases mono-
tonically and, at t � O((hv(li) � hN)hN/(hv(li)Gs(li))), levels off
to its pseudo-steady state value h ¼ hss(li), and remains
unchanged up to t � O(0.01teq) (Fig. 11).
This counterintuitive initial rise of the shear rate—even when
the imposed shear stress is much lower than the yield stress—has
been observed experimentally (e.g. see ref. 17 Fig. 3 therein), and
is explained through the present model as follows. At t ¼ 0 the
This journal is ª The Royal Society of Chemistry 2011
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
material is at rest but the microstructure is elastically deformed,
and the stress at the Maxwell element is sf. As time elapses, the
relaxing process continues at the Maxwell element, and thus the
stress at the Newtonian element (which was zero at t ¼ 0) must
increase, meaning that _g also must increase. It is thus clear that
this behavior is of elastic nature and is a consequence of a non-
zero retardation time. For this reason, the Maxwell-like model
proposed by de Souza Mendes6 is not able to predict such
behavior.
Similarly to what was observed with no imposed pre-shearing,
a viscosity bifurcation is also observed in Fig. 10, 11, and 12. In
this case the bifurcation indicates that a dramatic buildup of the
microstructure occurs when the stress is set to a value below
a local minimum slmin of the flow curve which exists in the range 1
< slmin/sod < 1.1. Actually this local minimum is slmin/sod ¼1.02694, and occurs at _glmin/ _g1 ¼ 6.2111 � 10�4 (see Fig. 2). In
this case it is clear that slmin is essentially equal to the dynamic
yield stress sod.
It is clear that the flows corresponding to Fig. 8 and 11 provide
respectively a lower and upper bound for the parameter _god that
appears in eqn (32), namely, _glmax and _glmin.
The above described trends predicted for startup flows at
constant shear stress by the present model are in qualitative
agreement with the experimental observations reported in the
literature e.g. see ref. 15,17.
Before proceeding to the next section, a final comment is in
order. The results presented in Secs. 3.2 and 3.3 were obtained via
simultaneous numerical integration of eqn (13) and (31).
However, the analytical expression given in eqn (35) turned out
to be an excellent approximation for s(t) in constant shear rate
flows, the same being true regarding eqn (41) for _g(t) in constant
shear stress flows, provided hv and Gs in these expressions are
seen as functions of l as given by integration of eqn (31).
In narrow time ranges containing stress overshoots and
sudden changes in time derivative, however, the just mentioned
approximation becomes poor. Nevertheless, this fact is certainly
potentially useful in the determination of some of the model
parameters.
3.4. Oscillatory flows
The model predictions for small-amplitude oscillatory flows are
now presented. As the results below will illustrate, the intro-
duction of a non-zero retardation time rendered the present
Jeffreys-like model far superior to the Maxwell-like model
recently proposed by de Souza Mendes,6 as far as the predictive
capability of oscillatory flows is concerned.
A sinusoidal shear g(t) ¼ gasin(ut) (ga is the shear amplitude,
and u is the frequency of oscillation) is imposed, resulting in
a shear rate _g(t) ¼ _gacos(ut), where _ga ¼ uga is the shear rate
amplitude. As usual in small-amplitude oscillatory flows, the
shear stress is assumed to be a linear function of both g(t) and
_g(t):
s(t) ¼ ga{G0sin(ut) + G00cos(ut)} (44)
where G0 and G00 are respectively the storage and loss moduli.
Differentiation with respect to time yields
This journal is ª The Royal Society of Chemistry 2011
_sðtÞ ¼ ga
�_G0 � uG00
�sinðutÞ þ
�uG0 þ _G0 0
�cosðutÞ (45)
The combination of eqn (44) and (45) with eqn (13), yields the
following differential equations for G0 and G0 0:
_G0 þ GsðlÞhvðlÞ � hN
G0 ¼ uG00 � hNu2 (46)
_G00 þ GsðlÞhvðlÞ � hN
G00 ¼ u
�hvðlÞGsðlÞhvðlÞ � hN
� G0�
(47)
The above equations are to be solved in conjunction with the
evolution equation for l, eqn (31). The expected solution for l(t)
is an initial transient—at the end of which the l level has adjusted
to the imposed shear stress level—followed by a periodic
response. However, when the typical times of change of l are
much larger than 1/u, then the amplitude of this periodic solu-
tion for l is negligibly small, and hence nearly time-independent
values of l, G0, and G0 0 are eventually reached. In this pseudo-
steady state, _G0 x 0, _G0 0 x 0, and the above differential equa-
tions reduce to the following expressions for G0 and G0 0:
G0ðlÞ ¼u2hvðlÞ
2GsðlÞ
�1� hN
hvðlÞ
�2
GsðlÞ2þu2hvðlÞ2
�1� hN
hvðlÞ
�2(48)
and
G0 0ðlÞ ¼uhvðlÞGsðlÞ2
1þ u 2
hNhvðlÞGsðlÞ2
�1� hN
hvðlÞ
�2!
GsðlÞ2þu2hvðlÞ2�
1� hN
hvðlÞ
�2(49)
Eqn (48) and (49) are equivalent to the well known expressions
for G0 and G00 as predicted by the classical Jeffreys model (usually
written in terms of the relaxation and retardation times, q1 and
q2), except that in the present model G0 and G0 0 depend on the
structure parameter l.
If the shear stress amplitude is sufficiently small to keep the
material structure intact, then hv ¼ ho [ hN and Gs ¼ Go �uho, and hence the above expressions reduce to
G0ohG0ð1Þ ¼u2h2
oGo
�1� hN
ho
�2
G2o þ u2h2
o
�1� hN
ho
�2x Go (50)
and
G00ohG00ð1Þ ¼uhoG2
o
�1þ u2
hNho
G2o
�1� hN
ho
�2!
G2o þ u2h2
o
�1� hN
ho
�2xuhN (51)
Actually, if the frequency u is not too low, then hv(l) [ hN
and Gs(l) � uhv(l) up to l values much lower than unity, and
hence, it is still true in this range that G0(l) x Gs(l) and that
G0 0(l) x uhN.
Therefore, the model parameters Go, m, and hN can in prin-
ciple be obtained directly from shear amplitude sweep data (see
Soft Matter, 2011, 7, 2471–2483 | 2481
Fig. 13 Structure parameter, storage, and loss moduli as a function of
the shear amplitude.
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
Fig. 13). This way of obtaining hN is particularly convenient due
to inertia problems that occur in steady-state flow measurements
in the high shear rate range.
The model predictions for a shear amplitude sweep test are
illustrated in Fig. 13. It is firstly observed that both G0 and G00 are
ga-independent (and equal to Go and uhN respectively) for
a wide range of ga, as discussed above. When ga reaches O(1), the
weak dependency of the structural shear modulus Gs on l
(because m ¼ 0.1) becomes visible in the G0 curve, i.e. it is clear
that in this range G0 increases mildly with ga, but subsequently it
drops sharply. In this same range, G0 0starts increasing steeply,
reaches a maximum, and from there on decreases with ga,
although less steeply than G0. The two curves cross (G0 ¼ G0 0) at
the same ga value where the G0 0 maximum occurs.
Fig. 14 Structure parameter, storage, and loss moduli as a function of
the frequency.
2482 | Soft Matter, 2011, 7, 2471–2483
The curve shapes of the storage and loss moduli obtained
experimentally for thixotropic fluids and reported in the recent
literature are typically very close to the ones seen in Fig. 13 (e.g.
see refs. 18,20,21). The observed trends are faithfully described
by eqn (48) and eqn (49). It is interesting to notice that the abrupt
changes in the G0 and G0 0 curves occur when the structural
viscosity is orders of magnitude lower than the one observed
when the material is at rest. This fact suggests that there is no
justification for the often employed method of yield stress
measurement based on stress amplitude sweep data.
Fig. 14 gives the model predictions for frequency sweep tests at
a fixed shear amplitude value. Both the moduli and the structure
parameter are given as functions of the frequency. The trends
observed in this figure are in full agreement with the ones typi-
cally observed experimentally for structured liquids (e.g., see ref.
22 p.92, Fig.12 therein). All the classical frequency sweep regions
are predicted: the viscous region, the transition-to-flow region,
the rubbery region, the leathery crossover region, and the glassy
region. The deepness of the G0 0 ‘valley’ and the slope of G0 on the
‘plateau’ and subsequent regions depend on the structuring level,
which varies with the imposed shear amplitude ga.
4. Final remarks
This paper describes a model for elasto-viscoplastic thixotropic
fluids that is a followup of the model recently proposed by de
Souza Mendes.6 It is composed of two differential equations, one
for stress and the other for the structure parameter. The equation
for stress is identical to the Jeffreys fluid viscoelastic model,
except that the relaxation and retardation times depend on the
structure parameter.
The model proposed here preserves all the advantages of the
former model, while achieving a significantly better predictive
capability. No additional parameter is needed.
The most remarkable features of the model are:
� The formulation is simple and all the assumptions are
justified by physical arguments;
� The concept of a neutral configuration that changes with the
microstructure introduced in ref. 6 is employed, which yields
a differential equation for stress that obeys a well defined
mechanical analog;
� A retardation time arises in the stress equation, yielding
a Jeffreys-like differential equation whose relaxation and retar-
dation times are structuring-level dependent. As a consequence,
the model is able to predict quite accurately even the most
complex trends related to viscoelasticity that are observed in the
rheological data for structured fluids found in the literature;
� As shown in ref. 6 the breakdown term of the evolution
equation for the structure parameter is assumed to depend on the
stress rather than on the shear rate;
� The evolution equation for the structure parameter is con-
structed such as to be consistent with the steady-state flow curve,
as proposed by de Souza Mendes.6 This is certainly one of the
main reasons for the excellent model performance;
� The non-monotonic steady-state viscosity function proposed
by de Souza Mendes6 is employed. This function allows the model
to accommodate two yield stresses, a static and a dynamic one.
Because the physics involved in the model are quite clear and
its predictions are in excellent qualitative agreement with
This journal is ª The Royal Society of Chemistry 2011
Publ
ishe
d on
09
Febr
uary
201
1. D
ownl
oade
d by
Im
peri
al C
olle
ge L
ondo
n L
ibra
ry o
n 01
/09/
2013
15:
34:3
5.
View Article Online
experimental observation for all flows investigated, then the
proposed model also constitutes an important tool to better
understand the fundamentals of the mechanical behavior of
structured fluids.
5.Acknowledgments
The author is indebted to Petrobras S.A., MCT/CNPq, CAPES,
FAPERJ, and FINEP for the financial support to the Group of
Rheology at PUC-RIO.
References
1 D. Bonn, S. Tanase, B. Abou, H. Tanaka and J. Meunier, Laponite:Aging and shear rejuvenation of a colloidal glass, Phys. Rev. Lett.,2002, 89(1), 015701.
2 J. R. Stokes and J. H. Telford, Measuring the yield behavior ofstructured fluids, J. Non-Newtonian Fluid Mech., 2004, 124, 137–146.
3 J. Mewis and N. J. Wagner, Thixotropy, Adv. Colloid Interface Sci.,2009, 147–148, 214–227.
4 H. A. Barnes, Thixotropy—a review, J. Non-Newtonian Fluid Mech.,1997, 70, 1–33.
5 A. Mujumdar, A. N. Beris and A. B. Metzner, Transient phenomena inthixotropic systems, J. Non-Newtonian Fluid Mech., 2002, 102, 157–178.
6 P. R. de Souza Mendes, Modeling the thixotropic behavior ofstructured fluids, J. Non-Newtonian Fluid Mech., 2009, 164, 66–75.
7 P. Coussot, A. I. Leonov and J. M. Piau, Rheology of concentrateddispersed systems in a low molecular weight matrix, J. Non-Newtonian Fluid Mech., 1993, 46(2–3), 179–217.
8 F. Yziquel, P. Carreau, M. Moan and P. Tanguy, Rheologicalmodeling of concentrated colloidal suspensions, J. Non-NewtonianFluid Mech., 1999, 86, 133–155.
9 D. Quemada, Rheological modelling of complex fluids: IV:Thixotropic and ‘‘thixoelastic’’ behaviour. Start-up and stress
This journal is ª The Royal Society of Chemistry 2011
relaxation, creep tests and hysteresis cycles, Eur. Phys. J.: Appl.Phys., 1999, 5(2), 191–208.
10 C. Derec, A. Ajdari and F. Lequeux, Rheology and aging: a simpleapproach, Eur. Phys. J. E: Soft Matter Biol. Phys., 2001, 4, 355–361.
11 P. Saramito, A new constitutive equation for elastoviscoplastic fluidflows, J. Non-Newtonian Fluid Mech., 2007, 145, 1–14.
12 P. Saramito, A new elastoviscoplastic model based on the Herschel–Bulkley viscoplastic model, J. Non-Newtonian Fluid Mech., 2009,158, 154–161.
13 R. B. Bird, R. C. Armstrong, O. Hassager, 1987. Dynamics ofPolymeric Liquids, 2nd Edition. John Wiley & Sons, New York,vol. 1.
14 K. Dullaert and J. Mewis, A structural kinetics model for thixotropy,J. Non-Newtonian Fluid Mech., 2006, 139, 21–30.
15 P. C. F. Møller, J. Mewis and D. Bonn, Yield stress and thixotropy:on the difficulty of measuring yield stresses in practices, Soft Matter,2006, 2, 274–283.
16 P. R. de Souza Mendes, Dimensionless non-Newtonian fluidmechanics, J. Non-Newtonian Fluid Mech., 2007, 147(1–2), 109–116.
17 P. Coussot, Q. D. Nguyen, H. T. Huynh and D. Bonn, Viscositybufurcation in thixotropic, yielding fluids, J. Rheol., 2002, 46, 573–589.
18 C. Derec, G. Ducouret, A. Ajdari and F. Lequeux, Aging andnonlinear rheology in suspensions of polyethylete oxide-protectedsilica particles, Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat.Interdiscip. Top., 2003, 67, 061403.
19 P. Coussot, Q. D. Nguyen, H. T. Huynh and D. Bonn, Avalanchebehavior in yield stress fluids, Phys. Rev. Lett., 2002, 88(17), 175501.
20 V. Carrier and G. Petekidis, Nonlinear rheology of colloidal glasses ofsoft thermosensitive microgel particles, J. Rheol., 2009, 53(2), 245–273.
21 M. Siedenb€urger, M. Fuchs, H. H. Winter and M. Ballauff,Viscoelasticity and shear flow of concentrated, noncrystallizingcolloidal suspensions: Comparison with mode-coupling theory,J. Rheol., 2009, 53(3), 707–726.
22 H. A. Barnes, 2000. A handbook of elementary rheology. University ofWales.
Soft Matter, 2011, 7, 2471–2483 | 2483