of COMPLEX FLUIDS
Jan Vermant
Department of Chemical Engineering
K.U. Leuven
The material in this lecture is covered by some textbooks, some of them however
being too limited, other ones being probably too detailed. For those among you who
want to know more about rheology we advice the following books:
Introduction to Rheology (Rheology Series) by H.A. Barnes, J.F. Hutton, K.
Walters, Elsevier Science Ltd (19992nd edition)
Rheology : Principles, Measurements, and Applications (Advances in
Interfacial Engineering) by Christopher W. Macosko, WileyVCH (1994)
The Structure and Rheology of Complex Fluids (Topics in Chemical
Engineering) by Ronald G. Larson , Oxford Univ Press (1999)
For those of you who want to see what is available on the WWW, good starting
places are:
The rheology index, a list of rheology related web sites
http://www.rheology.org/sorindex/default.asp
The European Society of Rheology : http://www.rheologyesr.org/
1
What is Rheology??
= flow
Hence the Newtonian fluid and the Hookean elastic solids are outside the scope of
rheology and material behaviour intermediate to these classical extremes will be
studied here. The term viscoelastic is used to describe this behavior. Some fluids
are however essentially inelastic, but have a viscosity which changes with the
deformation state, they are called NonNewtonian fluids.
Rheology is an applied science, and its aim is twofold: Firstly, rheologists try to
understand the relation between structure and flow properties.This important for the
intelligent design and/or formulation of materials for certain applications. Secondly, by
studying the material behavior using simple deformations, fundamental relations will
be derived between deformation and force. Such equations are called constitutive
equations. These equations can then be used to predict the material behavior in
complex deformation histories as they take place in typical process operations: e.g.
extrusion, polymer film blowing, spraying, pumping etc.
2
1. Viscoelasticity
I  Silly Putty
G(t)
t< t>
Polydimethylsiloxane + Borax
Dr. James Wright, a GE engineer, came upon the material by mixing silicone oil with boric acid.
Peter Hodgson borrowed $147, bought the production rights from GE, and began producing the goo.
He renamed it Silly Putty, and packaged it in plastic eggs because Easter was on the way.
Peter Hodgson's product left him an estate of $140 million at his death in 1976.
Probably the bestknown example of a material with a funny material is the socalled
Bouncing or silly putty. When this material is rolled into a small ball and dropped on
a surface, it will bounce. When it left to sit on a flat surface it will flow or like any liquid
will take on the shape of the container that holds it. Depending on the time scale on
which the deformations take place, the material will behave like either a solid (short
times) or a liquid (long times). But most of the the time the material will show an
intermediate behavior, displaying characteristic of a solid and a liquid at the same
time. We will see later that we can describe this by a timedependent modulus G(t).
We will also learn how we can quantify to which extent a material is fluid or solidlike,
by means of linear viscoelastic measurements.
Like the modulus, the viscosity is also not always a constant. Three examples are
given above. Mayonnaise is an example of a shearthinning fluid. When it sits on a
knife the only forces exerted on it are gravitational (low stresses) and the viscosity is
high. When the mayonnaise is applied to a sandwich, it is being sheared at high
stresses and the viscosity is low. The viscosity will be dependent on the rate at which
the material is being deformed, and decreasing as the shear rate is increased. This is
a behavior also typically encountered in polymers. Ketchup is an example of a
material which will not flow below a certain stress level, called a yield stress. It also
has a viscosity which, in addition is strongly dependent on the deformation history and
changes its value in time, this phenomenon is called thixotropy and is sometimes
encountered in suspensions. Concentrated dispersions can exhibit even different
features : a viscosity which increases with the deformation rate. This behaviour called
shearthickening.
3
2. NonNewtonian behaviour
II  Mayonnaise
 Ketchup "(!! , t )
 Starch
http://www.ccm.udel.edu/STF/index.html
Probably the bestknown example of a material with a funny material is the socalled
Bouncing or silly putty. When this material is rolled into a small ball and dropped on
a surface, it will bounce. When it left to sit on a flat surface it will flow or like any liquid
will take on the shape of the container that holds it. Depending on the time scale on
which the deformations take place, the material will behave like either a solid (short
times) or a liquid (long times). But most of the the time the material will show an
intermediate behavior, displaying characteristic of a solid and a liquid at the same
time. We will see later that we can describe this by a timedependent modulus G(t).
We will also learn how we can quantify to which extent a material is fluid or solidlike,
by means of linear viscoelastic measurements.
Like the modulus, the viscosity is also not always a constant. Three examples are
given above. Mayonnaise is an example of a shearthinning fluid. When it sits on a
knife the only forces exerted on it are gravitational (low stresses) and the viscosity is
high. When the mayonnaise is applied to a sandwich, it is being sheared at high
stresses and the viscosity is low. The viscosity will be dependent on the rate at which
the material is being deformed, and decreasing as the shear rate is increased. This is
a behavior also typically encountered in polymers. Ketchup is an example of a
material which will not flow below a certain stress level, called a yield stress. It also
has a viscosity which, in addition is strongly dependent on the deformation history and
changes its value in time, this phenomenon is called thixotropy and is sometimes
encountered in suspensions. Concentrated dispersions can exhibit even different
features : a viscosity which increases with the deformation rate. This behaviour called
shearthickening.
4
III. Weissenberg effect
N1
Dieswell of a fluid is another phenomenon which is typical for viscoelastic fluids. The
above photos compare the behavior of a Newtonian material with a polymeric fluid
exciting from a die. The ratio of the diameter of the jet to the capillary diameter is
113% for the Newtonian case (low Reynolds number) whereas it can be as high as a
few 100% for the case of polymeric fluids. This is a very important phenomenon when
extruding products. The dieswell will need to be accounted for (think about how you
would make a square duct by extrusion). Again the die swell arises mainly because of
differences in the normal stresses along and perpendicular to the stream lines in the
tube. They relax as the fluid exists from the tube causing a contraction in the
longitudinal direction. In the above pictures the Reynolds number is 0.001. The fluid
on the top is a Newtonian oil, the second fluid is an elastic fluid prepared in the same
manner as discussed previously (From Boger and Walters, Rheological Phenomena
in focus (Elsevier 1993).
5
IV. Die Swell
Weissenberg effect
McKinley et al (MIT)
N1
Dieswell of a fluid is another phenomenon which is typical for viscoelastic fluids. The
above photos compare the behavior of a Newtonian material with a polymeric fluid
exciting from a die. The ratio of the diameter of the jet to the capillary diameter is
113% for the Newtonian case (low Reynolds number) whereas it can be as high as a
few 100% for the case of polymeric fluids. This is a very important phenomenon when
extruding products. The dieswell will need to be accounted for (think about how you
would make a square duct by extrusion). Again the die swell arises mainly because of
differences in the normal stresses along and perpendicular to the stream lines in the
tube. They relax as the fluid exists from the tube causing a contraction in the
longitudinal direction. In the above pictures the Reynolds number is 0.001. The fluid
on the top is a Newtonian oil, the second fluid is an elastic fluid prepared in the same
manner as discussed previously (From Boger and Walters, Rheological Phenomena
in focus (Elsevier 1993).
6
V. Entry flows : vortex enhancement
N1, e
In more complex flows, the material characteristics will strongly alter the flow fields.
The above pictures are examples of flowvisualization of a contraction flow at different
flow rates. As the flow rate is increased, the fluid behaves more and more as a elastic
material. The Deborah number, which compares the characteristic time of the process
to the characteristic relaxation time of the fluid, hence expresses how important
elasticity will be. As the Deborah number is increased, Vortices are enhanced. These
vortices can even become unstable and start to swirl in the die. Two often related
aspects are important in entrance flow, the normal stresses and the elongational
viscosity. The fluid used in the experiments was a solution of 0.04% High
Mw polyacrylamide in a water/corn syrup mixture. This fluid is a socalled Boger Fluid,
which has an essentially shear rate independent viscosity and a pronounced elasticity.
The Reynolds number in the above experiments were small, ranging from 0.0005 to
0.02. The Deborah Number was 0.08 and increased to 0.2. If the flow rates are
increased further, the vortices become unstable and dance in the barrel, causing
unstable flow. This is not related to turbulence, but is an instability due to the elastic
nature of the fluid.
The above pictures are from : D.V. Boger, D.U. Hur, R.J. Binnington, J. NonNewt. Fluid
Mechanics, 20, 31 (1986) .
7
VI : Ductless Syphon
Tubeless syphon
McKinley et al (MIT)
8
VI : Reduction of Misting e
Elongational viscosity and elasticity of fluids can also be important in high Reynolds
applications, for example the breakup of a jet from a nozzle. The picture on the left
is for Water. The picture on the right is for a 200 ppm (!) solution of a high molecular
weight solution of a high Mw polyethylene oxide. (from JW Hoyt and JJ Taylor, Phys.
Fluis, 20 S253 (1977)). This is e.g. important in spraying or coating applications.
Another example of the same effect relates to drag reduction in turbulent flow.
9
Aim of this lecture:
10
Rheology
We now have a way to determine the state of stress in a material. Now we need a
similar threedimensional description of the state of deformation in a material. Relating
stresses and deformations leads us to the constitutive equations that we are looking
for. How do we proceed? Do we look to describe all the velocities in a fluid? This is of
not much use (fortunately) we need the relative deformations.
11
1. Deformation gradient tensor:
dx'
x
dx
z
Uniaxial elongation
dx = F.dx
Consider a cube of material as drawn above. Two points in that body are separated
by a vector dx at time t . Now let the body be deformed to a new state, e.g. the
deformation given above is a uniaxial elongation. The two points will move with the
body and the displacement vector between them will be stretched and rotated as
indicated by a new vector dx. Now we wish to relate dx (the present state) to dx (the
rest or past state at t). This will be given by another linear transformation, called the
deformation gradient tensor F.
Just as the stress tensor describes the force acting on any plane, the deformation
gradient tensor relates the state of deformation and rotation for any material point.
For the example of uniaxial elongation stresses will develop along the normal axis of a
piece of e.g. rubber. Keeping in mind Hookes law, we hence know what to expect for
the deformation tensor (which components will be nonzero?).
12
Simple shear ! = s/h
S
Above are two other simple deformations which can be given to bodies. They are a bit
more critical to check whether or not we have a good deformation measure: The
simple shear deformation will lead to shear stresses, and again based on Hookes law
(proportional) now know to expect shear components in the deformation measure. By
the symmetric nature of the stress tensor, both the xy and yx components should
also be present in a suitable deformation tensor.
Likewise, a solid body rotation forms another critical test. Material elements are not
stretched, only rotated and no stresses should develop. A suitable deformation
measure should hence yield the identity tensor for this type of deformation
13
! (x (x (x $
# (x' (y ' (z'&
# (y (y (y & Deformation gradient tensor
F=#
(x' (y ' (z'&
## (z (z (z &&
" (x' (y ' (z'%
F is hence not a good deformation measure for our purpose. It describes both rotation
as well as a shape change. Somehow we must eliminate this rotation. The material
response is determined only by the amount of stretching, not by a solid body rotation.
14
2. The Finger tensor:
Simple shear
Uniaxial elongation
" *2 0 0 %
!1 $ !1 '
C =$ 0 * 0 '
$ 0 0 * !1'
# &
We know from matrix algebra that multiplying a matrix by its transpose, yields a
symmetric matrix. Can we, by doing this also remove the rotation and get a symmetric
tensor? We can express the deformation gradient F as a product of both a stretch and
a rotation. F= V.R. as a rotation followed by an inverse rotation gives I:
F.FT = V.R.RT.VT = V.VT
The dot product of F with its transpose is called the Finger Tensor, and it is a good
deformation measure. Note that other deformations measures are possible e.g. FT.F.
The physical meaning of the Finger tensor is that it relates the change in area within
the sample.
The good news is now that we can write Hookes law in 3D!
=GC1
15
2. The Finger tensor:
Simple shear
Uniaxial elongation
" *2 0 0 %
!1 $ !1 '
C =$ 0 * 0 '
$ 0 0 * !1'
# &
We know from matrix algebra that multiplying a matrix by its transpose, yields a
symmetric matrix. Can we, by doing this also remove the rotation and get a symmetric
tensor? We can express the deformation gradient F as a product of both a stretch and
a rotation. F= V.R. as a rotation followed by an inverse rotation gives I:
F.FT = V.R.RT.VT = V.VT
The dot product of F with its transpose is called the Finger Tensor, and it is a good
deformation measure. Note that other deformations measures are possible e.g. FT.F.
The physical meaning of the Finger tensor is that it relates the change in area within
the sample.
16
3. NeoHookean Solid:
Using the Finger tensor we can come up with a general form of Hookes law
Simple shear
Uniaxial elongation
$+ 2 0 0 '
! = G.C = G.& 0 + "1 0 )
"1
& )
&% 0 0 + "1 )(
G
! xx = " p + G+ 2 = " G(1 + , )2
1+ ,
3, + 3, 2 + , 3
=G  3, G = E,
1+ ,
The good news is now that we can write Hookes law in 3D!
=GC1
Actually Hooke observed the proportionality between force and stress for a springy
body as early as 1678 and it took almost 150 years to formulate this model in 3D.
For an incompressible solid, the above calculation shows the link between the shear
modulus G and Youngs modulus E. For compressible materials, a Poissoncoefficient
needs to be used. For more details we refer to the textbook by Macosko
or the course Sterkteleer.
17
" )v x )v x )v x %
4. Velocity Gradient: $ )x )y )z '
$ )v y )v y )v y '
!v = $ '
$ )x )y )z '
)
$ z v ) v z ) v z'
# )x )y )z &
uniaxial extension
v x = *! x
(*!
vy = y
2
(*!
vz = z
2
" %
$ *! 0 0 '
$ (*! '
!v = $ 0 0 '
$ 2 '
$0 (*! '
0
# 2&
The velocity gradient tensor describes the steepness of the velocity variation as one
moves from one point to the other. This tensor has again two directions associated
with it : one of the velocity and one of the gradient. It operates on a local
displacement vector to to generate the magnitude and direction of the changes in the
velocity.
For the case of uniaxial elongation, it seems to give a good deformation measure.
But once again, solid body rotations and shear flows are more critical tests.
18
y
Simple shear U
v x = (!y d
vy = 0
vz = 0
x
no slip
" " 0 ( 0%
! U
!v = $ 0 0 0 '
linear velocity field: u x =  y
d
$ 0 0 0'
# &
stretching rotation
Lets consider the case of steady state shear flow. Again, we want the deformation
measure we use to yield a symmetric tensor in order to write our constitutive
equations. The kinematics of simple shear flow can be obtained by taking the time
derivatives of the displacement functions of simple shear we have used before.
We see that the velocity gradient tensor is not symmetric! Like the deformation
gradient F, the velocity gradient tensor contains both rotation as well as stretching.
A steady state shear flow can be viewed as a combination rotation and stretching. We
can remove the rotation from the velocity gradient by recalling that we could
decompose F in terms of a stretch and rotation
F= V.R if we differentiate this with respect to time:
F! = V
! !R + V !R
!
lim V = lim R = I
x'" x x'" x
lim F! = #v = V
! +R
!
x'" x
lim F! T = V
x'" x
! T
! +R ( )
! = 2D = (#v)T + #v
V
19
5. Rate of Deformation Tensor
T
2D = (!v) + (!v)
uniaxial extension
(" % " .! 0 +
1 *$ .! 0 0 0 .!
' + $ 0 !.! / 2 0 %'  = "$ 0 !.! / 2 0 %'
0 0
D= 0 !.! / 2 0
2 *$ ' $# 0 0
' $
!.! / 2&  # 0 0
'
!.! / 2&
)# 0 0 !.! / 2 & ,
Simple shear
! 1 $
# 0 ! 0&
' 0 ! 0$ 2
1 )!# ! 0 0 0$ * # 1 &
D= 0 0 0& + # ! 0 0& , = # ! 0 0&
2 )#" 0 0 0&% # &
( " 0 0 0% ,+ # 2 &
# 0 0 0&
# &
" %
Hence we obtained a new tensor 2D, which should be symmetric as is checked above
for the two types of deformation : uniaxial extension and shear flow.
20
General Newtonian Constitutive Equation
! = 2 (D y
U
" 1 %
$ 0 )! 0' d
$1 2 '
! = 2($ )! 0 0'
$2 ' no slip
x
$ 0 0 0' U
linear velocity field: u x =  y
$ ' d
# &
We now have derived all the tools that we need to extend Newtons law in three
dimensions. Note that the above equations relate the extra stress tensor and the rate
of deformation tensor. When we write the equation in its components for simple shear
flow, we find the definition that we know from e.g. transport phenomena or hydraulics.
The Newtonian constitutive equation is the simplest equation there is for viscous
liquids and forms the basis of fluid mechanics. When you are faced with a new
problem, this model will be the best place to start. Fluids which obey Newtons law are
e.g. lowmolecular weight liquids, dispersions at small volume fractions of dispersed
phase, but even high Mw polymers at low deformation rates
21
uniaxial extension
" (! 0 0 %
D = $$ 0 ! (! / 2 0 ''
# 0 0 ! (! / 2&
+! = (!
) = 2 *D
)11 = !p + 2 *(!
) 22 = ) 33 = !p  *(!
) ! ) 22
*e = 11
(!
*e = 3 * TROUTON ratio
The above calculation shows the value of using the tensor form of Newtons law. The
result that the extensional viscosity is three times the shear viscosity is also observed
for polymer melts at low rates. It is actually often used to check the validity of results
of elongational rheometry and we will probably come back to this.
22
Linear Viscoelasticity
In this chapter we will see one of the most important aspects of rheology: the description of linear
viscoelasticity. Linear viscoelasticity is important for a number of reasons:
It is an important goal of rheology that we are able to describe and characterize material
behavior that is intermediate between solid and liquid behaviour.
The linear regime is important because it is the building block for further work: how
complicated the nonlinear regime might be, we should always find the linear viscoelastic
behavior as a limiting case.
Last but not least linear viscoelasticity is of practical importance: For example for linear
polymer chains, measurements characterizing the linear viscoelastic properties provide insight in
to how the material will behave in some nonlinear flows.
23
What are viscoelastic fluids?
Memory = Memory = 0
Real fluid/solid
Viscoelastic fluids are defined as fluids which have a finite memory. The two material
classes we have worked with up to now; the perfect solid (Hookes law) and the
perfect liquid (Newtons law) do not obey this definition. A perfect solid will remember
the deformation indefinitely and return to its initial shape. A perfect liquid will stop
deforming instantaneously when the stress is removed and never recover to its
previous shape.
All real fluids and solids show intermediate behavior. I cannot recuperate the energy
that I put into a deformed solid. There will always be some internal dissipation of
energy. For example car tyres heat up during driving, not soo much due to the friction,
but due to the internal heat dissipation.
24
What are viscoelastic fluids?
Real fluid/solid
When we apply a stress and a body is deformed, this implies that workis being defined as force x
deformation, is done. The perfect solid will store this energy. According to Hookes law, the potential
energy
"
E = # d"
cannot be dissipated. If a perfect body would be deformed by a certain! xy xy
deformation g, and subsequently it
0 liquid on the other dissipates all the
would be released, it would oscillate indefinitely (in vacum). A perfect
energy , to keep the fluid going I need to put energy into the fluid.
All real fluids have a finite memory. Even water has a time constant of about 1012 s. Deformed slower than
that it will flow, deformed more rapidly it will behave as a solid. Some materials have time constants on the
order of 1017s. Most polymeric liquids will have time scales which are on the order of 103 to a few seconds.
In addition , they will often show a wide distribution of relaxation times.
Whether or not you have to take viscoelasticity in to consideration when designing a process or product will
depend on the time scales of the product and the process.
25
1. Viscoelastic phenomena :
1.a. CREEP
Compliance
! ( t)
J( t) =
"0
The inverse of the creep experiment is an experiment were a strain is suddenly applied and the
resulting stress is measured. Either the a constant strain rate can be applied or a a constant
strain is suddenly applied. The stress in a viscous (Newtonian) liquid (VL) will drop to zero
instantaneously. The elastic solid (ES) reacts to a constant strain by an instantaneous stress
that remains constant in time.
What happens with a VE fluid in a stress relaxation experiment? As the name of the experiment
suggest, the stress gradually relaxes. We can convert this data into a RELAXATION modulus .
Note that the above examples are for shear deformations but the concepts and models we
discuss can be applied to other types of deformations. But because we are dealing with small
strains, typically only shear
deformations are of interest.
Questions :
Stress relaxation experiments are not always in the linear response regime. Think about what
condition needs to be satisfied for a step strain experiment to be a linear measurement.
26
1.B. stress relaxation
fast
Apply strain
Relaxation modulus
!( t )
G( t) =
"
Questions :
Stress relaxation experiments are not always in the linear response regime. Think
about what condition needs to be satisfied for a step strain experiment to be a linear
measurement.
27
LDPE [Laun, Rheol. Acta, 17,1 (1978)]
The figure shown above contains the relaxation modulus for a low density
polyethylene at 150C (data from H.M. Laun, BASF, Rheol. Acta 1978). Data at two
different strains are shown. Note that the time scale over which the modulus
decreases expands up to a few 100 s!!! So this material has considerable memory for
the applied deformation. The data is represented on a loglog scale. The modulus
decreases rapidly, but it is not a single exponential decrease.
Remark:
Come back to this figure at the end of this chapter: then you will understand that the
line in the figures represents a fit to eight exponential relaxation times.
28
1.C. Sinusoidal oscillations
STRAIN:
! = ! 0 sin("t )
! = ! 0 exp(i"t )
STRAIN RATE
! = i "! 0 exp(i"t )
HOOKEAN SOLID
# = G! = G! 0 sin("t )
# = G! 0 exp(i"t )
NEWTONIAN FLUID
The creep experiments are especially suited for the long time scales. The stress
relaxation experiments can access time scales between 0.1 (because of transducer
resolutions) and a few 100 s. So we need other types of experiments in which the
time scale can easily be varied. The small strain oscillatory measurements present
us such a tool!
The sample is deformed sinusoidal. After a few cycles of startup, the stress will also
oscillate sinusoidal. For a Viscous Liquid (VL) and an Elastic Solid (ES) the response
can be easily calculated by invoking the appropriate constitutive equations. From the
above equations it can be seen that the stress in an ES will oscillate IN PHASE with
the applied deformation, whereas for a VL the stress will be DEPHASED by 90 with
respect to the oscillatory deformation. Again, we found a way to compare a elastic
solid and a viscous liquid. Note also that for the VL the stress amplitude will depend
on frequency whereas for an ES it will not!
29
VISCOELASTIC MATERIAL
! = ! 0 sin("t + # )
! = ! 0 exp(i"t + # )
! = G &$
! = G & $ 0 sin("t + # )
! = (G '+ iG " )$
G' '
Storage and Loss modulus tan ! =
G'
Now what do we get for a VE liquid: The total stress can be viewed as the combination of in
part a viscous and in part an elastic component, resulting in a complex modulus with a real
elastic part, and an imaginary (viscous) part The stress response will be dephased by an angle
. This angle will be a measure for how elastic or how viscous my fluid is. If the fluid is viscous
the phase angle will tend to 90. When the fluid is dominantly elastic, the phase angle will tend
to zero. The phase angle is hence a nice way to characterize the material on a viscoelasticity
scale. It is convenient to analyze the stress wave into two waves, one in phase and the other
one out of phase with the strain wave. This decomposition suggest the existence of two
dynamic moduli:
G the inphase, elastic or storage modulus
G the outofphase, viscous or loss modulus.
Hence we see the definition of VE in terms of energetic behaviour reappearing from this
experiment. The figure on the right shows oscillatory data on the same LDPE as shown before
in a creep experiment. We will try later to calculate the creep from the dynamic moduli (a nice
feature of linearity all Linear viscoelastic functions are interrelated)
30
Newtonian MATERIAL
# = %"!
# = [G '! sin($t ) + G"! cos($t )]" 0
# = ( G" ) "
G" =
G = Loss modulus tan ! = ?
!=?
Elastic " (Hookean) MATERIAL
$ = G#
$ = [G '! sin(%t ) + G"! cos(%t )]# 0
$ = (G ' ) # tan ! = ?
G' =
!=?
For Hookean solids, stress is proportional to strain and we find another limiting result
G is thus known as the elastic storage modulus. It characterizes the elastic
contribution to the stress response. Remind yourseld : what is the phase angle in this
case?
31
VISCOELASTIC MATERIAL
! = ! 0 exp(i"t + # )
! = $3 %!
G"
$' =
"
G'
$" =
"
1/ 2
' G '* 2 ' G "* 2 0 1 3
$3 = /) , + ) , 2 = G
/.( " + ("+ 2
1 "
Another way to view the experiments is in terms of of a sinusoidal strain rate. This
leads to the definition of a dynamic viscosity. This is also a material function. We can
now define a dynamic viscosity for the magnitude of the viscous stress to the
strain rate and an elastic part of the dynamic viscosity .
The typical low frequency behaviour for the materials function for polymer melts and
solutions is sketched in the above graph (A). It shows a plateau modulus G0 and
decreases with 2 for G and proportional to for G at low frequencies. This is an
important feature to check when doing measurements. For rubbers (fig: B) on the
contrary G is essentially constant over the entire frequency range.
32
Mirror
relation
Terminal zone :
Time (s)
When the frequency range for a polymer melt is extended, we find again the
behaviour that is known from the courses on polymer science. The different regions
are indicated on the graph.
The graphs on the right hand side also indicate the link between dynamic moduli
and the relaxation modulus. In an oscillatory experiment, the sample is
interogated at each frequency. The relaxation modulus corresponds to an integral
over the spectrum of relaxation times : i.e. at short time all modes contribute to the
stress, whereas some start to relax as time increases.
For a VE liquid the limiting relations for the terminal zone, shown above should
always be true when the material is tested in the linear viscoelastic regime. Hence
such relations are extremely useful for testing the internal consitency of data. You can
use the relations
above to check whether data from different types of experiments agree and even if
data from nonlinear experiments gives in the limit for vanishing shear rates the VE
result.
33
Effect molecular weight on
the LV moduli of polymer
melts
The above graphs show the dependence of the moduli on molecular weight for a
set of monodisperse PS polymer melts. These results point us to an application of
rheology as an analytical tool : LV measurements will give us detailed information
about the distribution of molecular weigths. To do this well need to develop a
theoretical description. Even than theres a catch : the calculation of the molecular
weight distribution is an example of an inverse problem (and hence intrinsically and
often numerically difficult).
34
Effect of volume fraction
on the dynamic moduli of
dispersions ! d 2* $
# 2&
" dr % 5+ ! G ') 6, c $
F( r ) = = # ( &
d c ' kT 0.639 N " kT +d 3c %
104 "c
0.421
10 3
0.426
: pairwise interaction potential
0.437 N : coordination number (7.5)
0.452
102 0.455
G: high frequency modulus
0.471 dc : particle core diameter
101
0.482
0.502
c : core volume fraction
0.515
0.534
100 !(r)/kT !d/kT
G' (Pa) G"(Pa)
1
10
102
102 101 100 101 102
! (rad/s)
!(r)/kT
The above graphs show the dependence of the storage modulus on frequency and
volume fraction. Again an application of the linear viscoelastic measurements arises :
we will be able to use rheological measuremenbts to characterize the interactions
between particles. As we will see in the chapter on microrheology of dispersions, we
will make use of rheology as a sensitive measurement technique to asses dispersion
stability and predict the (nonlinear) viscosity curves.
35
2. Differential models : The Maxwell model
Phenomenological models
Hookean Spring
! 1 = G" 1
Newtonian Dashpot
! 2 = " $ #! 2
Maxwell KelvinVoight
Lets now try to describe the dual nature of these fluids and solids by simple
mathematical models. We can start from simple phenomenological tools. A Hookean
spring will be our model for a perfect solid. A dashpot will be our conceptual model for
a perfect fluid. We can now start combining these elements in parallel or in series and
so on. We could as well draw an analogy on electrical elements, using resistors and
capacitances.
Maxwell (1867) first proposed the in series model for the viscosity of gases.
Question : from the above cartoons, which one will be the best for describing a fluid?
Which one will describe a solid?
36
Maxwell model = VE Fluid ! = !1 + ! 2
"! 1 " 2
!= +
G0 #0
%# (
" + ' 0 * "! = #0 !
& G0 )
Now lets assume we take the Maxwell model, using the elements in series. When we
apply a deformation to this model the total deformation will be split by both elements.
The stress in both elements is the same and equal to the total stress. Assuming a
homgeneous deformation, the deformation rate can also be assumed to consist of a
sum of the two individual motions.
We can then rewrite the equations, leaving us with a differential equation. This
differential equation is know as the MAXWELL model; In the next few pages we will
try to see if this model can capture some of the observed phenomena of linear
viscoelasticity.
37
Relaxation (Step strain) experiments for a Maxwell element
12
G0
! = !0 t $ 0 10
d" "
"+# =0
6
G(t)
dt 4
t = 0; " = G0 ! 0 2
"(t ) 0
! Time (s)
! [rad/s]
Fading memory
Lets first calculate the response of the Maxwell model to a step in strain. As boundary
conditions we impose the strain of g0 at times > 0. We can easily calculate the stress
at time zero (purely elastic response) and solving the differential equation gives use
the time evolution of the stress and of the relaxation modulus.
The full line on the curve gives the predictions of the above equations for a G0 of 10
Pa and a relaxation time of 1 s on a semilogarithm plot. In reality, a single
exponential is not always observed (remember the relaxation curve for the LDPE we
showed previously). The dotted line shows a more gradual evolution of the relaxation
modulus. We will see how we can modify the Maxwell model to take into account this
behaviour.
38
Dynamic moduli
10 G0
G0 ! 2 " 2
G '(! ) =
1 !
2 2
1+ ! "
0.1
G0 !"
G "(! ) =
Moduli [Pa]
1 + ! 2"2 0.01
G'
G"
0.001
w [rad/s]
Likewise you can calculate the response of a Maxwell model to an oscillatory strain. It
is most easily done by using complex notation.
What you can easily see from the evolution of the moduli is that for frequencies
smaller than the inverse of the relaxation time, a dominantly viscous response is
observed. This can be extended to all processes with slow motions indeed the
dashpot or Newtonian behavior will dominate. At high frequencies the elastic
behavior dominates. For rapidly changing stresses in general, the derivative term in
the Maxwell model equation dominates and at short times this model then approaches
elastic behavior.
39
Do Maxwell Fluids exist?
Do Maxwell fluids exist? Well at least some fluids do behave according to the simple
maxwell model, at least to a good approximation. These are fluids in which one very
dominant relaxation mechanism exists. Two examples are given above:
Emulsions : here the VE comes from the interface, the material now can store
elastic energy as the droplets, which have an interfacial tension, can deform. The
relaxation time of these fluids is determined by the medium viscosity, the droplet size
and the value of the interfacial tension. When the droplets are fairly uniform in size,
the system will respond with essentially one single relation time. The graph on the left
compares two emulsions which were subjected to different mixing histories.
Surfactants : A lot of shampoos come relatively close to being Maxwell fluids! Such
surfactants can form  in a certain range of concentration  rodlike micellar objects.
These objects are sometimes called Living Polymers as they resemble polymeric
fluids but they can break and reform without too much effort. The characteristic time
scale of these systems is the breaking or reconnection time.
40
Can we describe polymer melts?
Come back to this page and explain what the right hand side graph means.
41
Generalized Maxwell model
!TOT = % ! i
i
! i + " i !! i = #i $!
Jeffreys model
% d$! (
! + "1!! = #' $! + " 2 *
& dt )
An obvious way to generalize the Maxwell model is to allow for multiple relaxation
times. This can be done by considering multiple dashpotspring combinations in
parallel. In this manner, several relaxation times are available to fit our experimental
data.
The relaxation data of the PDMS on the previous page could be described by
selecting a set of relaxation times equidistantly spaced on a logarithmic axis and
determining the Gks for the different modes by a linear regression analysis. (k = 0.01,
0.1, 1, 10, 100 s, Gk=2*105 , 105, 104,102,10) .
Some molecular theories, like the Rouse theory for polymer dynamics (see later)
suggest an infinite series of relaxation times where only two fit parameters remain.
"k
Gk = G0
! "k
k
"O
"k =
k2
42
3. General viscoelastic Integral model
t
&(t ) = # G(t ! t ' )d%(t ' )
!"
t
d%
&(t ) = # G(t ! t ' ) dt '
!"
dt '
fluid
t
&(t ) = G(0)%(t ) ! G(")%( !" ) ! # %(t ' )dG(t ! t ' )
!"
Reference state @ t=0
s =t !t'
dG(s )
M (s ) = memory function =
ds
dg(s)
M(s) = is the memory function
ds
When the steps are taken to be infinitesimally small, the summation becomes an
integration. Note that in rheology it is customary to use the configuration at time t as
the reference. The equation can then be transformed by an integration by parts.
Noting that G()=0 for fluids and is finite at time t = that (t)=0 is a reference for
the deformation.
43
What is the relaxation function?
For a simple Maxwell model we used a single exponential
G(t ) = G0 exp( !t / +)
*
# !t &
G(t ) = ) F ( +) exp% ( d+
0 $ +'
Often one use a logaritmic scale : the relaxation spectrum is defined as follows
*
# !t & d+
G(t ) = ) H ( +) exp% (
0 $ +' +
Powerful way to characterize
materials  albeit in the Linear region
The use of G(s) does not lead directly to a specific equation for the stress relaxation
function. Specific expressions can be developed. By using a simplest expression is a
single exponential decrease. In this manner we obtain the integral formulation of the
Maxwell model. The integral form of the generalized Maxwell model can be easily
obtained by using the relaxation spectrum as defined above.
44
Oscillatory flow G and G from the general linear viscoelastic fluid:
! = ! 0 " cos("t )
t
# = ' G(t $ t ') % ! 0 " cos("t ')dt
$&
&
= ! 0 " ' G(s ) % cos("t $ "s )ds
0
& &
= ! 0 " ' G(s ) % cos("s )ds cos("t ) $ ! 0 " ' G(s ) % sin("s )ds sin("t )
0 0
&
G " = " ' G(s ) % cos("s )ds
0
&
G ' = " ' G(s ) % sin("s )ds
0
For all linear viscoelastic materials, the shear stress is expected to be a sinusoidal
function when a sinusoidal strain is applied. The above calculation demonstrated how
the response for an arbitrary linear viscoelastic fluid can be calculated (continued on
the next page). The only mathematical tool we use is that
cos(a+b)=cos a.cos b + sin a . sin b
45
( ( N
" %
G " = , * G(s ) ! cos(,s )ds = , * $ + Gi e )  i s ' ! cos(,s )ds
0 0 # i =1 &
N , i
= + Gi
i =1 1 + (, i ) 2
(
G ' = , * G(s ) ! sin(,s )ds
0
N (, i ) 2
= + Gi
i =1 1 + (, i ) 2
46
Oscillatory flow !'(" ) = ? Complex viscosity
G"
!'(" ) =
"
1,
= . G(s ) $ " $ cos("s )ds
"0
1 , %, H ( # )  # (
s
= . '. $ e d# * $ " $ cos("s )ds
" 0 '0 # *)
&
H ( # ) %, (
, s

= ' . $e # cos("s )ds * $ d#
.
0 # '
&0 *)
%, ( b
+ ' . $e  bt cos(at )dt * = 2 2
for b > 0
&0 ) a +b
,
H ( #)
!'(" ) = . d#
0 1 + ("#) 2
Here we give another example how a VE function, in this case the complex viscosity 
can be calculated. This time we have used a spectrum rather than a discrete set. You
can try to show the what the equation becomes for ()
47
Some relations between viscoelastic functions
2 # G'
G( t) = % ' ! sin('t)d'
&0'
#
( = % H( )))d ln )
$#
2 # G'(' )
(= % d ln '
& $# '
dG' 1 d2 G'
H( )) " +
d ln(' ) 2 d(ln ' )2 1/ ' = ) / 2
The relaxation spectrum also gives us a tool to interrelate the different viscoelastic
response functions. The different response functions can hence be calculated from
each other. This can be either time saving or can be used to verify the internal
consistency of experimental data. Combining data from different types of experiments
can also be used to calculate the relaxation spectrum (it is an illdefined mathematical
problem).
48
Conclusions
Defined and described linear viscoelasticity.
Using rheology as a spectroscopic technique!
MATERIAL FUNCTIONS :
This chapter we have taken a big step in describing rheological phenomena. Linear
viscoelasticity is important in practical applications. Some real life deformations will be
in the linear regime (especially for solid state rheology). Secondly, understanding and
describing the fading memory effect is an important aspect of rheology. We have also
provided a number of material functions that can be used to calculate relaxation
spectrum and that are interrelated. In the next chapter we will now see how these
material functions can be made nonlinear. Then we will learn how all these
properties can be measured.
49
Applied Rheology
Generalized Newtonian
Fluids
Although Newtons law is of much practical use, at the beginning of this century
researchers started to realize that not all fluids obey this simple linear relation. Many
colloidal suspensions and polymer solutions show a decrease of the viscosity as the
shear rate is increased.
In the following, we will try to extend Newtons law in the most simple fashion, by
making the stress depend on the rate of deformation (or better : on an invariant of the
rate of deformation tensor). But as we are talking about viscosities and rate of
deformations, maybe it is useful to start of with indicating a few orders of magnitudes
which are likely to be encountered
50
Liquid [Pa.s]
The SI unit of viscosity is Pa.s (Pascal.second). The cgs unit of Poise (P) is also often
used. 1 poise = 0.1 Pa.s. Centipoises are also often encountered, as 1 cP is the
viscosity of water.
51
Process shear rate [s1] Applications
The range of shear rates encountered during various processes spans many
decades. Rheological properties are important in slow processes such as
sedimentation; some products will be tailored too have a very high viscosity at low
shear rates for a good shelflife. Some processes such as rubbing or chewing
might seem odd to you, but the rheological properties determine to a large extent the
waqy these products feel . Coating flows and printing operate at high rates. The
flows, however, are still laminar.
It will be quite a challenge to develop measurement techniques that can embrace this
entire shear rate range!
52
Do real fluids behave according to Newtons law?
Just some examples of NonNewtonian fluids are given above. Human blood (a) is a
complex fluid with a very variable rheological behaviour. The most important
consituents (rheologically speaking) are the red blood cells, which implies that it can
be viewed as a dispersion of deformable particles. Another complex dispersion is
yogurt (rheology can be used here as a tool to detect fermentation). Another natural
material with a pronounced shearthinning behaviour is Xanthan gum (b) . Here the
structural units are macromolecules which interconnect to form a network that is
constructively broken down by the shear flow.
But increasing the shear rate does not always result in a decrease of the viscosity. In
some cases a shearthickening behaviour is observed, e.g. for concentrated
dispersions containing large particles or associating polymers. As an examoke data
for latex spheres in diethyleneglycol is given in fig (d).
Note that a nonlinear behaviour of the viscosity not necessarilly implies visco elastic
behaviour.
53
1. General viscous fluid
! = f(2D)
! = " pI + a1 # D + a2 # D2 + a3 # D3 +...
The simplest manner to extend Newtons equation is to have a stress tensor that only
depends on the rate of deformation tensor. The general expression can be expanded
in a power series. The isotropic term is of course the pressure term.
54
1. General viscous fluid
! '! 2 0 0$ I2D = tr 2D = 0
( )
!0 '! 0$ # & 1
2D = # '! 0 0& (2D) = # 0 '! 2 0&
2 II2D = I2D ! tr (2D2 )
# & 2
"0 0 0% ## 0 0 0&& IIID = det(2D) = 0
" %
$ "! 2 0 0'
$ 0 "! 0' & )
! = # pI + f1& "! 0 0) + f2 & 0 "! 2 0)
& ) && 0 0 0))
% 0 0 0(
% (
! xx # ! yy = 0
! yy # ! zz = f2 "! 2
Firstly : The invariants of 2D can be easily computed. As it turns out, we do not need
all invariants, in simple shear flow two of the three will be zero.
Secondly: we can reduce the equations even more: the second order term gives rise
to normal stresses that are however not physically observed. No N1 and a positive
N2 think about the rod climbing experiment this is not what we observed. Typically
N1 = xxyy will be positive and N2 = 0.1 N1
55
1. General viscous fluid
! = " pI + f1(II2D ) # 2D
()
! xy = "1 #! 2 $ #!
So we have gone through a lot of trouble to prove that the most sensible and
mathematically and physically correct thing to do is to replace the viscosity by a
scalar function that depends on the invariant of the rate of deformation tensor. For a
simple shear flow, the viscosity will depend on the magnitude of the shear rate.
Several forms have been proposed for this function. We will go through the different
models, distinguishing them by the number of parameters that are used.
56
Viscosity curves :
Styreneethylacrylate in H2O
102 # = 0.50
2a = 250 nm Data from Laun (1984)
101 0.47
100
0.43
(Pas)
101
"
102 0.34
0.28
0.18
103 H2O
! (Pa)
The graph above shows the viscosity versus shear stress for aqueous suspensions
of charged polystyreneethylacrylate spheres with a particle radius of 125nm. The
volume fraction was varied from 0 to 0.50, temperature was kept constant at 25C.
This graph reveals the typical fetaures that suspensions can display : At low volume
fractions the viscosity of the suspensions remains essentially Newtonian and is only
slightly greater than that of the dispersion medium. At higher volume fractions, the
occurrence of shearthinning and shearthickening is very obvious in this figure. At
very high volume fractions and low shear stresses, the Newtonian plateau tends to
disappear, and a yield stress starts to develop.
Not visible in the figure are eventual time effects. Especially for aggregated systems,
gradual timedependent changes in the viscosity can be observed, i;e. thixotropy is a
typical property of such systems.
57
Spherical particles : Zero shear viscosity
[
" = "m 1 + 2.5! + 6.2!2 ]
Up to maximum packing (KriegerDougerthy,1959)
)[ ( ]' max
0 ' + & ( ) (m #
( = (m .1 ) [(] = lim $$ !
. 'max + '*0% '(m !"
/ ,
Einstein calculated the viscosity dissipation produced by the flow around a single sphere. The result is given above. It is only valid for low
concentrations, when the suspensions is dilute enough that the flow field around one particle is not appreciably influenced by the presence of
neighbouring spheres. The equation states that the size does not affect the viscosity. This is true as long as the matrix cn be treated as a
continuum
When two spheres are close enough so that the drag on one particle is affected by the drag on a second one, the spheres are said to interact
hydrodynamically. The velocity field around one sphere is felt by the another one. Such interactions which are far more complicated to
calculate. Interactions between two spheres lead to a term that is proportional in 2, three body interactions lead to a term proportional to 3 and
so on. Following an idea by Arrhenius, we can treat the suspension as an effective medium. The addition of a particle leads to an incremental
increase of the viscosity. Ball and Richmond (1980) suggested that this viscosity increments is proportional to :
Where [] is the intrinsic viscosity, which is the dilute limit of the viscosity increment per unit volume fraction, divided by the particle volume
fraction.
Intergration of this equation leads to the Krieger Dougherty equation (1959), which is an emperical equation for suspensions of spherical and
nonspherical shape. Often the product of []max is constant and approximately equal to 2.
[)])md(
d) =
&1 ' ( #
$ (m !"
%
" # "m
["] = lim
!$0 !"m
58
Spherical particles : Zero shear viscosity
1000
max
Einstein
Batchelor
KriegerDougherty
100
s
"/"
10
This figure compares the predictions for the Einstein, Batchelor and Krieger
Dougherty equation. For the KD equation, the divergence of the viscosity as the
volume fraction approaches maximum packing is observed. For hard spheres,
randomly packed, m ~0.630.64 (which will be relevant for the low shear viscosity).
59
Spherical particles : Zero shear viscosity
60
Viscosity curves :
Styreneethylacrylate in H2O
102 # = 0.50
2a = 250 nm Data from Laun (1984)
101 0.47
100
0.43
(Pas)
101
"
102 0.34
0.28
0.18
103 H2O
! (Pa)
The graph above shows the viscosity versus shear stress for aqueous suspensions
of charged polystyreneethylacrylate spheres with a particle radius of 125nm. The
volume fraction was varied from 0 to 0.50, temperature was kept constant at 25C.
This graph reveals the typical fetaures that suspensions can display : At low volume
fractions the viscosity of the suspensions remains essentially Newtonian and is only
slightly greater than that of the dispersion medium. At higher volume fractions, the
occurrence of shearthinning and shearthickening is very obvious in this figure. At
very high volume fractions and low shear stresses, the Newtonian plateau tends to
disappear, and a yield stress starts to develop.
Not visible in the figure are eventual time effects. Especially for aggregated systems,
gradual timedependent changes in the viscosity can be observed, i;e. thixotropy is a
typical property of such systems.
61
1. Large particles: Hydrodynamic Forces
So far, we have only discussed the effects of particle volume fraction on the increase of the low shear viscosity. The viscosity of suspensions is,
however, strongly dependent on shear rate. The shear rate dependence of the viscosity occurs because shear flow will be able to distort the
equilibrium microstructure. The interparticle spacings will be altered, changing the magnitudes of both the hydrodynamic and the interparticle
forces. This will lead to shear thinning and shear thickening.
The microstructural changes in flowing dispersion can be complex in nature. The mechanism of shearthinning in hard sphere suspensions
has been elucidated by computer simulations. J. Brady and coworkers introduced a technique called Stokesian Dynamics (see e.g. Brady
and Bossis, Annual Review of Fluid Mechanics Vol. 20: 111157 (1988)), where the Stokes equations are solved for all particles. The distortion
of the equilibrium structure changes the interparticle spacing, and the Brownian contribution to the viscosity changes. Particles can be even
observed to lineup into strings parallel to the flow direction. When the shear rate is increased, particles can, however, be brought into close
proximity by the flow leading to the formation of socalled hydroclusters. The deformation of such clusters is difficult as the small separation in
between the particles lead to large lubrication stresses in the gap. For hard sphere systems, we will see that this leads to shearthickening.
When colloidal forces are important, the microstructural reorganizations can have an even more complex dependence on the flow rate.
We will now first address the question if we can predict the transition to shear thinning.
62
Scaling for Brownian hard spheres
! $1 < t d
a2 k T
td = = a 2/ B
D0 6#"m a
" a ?!
Pe = m
k BT
As increases:
replace medium by an effective viscosity
"susp ( #! ) $ #! $ a 3
!r =
kT
Distortion of the particulate microstructure will only occur when the rate at which the
particle equilibrium is recaptured is smaller than the rate at which it is being deformed.
The rate at which the structure regains its configuration is controlled by the particle
diffusivity, D0. For spherical particles the diffusivity is given by the StokesEinstein
relation: kBT
D0 =
6!"ma
The definition of the Pclet number implies that the particles are seeing the viscosity
of the medium. Especially at higher volume fractions, it is better to use the effective,
suspension viscosity as the medium viscosity that a particle feels. Hence a
dimensionless shear stress is more appropriate.
63
(
(r = (susp
med
= f &$% ',)r ,t ,Pe,Re #!"
(r = f &$% ',Pe #!"
10000
0r = f ., 1, / +)*
"r0
"r#
max!
1000
'[0]1max
& 1 #
0r = $1 ' ! r
100
% 1max "
"
10
& 0 ' 0r ( #
0r = 0r ( + $ r0 !
% 1 + b/r " 1
0.0 0.2 0.4 0.6 0.8 1.0
!
& 0 ' 0r ( #
0r = 0r ( + $ r 0 m!
% 1 + (b/r ) "
64
Shear thickening
16
14
12
10
r
!
6
4 Hydrodynamic viscosity
Brownian viscosity
Total viscosity
2
Pe
So far, we have only discussed the effects of particle volume fraction on the increase of the low shear viscosity. The viscosity of suspensions is,
however, strongly dependent on shear rate. The shear rate dependence of the viscosity occurs because shear flow will be able to distort the
equilibrium microstructure. The interparticle spacings will be altered, changing the magnitudes of both the hydrodynamic and the interparticle
forces. This will lead to shear thinning and shear thickening.
The microstructural changes in flowing dispersion can be complex in nature. The mechanism of shearthinning in hard sphere suspensions
has been elucidated by computer simulations. J. Brady and coworkers introduced a technique called Stokesian Dynamics (see e.g. Brady
and Bossis, Annual Review of Fluid Mechanics Vol. 20: 111157 (1988)), where the Stokes equations are solved for all particles. The distortion
of the equilibrium structure changes the interparticle spacing, and the Brownian contribution to the viscosity changes. Particles can be even
observed to lineup into strings parallel to the flow direction. When the shear rate is increased, particles can, however, be brought into close
proximity by the flow leading to the formation of socalled hydroclusters. The deformation of such clusters is difficult as the small separation in
between the particles lead to large lubrication stresses in the gap. For hard sphere systems, we will see that this leads to shearthickening.
When colloidal forces are important, the microstructural reorganizations can have an even more complex dependence on the flow rate.
We will now first address the question if we can predict the transition to shear thinning.
65
NonLinear Viscoelasticity
In the previous two chapter we have gradually added complexity to the description of material behaviour.
Still the cases presented before are only limiting cases of material behaviour. The generalized Newtonian
fluid describes nonlinear but inelastic materials at large deformations. The Linear Viscoelastic fluid
presents a viscoelastic material but it will only be valid for small deformations. Hence two parameters are
important, one being the magnitude of the deformation e.g. strain amplitude, the other being the
Deborah number which is defined as the characteristic relaxation time divided by the characteristic time of
the process.
#
De =
"! !1
???
0 N E
LVE
We now have to start describing a very wide range of nonlinear phenomena. Let us first try to make a round
up of the phenomena we at least want to describe
De
66
1. Nonlinear VE phenomena
2. Additional mathematical tools
3. The secondorder fluid
4. Quasilinear models
5. Nonlinear models
67
1. Nonlinear Phenomena
68
A. Normal stress differences in shear flow
The above figure presents the shear and normal stress differences for a solution of
1.2% polyisobutene in decalin (data from Keentok, et al, JNNFM, 1980). N1 is positive
and is the property responsible for rodclimbing. N2 is typically a factor of ten smaller
and has the opposite sign of N1 . The ratio between N1 and N2 is an important check
for the validity of models as it depends very strongly on how nonlinearity is being
described (biaxiality?).
N1 can be easily measured in rotational devices, e.g. in a cone and plate rheometers
it can be derived from the total axial trust on the cone (see next chapter on
rheometry). The measurement of N2 is much more difficult. It requires measurements
of the pressure gradients along a cone or can be measured using rheooptical
methods. It is rarely being measured in the industrial laboratories.
69
B. Shear thinning of and 1
The above figure (from Laun, Rheol. Acta, 1978 ) shows the steady state viscosity
and the primary normal stress coefficient for our favorite LDPE melt which weve
already encountered previously. This figure shows that the shear thinning in the
viscosity is typically accompanied by a decrease of the primary and secondary normal
stress coefficients for polymer melts and concentrated polymer solutions. We will
provide a physical explanation for this in the chapter on rheology of polymeric liquids
based on the concept of entanglements.
The data in the above figure were obtained by shifting data to a master curve by using
a shift factor. We will come back to the timetemperature superposition principle later
on.
70
B. Shear thinning shows up in timedependence
Shear thinning also shows up in time dependent viscosity after starting up the flow at
a constant shear rate. The above figure shows data from Menezes and Greassley
(1980) on a polybutadiene in an organic oil. The lowest shear rate gives us the
behavior that we already can describe in the framework of linear viscoelasticity. As
the shear rate increases, an overshoot appears, at least for deformations that are
sufficiently large and shear rates rates that are sufficiently high. The presence of an
overshoot will be typical for measurements in the shearthinning regime.
71
C. Relaxation modulus at large strains
Let us see what happens with another LV property, the relaxation modulus, when we
increase the strain amplitude. The above figure gives us the response for the IUPAC
LDPE (data from Laun). At small strains, we find the LVE response which weve
described previously. As the strain is increased, we find a systematic decrease of the
relaxation modulus both the modulus of the shear and the first normal stress
difference.
If you think about this figure it already indicates to us a route to describe nonlinear
viscoelastic behavior. The intrinsic shape of the modulus/time curve remains the
same, it is only shifted to smaller values on the vertical axis. Hence a description of
the time effects (the memory function) and the strain effects (a damping function),
would seem to be an attractive route to model NLVE behavior. A (mathematical)
separation of these effects might lead to tractable equations.
72
D. Extensional Thickening in uniaxial flow
1e+7
1e+6
1
1
0.1 s
0.5 s
1e+5 1
1s
(Pa.s)
1e+4
E
!
fiber windup
RME
1e+3 3*!LV
1e+2
0.01 0.1 1 10 100
t (s)
Up to now only nonlinear phenomena in shear flow have been discussed. One of the
annoying things about nonlinear behavior is that we cannot necessarily extrapolate
the behavior in shear flow to any other type of deformation.
In the chapter on rheology of polymeric fluids, we will discuss the physical origin of
this strain hardening.
73
E. Phenomena in complex flows
shear
stagnation point
The reliability of a constitutive equation needs to be verified for such complex flows,
as a successful prediction of properties in shear and uniaxial elongation do not imply
that the constitutive model works good.
74
2.Tools for NonLinear Models
Need additional mathematical tools:
A. Deformation measures:
C = FT ! F CauchyGreen
" = # A + v! " !A
A Material
#t time der.
We already introduced the Finger tensor in the chapter on finite deformation tensors. Physically, the Finger
tensor gives us the relative local change in area within the sample.
#xi #x j
C !ij 1 = "
#x'k #x'k
Another tensor which describes relative deformations is the Cauchy Green tensor. It relates the local
extension ratios with each other.
"x "xk
Both C and C1 are adequate measure of relative
equations.
Cij = kdeformations
! and are useful to build constitutive
"x' "x' i j
Now we have one more step to take: are the derivatives we know sufficent? Consider the following example
of an expanding duct:
for an external observer the fluid motion is stationary, but for e.g. a polymer molecule in this flow field (an
observer travelling with the stream), it observes a deceleration and hence the time derivative is not zero.
75
2.Tools for NonLinear Models
B.Time derivative : Upper convected
" = # A + v! " !A
A Material
#t time der.
We know from our courses on Transport Phenomena how to solve this: we use the
Material Time Derivative. This improves the description of what the molecule sees.
It is, however, not sufficient. Remember that solid body rotations should not result in
stresses, neither should the derivatives  neglecting inertia effects (which we can do),
it shouldnt matter how fast a molecule is being rotated.
It is also instructive to note that the UC derivative of the Finger tensor (which
describes relative deformations!) is zero, because of the way that they are both
defined. So we will need UC derivatives of D or to introduce nonlinearity.
76
3. The SecondOrder Fluid
$
& = #pI + 2%O D # "1, 0 D+ 4"2, 0 D ! D
the second order fluid is a logical extenion of the Newtonian CE. We already know
that the term in D.D is only capable of predicting a second normal stress difference.
Hence we need a term giving us a first normal stress difference. We will assume a
small perturbation of the behavior at rest. There is a class of constitutive equations
that we can form in this manner : they are called the retarded motion expansion .
The above equation belongs to this class and is called the secondorder fluid. It uses
an additional term containing the upper convective derivative of the rate of
deformation. This term introduces a weakelastic memory into the equation.
It can be shown that under quite general conditions, that a VE fluid will obey the 2nd
order fluid, provided the flow is sufficiently slow and slowly varying, to ensure that
the deviations from Newtonian behaviour.
77
2nd order fluid in steady state shear flow
T
)0 0 0& )0 +" 0 & )0 +" 0 & ) 0 0 0 & ) +" 2 0 0&
! 1' $ ' $ 1' $ ' $ ' $
"
D = D * ' +" 0 0 $ " ' +" 0 0 $ * ' +" 0 0 $ " ' +" 0 0 $ = *' 0 0 0$
2' 2' '0
(0 0 0% (0
$ ' 0 0 $% (0 0 0 $% '( 0 0 0 $% ( 0 0 $%
" = # D + v! " !D = 0
D
#t
Homogeneous
Steady state flow
$ xy $ xx " $ yy $ yy " $ zz
%0 = !1 = = !1, 0 !2 = = !2, 0
#! #! 2 #! 2
We assume a steady state flow the time derivatives are zero. When we are considering a
homogeneous shear flow, the divergence of the stress tensor will be zero as well (there will be no
gradients in the stress this is the definition of a homogeneous flow).
As you can see, the resulting equation is not all that hard to solve. You can also see how the upper
convected derivative works : when material elements are deformed, a stress is produced. And it produces
a term that gives rise to a first normal stress difference.
Note that we defined the coefficients of the secondorder fluid also in a correct manner. In shear flow, the
viscosity is predicted to be constant, and a first and second normal stress difference are predicted, with
constant normal stress coefficients.
Hence the model cannot deal with shearthinning, but predicts normal stresses. Let us consider the
behaviour in extensional flow.
78
2nd order fluid in steady state uniaxial elongational flow
" E = 3"0 +
3
(
#1 + 2#2 , 0 !
2 ,0
)
0 t
0 t
As a first example, solve for a steady state, homogeneous, uniaxial elongational flow. The elongational
viscosity at zero stretching rate equals the Newtonian one, but an extensional hardening is predicted when
2,0 / 1,0 < 0.5 , where E increases linearly with increasing stretch rate. Note that here we see the power
of the CE, once the three coefficients are known from the behaviour in shear flow we can predict the
behaviour in extensional flow.
As the secondorder fluid contains only terms in D and its higher order terms and a UC derivative, it will not
be able to do a good job for predicting time dependent phenomena. For example, for a stress relaxation
experiment the stress will vanish as soon as the flow is stopped. This will be a general result for ANY
retarded motion equation. Above are the predicitions for startup of shear and elongational flow.
This is what we mean by saying that the second order fluid predicts only weak elastic effcts, some
nonlinear effects are predicted, but timedependent characteristics, even in the LV regime are not
described. Most flow of polymeric liquids will not be slow enough for the 2nd order fluid or any of its
colleagues to be valid. So well have to abandon this route for writing good CEs.
Our efforts have, however, not been in vain. Any model that describes a smooth evolution of stress and
structure, should, in the limit of slow flows, yield the second order fluid as a limiting case. So apart from LV
behaviour, we have slow flow behaviour as a good test for any nonlinear models.
79
4. Quasilinear Models
"
! + # ! = 2 $0D
Extending the linear viscoelastic Maxwell model into the nonlinear regime can be done by replacing the
material derivative in the Maxwell equation with an upper convective derivative of the stress. This equation
is called the Upper convected maxwell model. The equation is nonlinear as the UC derivative contains the
product of the velocityb gradient and the stress tensor. For dmall strain amplitudes, the nonlinear terms
vanish and a material derivative is all that remains hence the limit of linear VE is recovered.
Likewise, for very slow flows, the rate is small and Newtonian behaviour is recovered. Analyzing what will
be the first perutbation as thedeformation rates is increased we have to evaluate the upper convected
derivative of the Newtonian equation and :
80
UCM in transient and steady state shear flow
&
( ) # ! " %! # ($v)= 2' D
T
!+% ! + %v # $! " % $v 0
&t
Steady state
Again, start from the steady state, homogeneous shear flow. The horrible looking
equation is easily solved. We find a prediction for a constant viscosity, a first normal
stress difference with a constant first normal stress coefficient and a zero second
normal stress difference.
This CE is not yet describing all the phenomena we want, but at least it looks
promising, as we are finding the right limits for VE and slow flows.
81
UCM in transient and steady state shear flow
.
#! xx
! xx + " = 2 "! xy $
#t t=0, no stresses
#! xy .
! xy + " = %0 $
#t
transient flow, predictions of an UCM model
!0=10, " = 1s
25
(
!( t) = !0 1 # e # t / " ) 20
* # $
t
t ' 15
5
!
#1
0
0 2 4 6 8 10 12 14
t [s]
We are more interested in the transient flows, again writing down the equations in
component form, for a homogeneous flow, you find that the only terms of the stress
tensor differing from zero are given above, and this can be easily solved.
We see that for the shear stress we get the same result as the linear Maxwell model,
BUT! we also predict an evolution of normal stresses in shear flow. The UCM nicely
predicts a gradual increase of the stresses afte inception of flow.
82
UCM in transient elongational flow
2 !0 !0
!E = . + .
1$ 2 " # 1$ " #
10000
1000
e
!
100
10
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
rate (s1)
Using the same parameters as on the previous page, and using the velocity gradient for uniaxial elongation,
the prediction of the elongational viscosity is obtained under steady state conditions as given above. Note
that this is a quite dramatic nonlinear behaviour. The viscosity is predicted to be extremely extensional
thickening, and it rises to infinity when the strech rate approaches 1/2.
In the limit of zero shear rate, the Trouton ratio of three is obtained, followed at slow rates by an evolution
that can be aproximated by a linear dependency on the extensional viscosity on stretch rate, as for the
secondorder fluid.
Typically, constitutive equations will differ more amongst themeselves in elongational flow predictions (this
is a strong flow).
Although the UCM model has some of the qualitative features of the rheology of polymeric fluids, it still fails
to describe some essential aspects, such as shearthiining, N2 and a limited rise of E. The UCM model or
related CEs will do a good job for polymer solutions which are very dilute, such that the polymer molecules
do not overlap. The viscoelastic properties of the solution will be fairly simple and determined by the
properties of the individual molecules. A molecular theory can be constructed that boils down to the UCM
(see part 3).
83
4. Quasilinear Models
( )
t
!= $"#
m( t " t' ) C"1 " I dt'
(
# t# t$ )
'0
(C )
t
#1
!= & "e ( # I dt $
2
#% (
Simplest generalization of the LV integral model
The linear Maxwell equation could be written in integral form. The nonlinear upper
convected Maxwell model can also be written in integral form, provided the proper
deformation measure, i.e. the Finger Tensor, is used. By using the fact that the
upper convective derivative of the Finger tensor is zero (the upper convective
derivative is a time derivative in a special coordinate frame whose base coordinate
vectors are stretching and rotating with material lines), the UCM model can be derived
from the Lodge model.
Although equivalent to the differential UCM equation, some problems will be easier to
understand or deal with with integral CE. The fact that real polymeric fluids posses a
distribution of relaxation times is one of them.
The Lodge model, like the UCM, is inadequate in describing most materials because
its elasticity is that of a simple Hookean solid. We will see how we can get a more
general CE.
84
5. Nonlinear Models
1 #
" + " + fC ( " , D) + fD ( " ) = 2G0 D
!
fC ( ! , D) Stress buildup
fD ( ! ) Stress decay
Starting from the UCM, we can try to obtain an equation that modifies the rate at
which the stress tends to buildup and decay. Two functionals will be introduced,
one depending on the stress and rate of deformation tensor  describing the rate at
which stress buildsup and a second one depedning only on the stress tensor that
can accelerate the rate at which stress decays.
Shearthinning and strain softening can hence be introduced through fc and fd. In
principle, they have a different effect on how nonlinearities are introduced, but when
sufficient relaxation times are present similar nonlinear effects can be introduced
through either fc or fd.
85
A.1. Differential form: Phan Thien Tanner (PTT)
Linear parameters , G0
Nonlinear parameters ,
Several forms have been proposed for fd and fc. Each of the equations has
weaknesses and strengths.
One example is the PTT model. The 1 parameter model describes shear thinning,
Nonlinear time effects and extensional thickening, without divergence. But some
weird oscillations are obtained during startup of flow.
The multimode model works much better and can be used to describe a fairly wide
range of polymer melts.
86
A. 2. Giesekus model
fC ( ! , D) = 0
#
fD ( ! ) = ! "!
G0
Linear parameters , G0
Nonlinear parameter
The Giesekus model is attractive as it contains only one parameter, and still does a
fairly good job. Note that the linear parameters can be obtained from Linear
viscoelastic measurements, the nonlinear parameter must be obtained from a non
linear experiment.
87
B. Really nonlinear integral models
$ [m ( t " t' ,I ]
t
!= 1 C "1
, IIC"1 ) C"1 + m2 ( t " t' , IC"1 , IIC"1 ) C dt'
"#
or assume m2 = m1.
"U "U
m1 = 2 , m2 = 2
"IC!1 "IIC!1
88
The above data refers to the LDPE by Laun. The damping function is obtained
experimentally, the approximation with one (dashed line) and two exponentials (sum,
line) is shown. For the single exponential version, only the exponent needs n to be
determined as nonlinear parameter. The Wagner model then can be used to predict
the nonlinear behaviour, and does a good job for example during startup of a shear
flow, for both shear and normal stresses. Calculating this evolution requires solution
of an integral equation.
89
B. Really nonlinear integral models
( ) (
m1 t ! t' , IC!1 , IIC!1 = m( t ! t' )h IC!1 , IIC!1 )
h is called the damping function
Wagner model
( ) (
h = f1 exp !n1 Ic !1 ! 3 ! f2 exp !n2 Ic !1 ! 3 )
h( " ) = exp( !n" )
When the straintime separability can be applied, the only nonlinear material
function needed to predict the behaviour is the damping function. It will, in general
depend on the invariants of the Finger tensor. In some case, a simple exponential or
for higher strains a sum of two exponentials is sufficient. This model was derived by
Manfred Wagner (1976).
90
Conclusions
yield
Gen. Newt
NLVE
N 2nd E
0 OF
Integral
Differential
..
LVE
#
De De =
"! !1
91
Conclusions of part 1:
Material functions that we can use to describe viscoelastic behavior and Non
Newtonian behavior have been rigorously introduced.
92
Outline
1. Introduction
6. Extensional rheometry
93
Introduction
94
Introduction: classification
type of straining:
95
Shear flow geometries
Sliding plates
Capillary
Couette
Slit
Cone and Plate
Annulus
Parallel plates
96
Requirements for shear rheometers:
97
1. Drag flows : Sliding plate rheometer
Fy v0
!=
h
Fx
Fx
h " 12 =
LW
L " 22 =?
98
Drag flows : Couette geometry
M. Couette (1890)
99
Drag flows : Couette geometry
L Equations of motion:
1 !(r" rr ) " $$ v2
r: % = %# $
r !r r r
$:
(
! r 2 " r$ )= 0
!r
% !p
z: + #g = 0
!z
100
Drag flows : Couette geometry
1. v=Ri at Ri 2Ri
2. v=0 at Ro
Shear stress
L
!:
( 2
" r # r! )= 0 % # r!
C
= 2)
'
"r r )) M
( # r! =
) 2$ & L & r 2
M )
# r! (Ro) & (2$ & Ro & L) = )*
Ro
Normal stress :
 causes rodclimbing
 stress difference @ Ri and Ro : hard to measure
101
Drag flows : Couette geometry
" #R Ri
!= if > 0.99
Ro $ Ri Ro
Ro + Ri
R =
2
if Ri/Ro <1: shear stress is not constant over the gap.
shear rate distribution
"#
!= r
"r
d#
! = 2$
d$
# (Ri ) $(Ri ) !
#(Ri ) = % d# = % d$
0 $(Ro ) 2$
d#
2$(Ri ) = ! ($(Ri )) & ! ($(Ro))
d$(Ri )
102
Drag flows : Couette geometry
103
Drag flows : Cone and plate
Equations of motion:
r:
( 2
1 ! r " rr )
&
" $$ + " %%
= & #
v $2
r2 !r r r
104
Drag flows : Cone and plate geometry
Boundary conditions
1. v(/2) = 0
2. v(/2)=rsin(/2)r
Shear stress
!:
( )
1 d " #! 2
+ cot ! & " #! = 0 ' " #! =
C *
( Cte ,
r d# r 2
sin # ,
, 3M
+ " #% =
, 2$ & r 3
2$ R
M = ) ) r 2 " #! drd! ,
0 0 ,
105
Drag flows : Cone and plate geometry
"R 2
Fz =
2
(
# $$ ! # %% )
2Fz
N1 =
"R 2
106
Drag flows : Cone and plate geometry
107
Drag flows : Parallel plates
Equations of motion:
1 !(r" rr ) " $$ v2
r: % = %# $
r !r r r
!(" $z )
$: =0
!z
!(" zz )
z: =0
!z
108
Drag flows : Parallel plates
vr " #r
!= =
h h
Normal stresses
Fz " d ln Fz %
N1 ! N2 = 2 $
2+ '
(R # d ln )! R &
109
Drag flows : Parallel plate geometry
110
Rotational rheometers : practical designs
Transducer
Torque control
Servo Controlled
Geometries Motor
environments
111
Pressure driven flows : capillary rheometry
Hagen (1839)
Poisseuille (1840)
Nonhomogeneous flow
typically for high shear rates
simpler to operate
yet yield precise data!
112
Pressure driven flows : capillary rheometry
Q
R 1. Steady, laminar, isothermal flow
2. No slip at the wall, vx =0 at R=0
3. vr=v=0
4. Fluid is incompressible, f(p)
! r
Equation of motion:
Z
1 !(r" rz ) !p
L z: # =0
r !r !z
Only a function of r
Only a function of z
1 d (r! rz ) dp
=
r dr dz
113
1 d (r! rz ) Po " PL
=
r dr L
Po " PL r C2
! rz = # +
L 2 r
C2 = 0 because ! rz $ % @ r =0
Po " PL R
! rz (R ) = ! w = #
L 2
r
! rz (r ) = ! w
R Note : independent of fluid properties
114
R
Q = 2) ! v z (r )rdr Shear rate calculation from Q
0
R
R dv z
Q = 2 )v z r 0
" 2) ! r 2 dr Integration by parts + assume :no slip
dr
0
R R substitute r and dr
r= * and dr = d*
*w *w
*w # 2
R
dv z R & dv z R
Q = "2 ) ! r 2 dr = "2 ) ! % *( d* Typical trick to deal with inhomogeneous flows:
0
dr
0 $ * w '
dr *w changing of variables
Q* w 3 *w 3Q! w 2 !w 3 dQ
= " ! (* ) = % ! w 2 dv z
2 dv z
d* + $
)R 3
0
dr "R 3
"R 3 d! w dr !w
#! a & d ln Q )
#! w = (3 + +
4 ' d ln ! w *
115
!a # d ln Q &
WeissenbergRabinowitsch Correction !w = %3 + (
4 $ d ln " w '
Physical meaning: 2
n=1
n=0.7
n=0.3
>
!a "
z
1% 1
!w = $3 + '
/ v<v
z
4 # n&
Steepness of the velocity profile changes
Shear rate is increased with respect to 0
0 1
the Newtonian case: r/R
4Q
!a =
"R 3
116
Rabinowitsch Weissenberg correction : be careful  requires numerical differentiation
1000
Raw data
"corrected data"
100
[Pa.s]
a
!, !
10
typical data of a PP sample @ 230C
117
Are these conditions met ? 1. Steady, laminar, isothermal flow
2. No slip at the wall, vx =0 at R=0
3. vr=v=0
4. Fluid is incompressible, f(p)
Problem 1: Melt distortion
118
Are these conditions met ? 1. Steady, laminar, isothermal flow
2. No slip at the wall, vx =0 at R=0
3. vr=v=0
4. Fluid is incompressible, f(p)
Problem 2: Viscous heating
.
!"R2 #
Na = 4k
119
Are these conditions met ? 1. Steady, laminar, isothermal flow
2. No slip at the wall, vx =0 at R=0
3. vr=v=0
4. Fluid is incompressible, f(p)
Problem 3: Wall slip
4vs
"! a = "! + @ ! = cst
R
120
Extensional Rheometry
Extensional viscosities can provide an
indication of chain structure.
3
thinning: rigid systems
thickening: branched melts
3
thinning: linear melts
Most processing flows are complex superpositions of flow types and many contain substantial
amounts of extensional character.
The extensional viscosity is qualitatively different than the shear viscosity and complete
characterization should include its measurement.
Extensional flows are extremely difficult to realize in the laboratory. The available techniques
can be divided into methods for melts and methods for solutions.
121
Extensional Viscosity of Melts
Extensional flows for melts can be generated more easily than
solutions because these systems are sufficiently viscous to
allow them to be mechanically stretched. Below is the Meisnner
device (Rheol. Acta 33, 121, 1994), other examples include the Mundstedt
device.
sample
frit
air support
The major advantadges of the Meissner device are that effects of gravitation are
nicely balanced by the air flow and that the deformation field is homogeneous.
Nevertheless, sample preparation is not at all easy, slippage at the belts can occur
and the range of deformation rates and forces is typically limited.
122
Macoskos Windup Technique
J. Rheol. 40, 473481 (1996)
!R D
"! = ,R D windup drum radius,
L0
L 0 distance between clamp and drum
M exp("! t )
#11 $ # 22 = ,M torque,
!R D R 02
R 0 initial radius of the sample
This technique goes back to the 60s but was reintroduced and improved by Macosko.
The deformation field is however not perfectly homogenous (one end is clamped in 
think about this) and sagging under the effect of gravitation cannot be avoided.
123
LDPE (ATOFINA) @ 190C
1e+7
0.1 s1
0.5 s1
1 s1
1e+6
3*!
[Pa.s]
E
!
1e+5
3*!afschuiving
1e+4
0.1 1 10 100
t [s]
These are typical data taken on a commercial LDPE sample in our lab using the FWU
technique. The pronounced strain hardening that can be observed makes this a
suitable material for e.g. film blowing.
124
Capillary Breakup methods
125
Current Research Topics
 Microrheological methods
126
Current Research Topics
127
 Linking rheology and structure:
Rheooptics, Rheoscattering,...
128
 Rheology of complex fluid interfaces
129
Aim of this lecture:
130