Module 1: Viscoelastic behavior of polymers

Table of Contents

Foreword

1. Introduction
1.1: Recall of material properties
1.2: Recalls of concepts and definitions of fluid and solid mechanics

2. Mechanical behavior in the fluid state

2.1 Rheology of polymers
2.2 Effect of internal and external variables
2.3 Examples of flow in a duct

3. Mechanical behavior in the solid state

3.1 Viscoelasticity of polymers
3.2 Effect of temperature: the time-temperature equivalence postulate
3.3 Effect of internal and external variables

4. Mechanical spectroscopy

References
CHAPTER 1

1.1 WHAT CAN BE CALLED A “PROPERTY” OF THE MATERIAL?

The traditional concept of material properties is based on the identification of a "value" or

"quantity" which is represented by a number with an associated unit of measurement. Most
material information available in literature or technical datasheets is coded in this way (Figure
1.1).

In addition to the name, value and units of measurement of the property, there is always a
reference to a document specifying the test method, which may be an internal document of the
material manufacturer or - increasingly - a standard issued by a national or international
standardization body, such as ASTM, EN or ISO; this document contains all the details required to
understand the meaning of the property and the test conditions under which the property values
reported in the bulletin were measured in the laboratory. It is therefore essential for
understanding the meaning of the numeric value reported.

In the example of Figure 1.1 most of the measurements refer to ISO (International Standards
Organization) standards or, for the measurement of combustion or electrical properties, to IEC
(International Electrotechnical Commission) standards. It can also be noted that not all properties
are characterized by a numerical value: for example, the burning behavior is characterized by a
"class" - e.g. V-2 - which is defined by the corresponding IEC 60695-11-10 standard.

Figure 1.1

2
Finally, since the material of the example (polyamide 6) is subject to moisture absorption, most of
the property values were measured in both "dry" and "conditioned" conditions for comparison
(the conditioning procedure for water absorption is also specified by a relevant standard, not
reported here).

Although this type of information can be very useful in the preliminary phase of identifying a
shortlist of candidate materials for a given application (material selection), it is important to note
that not all values are adequate for the subsequent design phase of a plastic component. Because
they are single numerical values, they are usually obtained by sampling during a continuously
varying material response. An example which is very familiar to materials engineers is shown in
Figure 1.2 for a loading curve measured in uniaxial tension.

Figure 1.2

The initial slope of the curve allows for the calculation of the Young's modulus; when the linear
trend is no longer obeyed, it is necessary to recognize that phenomena other than elastic
deformation may come into play, such as non-linear elasticity, viscoelasticity or damage. All of
these factors are very hastily - and often erroneously - attributed to the onset of plastic
deformations in the material which subsequently result in a localized collapse of the specimen
section, called necking, which produces a relative maximum in the curve usually identified as the
“yield stress”.

After the subsequent stages of cold drawing and strain hardening, the curve usually passes
through another maximum – this time an absolute maximum – which is classified as "ultimate
strength". The last stress value recorded before the final failure of the specimen is finally called
the "fracture point".

It is therefore clear that this quantification of the load curve "by points" can only provide a very
limited description of the actual material behavior. This is also the reason why some “simplified”
definitions of material behavior have come into use and are also shown in Figure 1.2.

3
In an attempt to overcome the limitation of traditional datasheets, the technical community
introduced a new way of characterizing the behavior of polymeric materials based on the use of
'functions' instead of single values. This innovation was sanctioned by
the publication - in 1999 - of the ISO 11403 standard entitled 'Plastics - Acquisition and
presentation of comparable multipoint data', which provided scientists and business researchers
with a new methodology for representing materials data.

The two examples of rheological (a) and isochronous stress-strain curves (b) taken respectively
from part 2 and part 3 of the above-mentioned standard provide a good representation of the
potential given by describing the "trend" of the property when certain characteristic variables of
the physical phenomenon under examination vary (Figure 1.3).

Figure 1.3

Example (a) shows the effect of the shear rate - shown on the horizontal axis as an independent
variable - on the shear viscosity of a polymer. Example (b) shows the effect of time on the shape of
stress-strain curves obtained from creep experiments. Although the meaning of these curves can
only be fully understood after examining the mechanical behavior of polymers in the following
Chapters, it is immediately apparent that this type of data contains very valuable information for
both the design of the conversion of the material into a product - which implies the melting and
flow of the material - and of the behavior of the product under 'static' loading conditions.

The limitation of using "single" data is not only due to the limited information content but often
also to the fact that forcing complex behavior into simple information often implies that the
physical meaning of the measurement is lost. In the example shown in Figure 1.4, the
experimental setup called “heat deflection temperature" (HDT) is used to characterize the
"thermal resistance" of a solid material; this is done by recording the temperature at which the
deflection under load of a bar reaches the given value of 0,25 mm.

4
 Figure 1.4

The experiments return a ranking of the thermal resistance of different polymeric materials that
does depend on the applied load (Table 1).

Table 1: ranking of HDT for different materials under two different loading conditions, labeled Method A and
Method B respectively

Ranking HDT Method A HDT Method B

1 PA6 + 30% glass fiber PA6 + 30% glass fiber
2 PBT + 25% glass fiber PBT + 25% glass fiber
3 PC PA66
4 PA66 PA6
5 ABS PBT
6 PA6 PC
7 PBT ABS

Furthermore, the fact that the measurement is conducted under non-isothermal conditions makes
the value of the measurement itself very questionable, as it will certainly be influenced by the
temperature gradient within the material, which in turn is controlled by the thermal conductivity
coefficients. Clearly, this method has been developed in order to be able to make rapid
comparisons between materials under given conditions, but it lends itself poorly to engineering
use. This is one of the many examples in which the material property – the thermal “resistance” –
is just “operationally defined” by the test used to measure it, but is not reflected in a general
scientific framework.

A question then arises: why have these test methods been developed if they are not based on a
sound scientific approach? The answer lies in the need to develop 'indices' for comparing
materials that can simplify testing procedures and reduce characterization times, especially during
material development or process quality control. Certainly, as long as these data are kept within
the boundary for which they have been generated, there is no ambiguity; however, any attempt to

5
extend their validity beyond this boundary, for example by using them for the engineering design,
carries with it the risk of mis-estimating the actual characteristics of the materials.

Another particular aspect that must always be carefully considered is the possibility that the
values obtained for a given property are dependent on the geometry of the specimens with which
they are measured. The effect of geometry on the behavior of structures - and a laboratory
specimen is in fact a miniature structure - is well known to engineers: Figure 1.5 shows as an
example the ranking the flexural stiffness of beams having a constant cross-section but different
cross-sectional shapes.

Figure 1.5

This problem has found a mathematically exact - and therefore very elegant - solution in the
framework of Structural Mechanics, which very clearly separates the geometry of the body and
the external boundary conditions (constraints and loads) from the material property that controls
the elastic behavior of the structure (i.e. the Young's modulus, E). An example of the analysis for a
simply supported beam under constant bending moment is shown in Slide 1.8.

This is not, however, always strictly recognized in standards and industrial practice. A striking
example is the impact strength measured according to the IZOD or Charpy methodologies (see
Slides): although the fracture energy is correctly ‘normalized‛ to the fracture surface area, thus
obtaining a ‘specific‛ fracture energy (expressed in J/m2), the resulting values change significantly
with specimen geometry (namely notch tip radius and thickness). This clearly prevents not only
the use of such data for engineering design, but also to attain the less ambitious target to
unambiguously classify different materials according to a specific performance, since the latter is
not truly representative of the real behavior of the materials.

6
In the case of polymers, moreover, there are some additional complications to the general picture
outlined above: the properties are for example highly dependent on the microstructure of the
material (which in turn depends in a combined way on the chemical structure and the processing
conditions for obtaining the product), on the effect of environmental conditions (temperature,
solar radiation, humidity, etc.) and on the time of observation (e.g., due to the fact that their
mechanical behavior is viscoelastic). As a consequence, such PROPERTY DATA normally reported in
the technical bulletins or commercial datasheets provided by the polymer producers are generally
of little ‘predictive‛ value, and they cannot in general be used to predict the material behavior
under real end-use conditions. Put in simple terms, they lack “transferability” from lab conditions
to service conditions.

It is therefore required, before entering into a detailed examination of the physical-mechanical

behavior of polymers, to clarify in advance what the minimum conditions are for the behavior of
materials to be described appropriately, so that the values of their properties can be reliably and
accurately used for engineering design of the industrial components.

DEFINITION OF INTRINSIC PROPERTY

In general, it is not possible to say that a material possesses a certain property "by itself": for such
property to be identified, in fact, the material is to be subjected to an external stimulus and the
corresponding “reaction” should be recorded. Indeed, it is only by measuring the material's
response to the aforementioned stimulus that it is possible to define the property under
consideration (see Slides).

The stimulus can be any kind of chemical-physical entity such as a mechanical force, an electric
voltage, a temperature gradient, etc. The response can be of the same nature as the stimulus (for
example a deformation when a force is applied) or of a different kind (as in the case of
piezoelectric materials which generate an electrical polarization upon deformation).

It is important to emphasize that the material response that is measured as a result of the
application of the stimulus is not a "property" per se: for example, the color of an object is not a
property of the object itself, since it depends on the wavelength of the incident light. Rather, it is a
relationship between the stimulus and the response: therefore, the property turns out to be a
quantity that is defined through an equation or a function ...!

This function is usually multi-variable since it is known, for example, that the response of a
material to a mechanical stress depends on the type of stress applied (force, strain, etc.) but also
on the geometry of the body and the boundary conditions (beam, plate, bending, torsion, etc.) as
well as on the environmental conditions, observation time and structural state of the material.
These categories are in turn a set of specific variables and parameters: environment is defined, for
example, by temperature, humidity, pressure, etc., while for the structural state of the material
one may speak about morphology, orientation of the molecules, degree of crystallinity, degree of
cross-linking, free volume fraction, etc.

This complexity cannot be reduced or eliminated, but it’s possible to schematize the approach to
the problem by trying to separate the variables which are intrinsic to the behavior of the material

7
(for example, it is certainly true that the material behavior should change according to the degree
of crystallinity or the temperature) from those who are not. Indeed, in most cases there is the
possibility of "normalizing" the stimulus and response with respect to the "geometry" of the
system, thus making sure that the measured property is always the same regardless of the shape
and boundary conditions of our experiment. Referring again to a mechanical experiment, such
normalization is done by calculating the stress (in fact a force normalized with respect to the
geometry, obtained, for example, by dividing the applied force by the cross-sectional area of the
specimen in the case of uniaxial tensile test) and the strain (obtained, for the same example, by
dividing the elongation of the specimen by its initial length).

When this is done, the function defining the property - and therefore the property itself - becomes
independent of the geometry, thus characterizing the behavior of the material “per se” regardless
of the shape, size and boundary conditions of the physical system on which the same property is
measured. A new function is thus obtained, which is called the “constitutive equation” of the
material.

The advantage of this approach is remarkable: attempting to define a property as a simple

relationship between stimulus and response inevitably leads to a result that lacks generality, since
it depends on the boundary conditions of the experiment (see the example of the HDT test given
in Tab. 1). If, on the other hand, normalization of the variables is operated preliminarily, then this
relationship - and consequently the property - will turn out to be independent of the geometry of
the experiment: we will call this relationship an “intrinsic property” of the material (namely
because it depends only on the material and not on the body or specimen used in the
experiment).

The example of measuring the mechanical behavior of an elastic material at small deformations
can be useful to better clarify this difference. Assuming that the behavior of the material is linear
elastic, the simple ratio of the change in the size of the specimen (elongation or deflection
depending on the type of experiment) to the force that originated it will turn out to be a constant
called “compliance” (of the specimen). The fact that this value is constant at varying the applied
force is a direct consequence of the linear elastic behavior of the material, but it does not mean
that it is an intrinsic material property: in fact, if the cross-section of the specimen changes, the
value of the compliance varies accordingly.

Many other examples of the advantage of referring to intrinsic properties instead of empirically
defined quantities can be found in many other areas of science and technology. The viscosity of a
fluid, for example, can be used to calculate the flow rate of a fluid exiting a pressurized pipeline
regardless of the shape and size of the pipe section. Another example is fracture toughness, which
is measured according to fracture mechanics theory on notched specimens and depends neither
on the shape of the specimen nor on the depth of the notch; this property is therefore considered
an intrinsic property of the material, in contrast to other types of measurement (such as tensile
failure stress or IZOD or Charpy impact). Other examples can also be found for other categories of
material properties such as magnetic (susceptibility), electrical (dielectric constant), optical
(refractive index), thermal (conductivity, heat capacity, coefficient of thermal expansion) and
physical (density).

In conclusion, the distinction made above between a generic property and an intrinsic material
property sets up an important hierarchy for the purposes of engineering design: in fact, only

8
intrinsic properties can be used with confidence and reliability for purposes that are not simply a
generic comparison between different materials but rather an engineering use for designing
components. Accordingly, in the remainder of this course, only those intrinsic properties that are
of interest for characterizing the mechanical behavior of polymeric materials - both in the fluid
and solid state - will be examined.

2. RECALL OF BASIC CONCEPTS AND DEFINITIONS IN SOLID AND FLUID MECHANICS

Before proceeding with the analysis of the course topics, in light of what was reviewed in the
previous section, it is important to briefly recall the main definitions of quantities and properties
used in the analysis of the mechanical behavior of materials based on the theory of continuum
mechanics. This theory, as its name suggests, is based on the assumption that material subjected
to mechanical action maintains its continuity and thus the effect of applying a stress at one point
propagates instantaneously to all points of the body. Which is certainly true in practice if the
stresses – or the strains – are kept low (infinitesimal).

Under these conditions, it is possible to conceive and mathematically define the stress acting at
each point of the body (see slide 2.1) by imagining to cut the body with a plane through P so as to
show off the internal forces acting on this imaginary surface (red arrows) and in particular on the
small area ∆A surrounding P. The stress acting at point P on the plane section through P is so
defined:
𝐹"
𝜎" = lim
∆"→$ ∆𝐴

Both the stress, 𝜎", and the applied force, 𝐹" , are vectors, and they will depend on the orientation
of the plane section. The overall state of stress at point P in the body is defined as the totality of
vectors acting at that point (corresponding to the infinitude of cut planes we can make through
point P): that set of vectors is called a “tensor”, 𝜎+.

Such a representation requiring an infinite set of vectors would be of course impractical: however,
Cauchy was able to demonstrate that, if the stress vectors on three mutually perpendicular plane
sections (e.g. the stress vectors 𝜎"% 𝜎"& , 𝜎"' on the three planes normal to the coordinate axes, 1, 2
and 3 in slide 2.2) are known, then the stress vector acting on any other plane (e.g. acting
on the plane having normal n different from 1, 2 and 3) can be derived from the previous ones [via
a relatively simple trigonometric function based on the equilibrium conditions].

Therefore, in order to fully characterize the stress state at any point P of the body, only the three
following stress vectors are required:

𝜎"%
𝜎+ = ,𝜎"& -
𝜎"'

For practical reasons, any stress vector can easily be reduced to scalar quantities by simply
decomposing it along each one of the reference axis, as shown in slide 2.3. Each projection can in
this case be easily identified by multiplying the modulus of the vector (a scalar quantity) by the

9
versor (a unit vector) oriented in the direction of the relevant reference axis. A set of nine scalar
components is therefore obtained, which can be expressed in matrix form as follows:

𝜎"% 𝜎%% 𝜎%& 𝜎%'

𝜎+ = ,𝜎"& - = .𝜎&% 𝜎&& 𝜎&' /
𝜎"' 𝜎'% 𝜎'& 𝜎''

in which the first index identifies the normal to the plane on which each stress vector is applied
and the second index identifies the direction of the projection. It therefore follows that the three
components along the main diagonal - which have the two equal indices - refer to “normal”
stresses (i.e. acting in the direction normal to the plane to which they refer) while the components
outside the diagonal refer to so-called "shear" stresses, i.e., acting in the reference plane.

Furthermore, using equilibrium it can be shown that 𝜎() = 𝜎)( for i ≠ j (i.e. the matrix is symmetrical)
leaving only six independent components:

Three normal stresses: 𝜎%% ; 𝜎&& ; 𝜎''

Three shear stresses: 𝜎%& (= 𝜎&% ) ; 𝜎%' (= 𝜎'% ) ; 𝜎&' (= 𝜎'& )

Based on this representation, different combinations of stresses have been developed to better
characterize the stress state at point P. The main functions are listed below:

Pressure, p

It’s a scalar value obtained by taking the average (changed by sign) of the normal applied stresses
at the point P:

%
p = - ' (𝜎%% + 𝜎&& + 𝜎'' )

Deviatoric stress

It’s the tensor calculated by taking the difference between the stress tensor 𝜎+ and the hydrostatic
pressure tensor 𝑝̿ = 𝑝 𝛿̿:

𝜏̿ = 𝜎+ − 𝑝 𝛿̿

in which 𝛿̿ is the unit tensor. It accounts for the stress components that will induce only a change
in shape of the material, which is an important distinction in the definition of failure criteria.

Stress invariants

These are scalar values made of special combinations of stress components that are invariant with
changing the reference axes.

I1 = 𝜎%% + 𝜎&& + 𝜎''

& & &
I2 = 𝜎%% 𝜎&& + 𝜎&& 𝜎'' + 𝜎%% 𝜎'' - 𝜎%& - 𝜎&' - 𝜎'%

10
 & & &
I3 = 𝜎%% 𝜎&& 𝜎'' + 2 𝜎%& 𝜎&' 𝜎%' - 𝜎%% 𝜎&' - 𝜎&& 𝜎'% - 𝜎'' 𝜎%&

in which the index of the invariant identifies the degree of the function (first, second and third).
Similar invariants can also be defined for the deviatoric (shear) stresses:

J1 = 𝜏%& + 𝜏&' + 𝜏%'

% %
J2 = + [(𝜎%% − 𝜎&& )& + (𝜎&& − 𝜎'' )& + (𝜎%% − 𝜎'' )& ] + 𝜏 %&
&
+ 𝜏 &&' + 𝜏 &'% = ' 𝐼%& − 𝐼&
& %
J3 = det?𝜏,- @ = &. 𝐼%' − ' 𝐼% 𝐼& + 𝐼'

Based on the previous analysis it has been concluded that, for any arbitrary reference system, the
stress state at any point P is fully characterized by three normal stresses and three shear stresses.
In simplified engineering notation, those two categories of stresses are usually distinguished by
the use of two different symbols:

Normal stresses: 𝜎% ; 𝜎& ; 𝜎'

Shear stresses: 𝜏%& ; 𝜏%' ; 𝜏&'

in which the double index is dropped for the normal stresses, and the symbol 𝜏 is used instead of
𝜎 for the shear stresses.

It is possible to demonstrate that do always exist three particular, mutually perpendicular

directions I, II and III (slide 2.5) such that on the respective normal planes only normal stress
components exists 𝜎/ , 𝜎// , 𝜎/// while all shear stress components values are zero (𝜏%& = 0, 𝜏%' = 0,
𝜏&' = 0). The three directions I, II and III are called the “principal” directions and the three normal
stress components 𝜎/ , 𝜎// , 𝜎/// are called the “principal” stresses.

It is interesting to note that, when written in terms of the principal stresses, the stress invariants
expressions simplify as follows:

I1 = 𝜎/ + 𝜎// + 𝜎///
I2 = 𝜎/ 𝜎// + 𝜎// 𝜎/// + 𝜎/ 𝜎///
I3 = 𝜎/ 𝜎// 𝜎///

J1 = 0
%
J2 = + [(𝜎/ − 𝜎// )& + (𝜎// − 𝜎/// )& + (𝜎/ − 𝜎/// )& ]

since all shear stresses are zero.

There is always one principal stress which coincides with the maximum normal stress acting on the
material, which is the basis for the criterion of the maximum principal stress theory of failure. In
case of a plane stress condition (which is typical of beams and shells) the values of the principal
stresses and their directions can be found analytically using the following formulae:

(𝜎% + 𝜎& ) (𝜎% − 𝜎& )&

𝜎012 = 𝜎/ = + B &
+ 𝜏%&
2 2

11
 (𝜎% + 𝜎& ) (𝜎% − 𝜎& )&
𝜎0,3 = 𝜎// = − B &
+ 𝜏%&
2 2

or graphically using the Mohr’s circle.

When dealing with yielding of materials – since hydrostatic stress alone does not cause yielding –
we can find a material plane called the octahedral plane (slide 2.6) where the stress state can be
easily decoupled into dilation strain energy and distortion strain energy. On the octahedral plane,
the octahedral normal stress solely contributes to the dilation strain energy; this effect is
summarized into a scalar quantity, called “hydrostatic” stress 𝜎4 , obtained by calculating the
average of the principal stresses:

%
𝜎4 = 𝜎0 = '
(𝜎/ + 𝜎// + 𝜎/// )

which is also sometimes referred to as the mean stress, 𝜎0 . Such a value should not be confused
with a quantity with the same name used in alternating (e.g., fatigue) stresses to represent the
average value of the stress acting on the material.

The remaining dilation strain energy in the state of stress is determined by the octahedral shear
stress which is obtained by calculating the square root of the sum of the differences squared
between the principal stresses:

1 2
𝜏567 = E(𝜎/ − 𝜎// )& + (𝜎// − 𝜎/// )& + (𝜎/// − 𝜎/ )& = B 𝐽&
3 3

and represents the tangential component of stress across the faces of a regular octahedron whose
vertices lie on the principal axes of stress. This quantity is used for example as a measure of the
strength of the material in the von Mises octahedral shear stress failure criterion.

The final meaning of the J2 invariant is therefore the energy of distortion, e.g. the energy per unit
volume required to change the shape but not the volume of the material. Strictly speaking, the
energy of distortion is not an invariant of the stress tensor because its value is dependent on
material properties. Nevertheless, it has the same form for all linear, isotropic materials and its
value is directly proportional to J2. Hence, for a given material, if one is known, so is the other. The
energy of distortion, Ud, can be calculated based on the value of J2 using the following formula:

1
𝑈8 = 𝐽
2𝐺 &

The strain – which can be seen as the consequence of applying a stress state to a material – can be
represented in very much the same way as the stress. The corresponding set of nine scalar strain
components is as follows:

𝜀̅% 𝜀%% 𝜀%& 𝜀%'

𝜀̿ = ,𝜀̅& - = 𝜀
. &% 𝜀&& 𝜀&' /
𝜀̅' 𝜀'% 𝜀'& 𝜀''
12
of which, for the symmetry condition of the tensor, only six components are independent, namely:

Three side extensions or dilatations: 𝜀%% ; 𝜀&& ; 𝜀''

Three variations of dihedral angles or distortions: 𝜀%& (= 𝜀&% ) ; 𝜀%' (= 𝜀'% ) ; 𝜀&' (= 𝜀'& )

as shown in slide 2.7. In simplified engineering notation, those two categories of strains are usually
distinguished by the use of two different symbols:

Dilatations: 𝜀% ; 𝜀& ; 𝜀'

Distortions: 𝛾%& ; 𝛾%' ; 𝛾&'

in which the double index is dropped for dilatations, and the symbol 𝛾 is used instead of 𝜀 for the
distortions. The total volume variation, ∆, due to deformations is simply obtained by calculating
the trace of 𝜀̿ (e.g. by summing up all dilatations along the main diagonal) as follows:

∆ = 𝜀% + 𝜀& + 𝜀'

Based on the previous definitions of stresses and strains, a set of simple mechanical experiments
can be designed (Fig. 2.7) in order to measure the property of the material that controls the
relationship between them. By assuming, for the sake of simplicity, that the material is an
isotropic solid having a linear elastic behavior one obtains:

Uniaxial tensile test

9!!
1. Tensile modulus (Young’s modulus) 𝐸= :!!

:!!
2. Tensile compliance 𝐷= 9!!

; :"" ; :##
3. Lateral contraction ratio (Poisson’s ratio) 𝜈= :!!
= :!!

Simple shear test

<!#
4. Shear modulus 𝐺= =!#

=!#
5. Shear compliance 𝐽= <!#

Hydrostatic compression test

9$
6. Bulk modulus 𝐾= ∆

∆
7. Bulk compliance (compressibility) 𝐵= 9$

Therefore, depending upon the type of stress being applied and the resulting strain, the modulus of
elasticity can basically be classified into the following three types: Young’s modulus (E) shear

13
modulus (G) and bulk modulus (K). These three values are not independent from each other:
typically, for engineers the bulk modulus and the shear modulus are considered to be functions of
the Young’s modulus and the Poisson’s ratio. This comes from the fact that both of them can be
measured in one single, relatively simple (and cheap) type of experiment: the uniaxial tension. The
relations are:
𝐸
𝐺=
2 (1 + 𝜈)

𝐸
𝐾=
3 (1 − 2𝜈)

It is important to note that bulk and shear are not interrelated, since they are independent,
fundamental mechanisms of deformation which refer to size and shape changes of the body
respectively.

Another important example comes from the flow of a liquid material; in this case, the two
fundamental types of flow are simple shear and elongational (Fig. 2.8). The first type of flow can
be obtained by applying either a shear stress – like in the case of the interaction between the hull
of a sailing boat and the water of the ocean – or a shear strain – which arises, for example, during
the flow of a liquid into a pressurized pipe due to the fiction with the inner wall of the pipe itself. It
turns out that the simple shear flow is controlled by the shear rate, 𝛾̇ , which – if the viscosity of
the fluid is known – can be calculated based on the value of the shear stress, 𝜏, as follows:
<
8. (Newtonian) shear viscosity 𝜂= =̇

It should be noted that in this case the indexes have been dropped due to the fact that only one
stress and strain rate component is applied during the experiment. This type of flow is also
sometimes referred to as “laminar” flow, since the fluid particles are following smooth paths in
layers, with each layer moving smoothly past the adjacent layers with little or no mixing between
them – very much like sliding a deck of playing cards. The motion of the particles of the fluid turns
out to be very orderly, with particles close to a solid surface moving in straight lines parallel to that
surface. Laminar flow is therefore a flow regime characterized by high momentum “diffusion” and
low momentum “convection” (e.g. no cross-currents patterns perpendicular to the direction of
flow).

As for the second type of flow, the analogy with the uniaxial tensile test for solids is
straightforward and the material property is defined as the ratio between the normal stress and
the elongational strain rate as follows:
9
9. Extensional (elongational) viscosity 𝜂? = :̇

Examples of this type of flow are frequently encountered during processing of materials (Fig. 2.9)
in the case of both constrained flow (for example, the acceleration of the fluid motion in the
tapering section of the extrusion die, the material flow in the gap between the rolls of a calender
and the expansion due to mold filling in injection molding) and unconstrained flow (like drawn

14
fiber spinning). For low molecular mass fluids – like water or oil – only the first type of viscosity
can be measured, since elongational behavior requires some elastic character of the fluid for it to
be able to sustain normal (tensile) stresses; for high molecular mass fluids like polymers, on the
other hand, both are types of behavior are present.

All the properties listed above are intrinsic properties of the material, in that their value does not
depend on the geometry and boundary conditions of the experiment in which they are measured.
Therefore, they can usefully and safely be employed in the design of engineering components.

DESIGN FOR DUMMIES

Before proceeding with a detailed examination of the liquid and solid state mechanical behavior of
polymeric materials, it is appropriate to recall here briefly how the properties defined above can
be used for a generic product design procedure. The scheme has been deliberately simplified here
so as to emphasize the logical flow rather than the actual procedural details.

In general, the process can be schematized in 5 distinct steps. The first step is the idea: the
stimulus for this idea may come from a market prompt, a social suggestion, and an unexpressed
need. Normally, these ideas come unexpectedly at times of special relaxation such as upon waking
up in the morning, traveling, or on vacation. These ideas can be usefully fleshed out with
diagrams, drawings, or notes that constitute the output of this stage, to be used as input for the
next stage.

The next step is the technical sketch. For this activity, one needs to use relatively simple technical
tools such as CAD software or even just a drafting square or ruler. It is necessary at this stage to
only master the most basic graphic representation techniques in order to be able to effectively
communicate our idea to the other team members as well. Once the sketch is approved, it will
constitute the input for the next stage.

The third step is the definition of the design specifications and the selection of materials. This
stage can be very complex and time consuming and may involve several skills including not only
the technical ones; its purpose is to prepare all the "ingredients" required for the complete
definition of the product that must lead to the working plan (executive project). Activities can take
place in front of the computer, at the phone or even in the laboratory for the purpose of
identifying all intrinsic properties of materials that are associated to product specifications, so as
to make sure that all characteristics required by the product can actually be met by the material
(such as stiffness, strength, thermal or electrical conductivity, color, tactile feel, etc.).

The output of this stage should be the selection of one or more materials that can be the best
candidates for the project: they will be those materials that meet the requirements of the
specifications to the highest degree. The question one has to ask himself is the following: for a
given material under examination (e.g., a 30% glass fiber reinforced polypropylene) have I been
able to find all the intrinsic property values that I will need to ensure that the product performs
correctly under the operating conditions? If the answer is yes, the material can be included in the
list of eligible materials; if the answer is no, the material must be discarded.

15
For example, if the product must bear a constant load of some significance for a long time under
high temperature conditions, it will be mandatory to have the creep compliance "master" curve of
the material and the corresponding temperature shift factor. Although the meaning of these
terms will have to be clarified later, it is clear that without this information the performance of the
product cannot be guaranteed.

Once all suitable materials have been identified and it turns out that there are too many of them,
a selection can be made based on other criteria, such as ease of sourcing, familiarity of the
production department with that specific material, sustainability, etc. Material cost cannot be
considered a parameter of selection at this stage, since the volume of the product – and hence its
mass, which is the benchmark for calculating the cost – has not yet been defined.

The fourth step is the detailed (executive) project, which will prepare all the information (final
shape, shape and size of the cross-sections, couplings, etc.) to move on to production set-up,
which typically begins with the fabrication of the production mold. This is a high-level engineering
activity, which today is almost always conducted using finite element codes. The output is the
complete definition of the part geometry for mass production. At the end of this stage, it will also
be possible to estimate the cost of the raw material to be used in the production of the part,
which will presumably be different for the different selected materials mainly because the values
of their properties are different (for example, one material with a lower E modulus than another
will require the use of more material to achieve the same overall stiffness of the part). Having thus
defined the volume of each part made of different materials, the cost of the raw material required
to manufacture them will simply be calculated based on the product of the volume times the
density times the cost per kg of the raw material itself. This this value can be a reasonable starting
point for arriving at the final choice of the single material with which to produce the part.

Before going into mass production, however, it is always advisable to carry out a pre-series of a
limited number of parts (typically about a hundred) on which to make the necessary tests of
functionality and stability of the product features over time. This step (fifth and final stage) should
be repeated iteratively until the product is fully tuned.

16
RHEOLOGY

Rheology should not o be confused with theology. It is not something related to religion, although some
people may find it so mysterious that they tend to develop some kind of belief.

The word literally means "the study (logos) of flow (rheo)" which is clearly related to the mechanical
behavior of fluids, and the term was coined expressly by polymer scientists to distinguish this branch of
science from the more general field of fluid mechanics.

The term was actually originally proposed by

Eugene C. Bingham, a professor at Lafayette College, in 1920, from a suggestion by a colleague, Markus Rei
ner. The term was inspired by the quotation mistakenly attributed to Heraclitus, (but
actually coming from the writings of Simplicius) "panta rei", "everything flows".
The theoretical aspects of rheology are the relation to the flow/deformation behaviour of material and the
rearrangement of its internal structure (e.g. the orientation and elongation of polymer molecules under
flow).

One may wonder why polymer people don't use "fluid mechanics" as their colleagues from other areas of
engineering. This apparently strange habit can be explained by the typical differentiation strategy so
common in nature to survive competition, and is basically adopted to distinguish between newtonians and
rest-of-the-world fluids. One must admit that the exact boundaries of the discipline are not perfectly
established, and the gray-shaded area has been progressively expanded to biological fluids, food and soft
solids well beyond its original limits.

Rheology has also the merit of establishing a connection between two seemingly unrelated fields of
mechanics (flow and plasticity) by recognizing that materials undergoing these types of deformation are
unable to support a stress, and particularly the shear stress, in static equilibrium. In this sense, a solid
undergoing plastic deformation behaves very much like a fluid – although no viscosity coefficient is
associated with this flow.

The constitutive equations used in rheology may change depending on the scope: if the emphasis is on the
relation between the input of the experiment (for example the shear rate, which can be controlled by
controlling the flow throughput from a capillary) and the output (the shear stress in this case, which can be
inferred from the force required to maintain a given throughput) they are called flow equations, and their
graphical representation is called the flow curve of the material.

If the emphasis is instead on the material property rather than on the flow behavior, then the results of the
experiments can be reported as viscosity curves, namely 𝜂 = 𝜂(𝛾̇ ) and 𝜂 = 𝜂(𝜏). The most frequently
used version of the viscosity curve is the first of the two. Due to the (relatively) large variations of the
property over a (relatively) wide range of variation of the controlling variable (which can be either the
strain rate or the stress), more frequently the viscosity curves are represented in logarithmic scales.

Clearly, the value of the viscosity represented in the viscosity curve depends on the constitutive equation
which is used to fit the flow curve. If the curve is a straight line (newtonian behavior) then the value of the
viscosity is uniquely determined by the slope of the line according to the following equation:

𝜏 = 𝜂! 𝛾̇

in which 𝜂! is the so-called newtonian viscosity. While this approximation may hold true – at least in the
first approximation – for polymers at very low shear rates, while increasing the shear rate the flow curve
turns out to be highly nonlinear (Fig. 4.1). Therefore a problem arises when using the previous equation on
how to calculate the slope of the curve: of the two most commonly used options, i.e. the secant value

17
(“apparent” viscosity) 𝜂" and tangent value 𝜇, the first is the most commonly used but the last is normally
to be preferred. When referring to the apparent viscosity 𝜂" , the subscript is often omitted for simplicity.

Based on the viscosity curve, the rheological behavior of fluids can in general be classified into four main
categories:
1. Newtonian 𝜂 = 𝑐𝑜𝑛𝑠𝑡
2. Time independent 𝜂 = 𝜂 (𝛾̇ )
3. Time dependent (momentarily) 𝜂 = 𝜂 (𝛾̇ , 𝑡)
! $#
4. Viscoelastic (elastoviscous) 𝜂 = ℱ##! $%& [𝛾̇ (𝑡 ' )]

Obviously, categories 2, 3 and 4 fall into the non-newtonian type of behavior; in the remainder of this
analysis, only categories 1 and 2 (steady state situations) will be examined into the detail.

The time independent non-newtonian type of behavior can be divided into two subcategories:

2.1 Pseudoplastic (shear thinning) behavior

2.2 Dilatant behavior

In the first case, the viscosity curve is a monotonically decreasing function of both the shear rate, 𝛾̇ , and the
shear stress, 𝜏, while the opposite holds for the second case (Fig. 4.2). The effect on the flow curve is also
shown in Fig. 4.3. The vast majority of polymers shows – to a different extent – a clear preference for the
pseudoplastic behavior: this has been explained from a molecular point of view by assuming that, being the
macromolecular random coils highly deformable in the fluid state, they will tend to deform under the
action of the shear stress field by orienting in the flow direction (Fig. 4.4). This change in the
macromolecular conformation is expected to severely reduce the resistance to flow, which in turn reduces
the value of the viscosity for increasing the shear rate or the shear stress. This reduction is seen to attain a
limiting, steady state value for very high shear rate/stress values.

This type of behavior, which basically imply that the pseudoplastic behavior is simply a transition between
one equilibrium state (for the liquid being at rest, macromolecules are in the random coil equilibrium
configuration) and the next (for the liquid being highly stressed, macromolecules are more and more
oriented until they reach the maximum possible extension) allows for the identification of three different
ranges in the viscosity curve:

I: lower newtonian regime

II: pseudoplastic (transient) regime
III: upper newtonian (steady state) regime

The fact that the lower newtonian range is not classified as a steady state regime comes from the trivial
observation that the asymptotic value for lower shear rates/stresses is an artifact due to the use of the
logarithmic scale on the horizontal axis, which stretches the lower values to - ∞: in fact, when plotted in a
linear scale (Fig. 4.5) the viscosity curve lower asymptotic value completely disappears …

In spite of the fact that it is clearly not a "real" asymptote, the lower newtonian asymptote has become
very popular among rheologists because it turns out to be very sensitive to both internal (i.e. molecular)
and external (i.e. environmental) variables. These topics will be better explored later in this Chapter.

The last category of flowing behaviour of some interest for polymers is the one which requires to overcome
a stress threshold. It is named after Eugene C. Bingham (the “founder” of rheology) who was able to
develop its mathematical description:

𝜏 = 𝜏( + 𝜇 𝛾̇

18
When an increasingly higher stress is applied to a Bingham's fluid, it will have a solid-like behavior (e.g. no
flow) until the stress reaches the threshold 𝜏( , which is called Bingham's stress or “yield” stress. Therefore,
the liquid behaves like a solid at lower stresses and becomes a fluid at higher stresses. Bingham originally
observed that beyond threshold 𝜏( the flow behavior was of newtonian type, with the shear rate linearly
increasing with the applied stress, but it doesn't have to be that way for all fluids: the “post-yielding” flow
behavior, in fact, can also very well be non-newtonian (Fig. 4.6). The model applies not only to some
polymeric fluids, but also to mayonnaise, toothpaste, paints, muds and slurries among the others.

Plotting the viscosity curve of a Bingham newtonian fluid can be more tricky: if one chooses the version 𝜂 =
𝜂(𝛾̇ ) nothing changes compared to a regular fluid, since the viscosity transition from 𝜂 = ∞ to 𝜂 = 𝜂!
strictly speaking occurs at 𝛾̇ = 0. Therefore, the other representation 𝜂 = 𝜂(𝜏) is recommended in this
case (Fig. 4.7), for which the viscosity curve shows a asymptotic vertical axis right at 𝜏 = 𝜏) . For real
Bingham fluids, however, a transition is observed which is much less sharp than predicted by the
mathematical model.

APPENDICE

Viscoelasticity for Dummies

(simple tips to help you in solving the exam problems)

First of all, we should clarify that problems and exercises are not the same: while an exercise is
designed to teach you a solving method for a specific task (for example, how to find the coefficient
of the Eyring’s equation based on experimental data) and you may expect that the only variations
from exercise to exercise are the experimental data, a problem is designed to simulate a real
situation in a working engineering environment. While in the exercise everything is known from
the beginning (the formula, the math required to solve it and the experimental data) you should
not expect the same situation in a problem: in most of the cases the above boundary conditions
are only partly known, and you should find a way to combine the available information with your
background knowledge to complete the overall picture. You should therefore not expect for a
problem to have a “standard” procedure that allows you to find the solution.

19
Given this situation, you should reinforce your basic knowledge and train your solving skills before
affording an exam problem. Since all formulas – together with their theoretical foundation –
required to solve the problems are clearly explained in the lecture notes you should first make
sure that you know them perfectly in all details. The extra – very basic – skills required to
effectively accomplish the task of getting the “right numbers” out of them within the time
available during the exam are listed below in order of importance:

- Practice: in order to become familiar with the solving strategies, you should practice on
your own using the same tools (ruler, pen, paper and pocket calculator) and within the
same time (approx. 45’) which are allowed during the exam. Take note of your progresses
and - after 45’ – stop writing and carefully go back through the procedure you adopted to
identify and remove possible problems. Use the solutions which are made available on
Beep for classroom exercises and after each exam session. Ask your teachers if you need
more support.

- Math: some knowledge of basic calculus is required to solve the integro-differential

equations of viscoelasticity and to get around possible difficulties. This is something that
you should already know but - due to the fact that you probably have seen it a looong time
ago - it is briefly recalled here.

- Plotting: this is a basic skill which might be seen as trivial, but it proves to be very useful in
many situations. Sketching a plot of everything – loading history, thermal history, material
properties like creep compliance or relaxation modulus, expected solutions, final results,
etc. – may greatly help you to avoid stupid mistakes, like the use of wrong units of
measurement or failure in making proper calculations.

- Units: you should know how to manage the different units in order to get the right
numbers out of your equation.

Some useful tips are found below.

MATH

There are many textbooks available to teach you how to solve integro-differential equations. A
good book for students who need a refresher is “Calculus for dummies” by Mark Ryan, Wiley
(2014) ISBN-13: 978-1119293491.

Some general methods of integration are listed below:

a) Power law: the following formula can be applied to algebraic as well as trigonometric,
logarithmic and exponential functions:

@%&!
∫ 𝑢3 𝑑𝑢 = 3A%
+ 𝐶 for any (n ≠ -1)

in which u(x) can be any function of x. In order to apply this method you should use a
substitution; for example, if the following integral has to be solved:

20
 2
∫ √2 " A% 𝑑𝑥

one can make the following substitution:

u = x2 + 1; du = 2x dx

and therefore the integral reads:

1 1 1 %
X 𝑑𝑢 = X 𝑢;& 𝑑𝑢
2 √𝑢 2

Based on the power method, the solution turns out to be:

%
1 𝑢& %
+ 𝐶 = 𝑢 & + 𝐶 = E𝑥 & + 1 + 𝐶
2 1Z
2

This method is very useful since power-law functions are frequently used in viscoelasticity.

b) Exponential function: the formula is:

X 𝑒 @ 𝑑𝑢 = 𝑒 @ + 𝐶

which is remarkable, since the integral here is the same as the expression we started with.
As well as for the previous method, a substitution is required for u(x). For example:

X 9𝑒 &2 𝑑𝑥

can be solved by making the following substitution:

u = 2 x; du = 2 dx

The integral then becomes:

𝑑𝑢 9 @
X 9𝑒 @ = 𝑒
2 2

which implies that the solution is:

9 &2
X 9𝑒 &2 𝑑𝑥 = 𝑒 +𝐶
2

Exponential functions are quite frequent in viscoelasticity - especially when dealing with
analogue models.

c) By parts: this method is particularly useful when you meet an integration which is the
product of 2 functions, say u(x) and v(x):

21
 X 𝑢𝑑𝑣 = 𝑢𝑣 − X 𝑣𝑑𝑢

This formula allows you to turn complicated integrals into more simple ones by
choosing “u” and “dv” carefully. When you have a mix of functions in the expression to be
integrated, in general, the best choice is the one that allows du/dx to be of a simpler form
than u. For example, the following choices for “u” are suggested in order of preference:

1. Let u = ln (x)

2. Let u = xn

3. Let u = enx

An example is given below for the integration of the logarithmic function:

X ln(𝑥) 𝑑𝑥 = X 𝑢 𝑑𝑣

𝑑𝑥
= x ln(𝑥) − X 𝑣 𝑑𝑢 = 𝑥 ln(𝑥) − X 𝑥
𝑥

= 𝑥 ln(𝑥) − X 𝑑𝑥 = 𝑥 ln(𝑥) − 𝑥 + 𝐶

d) Change the function: this method can be used when we meet a function which cannot be
easily integrated. It consists of replacing the current function with one you can better
handle, like a power law (method a) or an exponential (method b).

For example, when applying a series model of two compliances to the solution of a
relaxation problem – which is a typical case in experimental testing, when you have to
account for both the compliance of the specimen and the compliance of the mechanical
frame – you may end up with a function like 1/(1 + A xn) which does not have an analytical
solution for integration. First of all you recognize that the function is a monotonically
decreasing function of x, having an asymptotic value of 1 for x = 0. You can therefore use
the customary power law f (x) = B x-n which has of course to be limited for x < 1 to f (x) =
const. = 1. This is a reasonable assumption for short-to-medium times. For larger times, a
logarithmic function may be more appropriate (see solution of Nov. 2017 midterm
examination problems for more details).

e) Numerical solution: if nothing else works - or you don’t have enough time to find a proper
analytical solution method - you can always apply a numerical method for integrating your
function. An example in which the trapezoidal rule is applied is provided below.

From a plot of the shift factor vs. temperature you can get the following dataset:

TEMPERATURE
a23
−10

22
 [ °C ]
23 1
10 102
0 104
-10 106

and, based on the fact that the temperature is changing linearly in time, you should replace
the temperature values with time values based on the value of the heating/cooling rate.
For integration purposes – as in the case you should solve a variable temperature problem
- two options are available: either fit the data to a suitable function which can then be
integrated analytically or integrate numerically. The first method limits the possible choices
to a small number of functions; in general, the exponential function is the most
recommended (see Handouts, page 24). The second solution is obviously more practical. By
recalling that the - finite - integral value of a curve is the value of area below that curve
within two specified abscissa limits, a reasonable approximation is to calculate the area
below each couple of adjacent datasets by assuming that it can be approximated by the
area of a trapezium having the two parallel sides equal to the ordinate values of the points,
and the height equal to the difference of the abscissa values.

The application of the method to the 4 data points shown in the table above will provide a
method to calculate the reduced time in case of the temperature varying with a linear
ramp from -10°C to room temperature for t1 = 2135, t2 = 2428 and t3 = 2700 s:

2135
ξ1 = ∫ T =
t (
1+10−2 ⋅ 335 )
= 170 [s]
1800 aT
0 2

2428
ξ2 = ∫ T =
t ( ) (
1+10−2 ⋅ 335+ 10−2 +10−4 ⋅ 293 )
= 171 [s]
1800 aT
0 2

2700
ξ3 = ∫ T =
t ( ) ( ) (
1+10−2 ⋅ 335+ 10−2 +10−4 ⋅ 293+ 10−4 +10−6 ⋅ 272
= 171
) [s]
1800 aT
0 2

23