Department of Engineering Science

University of Oxford

Deformation of Polymers

4 Lectures for Paper C3

Prof. J.C. Tan

Hilary Term 2016

1
 Contents of the lectures
1. Introduction
2. Preliminaries (B5 lectures): Large
deformations & rotations, Stresses in the
presence of finite deformations and rotations
3. Elasticity of polymers: Linear elasticity;
Hyperelasticity; Phenomenological &
physically-based models
4. Viscoelasticity: Linear viscoelasticity in 3D;
Relaxation and retardation spectra; Thermal
activation; Glass transition …

2
 1. Introduction
• The subject of these lectures is ‘materials
modelling’ applied to the deformation of
polymers.

What does this mean?

• ... the business of describing quantitatively, via

mathematics, the 3D deformation that occurs in
polymer solids under the action of applied loads.

3
Why is this subject important in Engineering?

• Designers of manufacturing
processes and of products need to
be able to predict quantitatively in
advance the deformation that will
occur under applied loading. O-rings, gears, impellers, …

• Materials engineers, developing

and selecting materials for
particular applications, need to
predict how the (3D) deformation of
a polymers depends on their
composition and structure.
Orthopaedic implants
(UHMW-PE knee meniscus) 4
 Computer-based engineering analyses
• Engineers nowadays use finite element
(FE) numerical analysis to predict the
complex patterns of stress and strain
that usually arise in 3D ...
• But as always, such computer-based
FE model of an
analyses follow the general rule of artificial knee
computer programs: implant

(RiRo) “Rubbish input gives rubbish output”

• So it is essential to feed FE analyses with good

models of the material behaviour!
5
How do we create mathematical
models of material behaviour?

• Where does the information come from?

• It comes to us from two sources at two

different length-scales, which comprise
the “macro-world” and the “micro-world”

6
Handouts P.4-5

(I) “Macro-world” > 1 mm à few m

This is the everyday world you are familiar with, visible
with the naked human eye!

(a) A few known physical laws applying at this length scale

(e.g. Newton’s laws, the laws of thermodynamics...)
(b) Our idealisations of the material (a single continuum,
assume some known symmetry...)
(c) Experimental data of specific materials, specific test
conditions & parameters, specific samples...

“Phenomenological” Models

7
 (II) “Micro-world” < 1 mm
(a) Using microscopy and other advanced
characterisation tools (eg. EMs, X-rays, TEM

neutrons) to gain insights into materials

composition and microstructure
(b) Known physics and chemistry provide useful
theories to devise more powerful models that
capture more aspects of material behaviour
(c) Guide us in making predictions of how the
material is likely to behave in the macro-
world

“Physically-based” Models Synchrotron radiation

(Diamond Light Source @ Harwell)
8
Handout P.5

What polymers are?

• Polymers are materials consisting of very long
molecules, made up of many repeating chemical
units (the monomer units), covalently bonded
together.

Amorphous (e.g. PS, Semicrystalline (e.g. HDPE,

PMMA, LDPE, atactic PVC) Nylon, isotactic PVC)

9
 Molecular topology
linear
e.g. linear PE (“high density
polyethylene” or HDPE)

branched e.g. branched PE with 6 - 10

branches per 100 carbon atoms
(“low density polyethylene” or
LDPE)
crosslinked

e.g. cross-linked polyethylene

10
 What are the key features of the
chemical architecture of a polymer?
1. Chemical composition(s) of its monomers
2. The number of monomers in each molecule
(the ‘degree of polymerisation’ = n)
3. The manner of covalent connection between
the monomers (the ‘molecular topology’)
H H CH3

* C C C O C O
n* *
n
*
H H CH3 O
Polyethylene (PE) Polycarbonate
(2 main-chain covalent bonds (nb0), (6 main-chain covalent bonds,
28 amu: molar mass M0 = 28 g/mol) M0 = 254 g/mol)
11
Handout P.7

Stretched out lengths of polymers

Molecule b
a polyethylene
axis
(PE) molecule
bp
Let the number of main-chain bonds per monomer be nb0

Then the total number of such bonds per molecule is

nb = n nb0 n
Let the mean main-chain bond length, as projected onto the axis of the
molecule when stretched out straight, be bp

Then the stretched out length of one monomer, L0, and that of the
whole molecule, L, are respectively

L0 = nb0bp ; L = nL0 = nbbp (1.1)

12
 Structure of polymer crystals
e.g. A lamella-shaped single crystal of polyethylene (PE)
• The crystals are tiny
(10–50 nm thickness) Its unit cell
the chains
TEM pack parallel
along c-axis
looking down
c-axis c (254 pm)
−
= 0.254 nm
Thickness

looking
down
c-axis

c-axis

The molecules are aligned normal to

the lamella-shaped crystal 13
Handout P.7

“Microstructures” of spherulites in PE
In a polymer crystallised from the melt, instead of lamellae
forming independently, stacks of them grow outwards
together from each nucleus, forming a ‘spherulite’.

200 μm
“Maltese crosses” visible under cross-polarisers
14
Handout P.8

Sizes of typical structural

features of polymers
“Micro-world”

vs.
Typical size of polymer-based engineering
components/products in the “Macro-world”:
a few millimetres (mm) up to several metres (m)
15
Handout P.9

2. Preliminaries (B5 Revision)

2.1 Large deformations and rotations

• We must use a
mathematical description
that does not suffer from
restriction to small strains

• Require a formulation that is

valid for arbitrarily large
deformations and rotations

Assume it is a continuum:
the new position x is a
smooth function of the
The vector displacement:
old position X u=x─X
16
Handout P.10

Deformation gradient F
• It follows that we can
arbitrarily pick another point
in the material, initially at
position Q0, close to P0,
• ... track it to its new position
Q, by applying a
transformation F to their
initial vector separation dX
to obtain their new vector
separation dx

dx = F dX ; or in component terms dxi = Fij dX j

F is called the Deformation Gradient ∂xi
Fij = = ∇0 x
It fully specifies the deformation and rotation of the ∂X j
sample of material as it transforms from B0 to B
17
 Rigid body rotation R
F can always be expressed as the product of an
orthogonal matrix R and a symmetric matrix U or V:

F = VR = RU
• R corresponds to a pure rigid body rotation
• U and V correspond to pure deformations
• The eigen values of U and V are equal (i.e. differ by only
a rotation), and they are called the Principal Stretches

Definition:
“stretch”
λ=
new length λ1, λ2, λ3
original length
18
Handout P.11

Left Cauchy Green tensor B

“Principle of Material Frame Indifference”:
• Stresses that give rise to the deformation must be
independent of any rigid body rotation R

• Consequently, to describe the deformation we do

not use F itself, but instead we could use U or V
• ... but actually it turns out to be more convenient to
use instead the left Cauchy Green tensor B or right
Cauchy Green tensor C defined

Here we T T
shall use: B = FF ; C =F F
(3x3 Symm.)
2 2 2
The eigen values of both B and C are λ1 , λ2 , λ3 19
Handout P.14

True stress in 3D
Stresses in the presence of finite deformations and
rotations (j)
Assuming a continuum, we define
n
n
(i)
traction vector t as the local force
ΔP
per unit area, in the limit
t ΔP
Δa t=
Δa Δa →0

Cauchy Stress Tensor σ

This is a measure of true stress; it relates the traction vector t to the
direction n of the normal of the plane on which it is acting:

t = σn or in component terms ti = σ ij n j
σij is the i-th component of the force per unit current area acting on
a plane whose normal is currently in the j-th direction. 20
Handout P.15

3. Elastic Deformation of Polymers

3.1 The general case Polymer sample as a thermodynamic
system with U and S at absolute T
Under some circumstances
Q W
solid polymers are elastic in T
their mechanical response, U, S
particularly at
• very low temperatures, and
• at high temperatures (>Tg) if First and Second Laws of Thermodynamics:
they are crosslinked or ΔQ
sufficiently entangled
ΔU = ΔQ + ΔW ; ΔS ≥
T
Helmholtz Free Energy (per unit initial volume) A ≡ U ─ TS
Rate of work done
(power supplied) by ! ≥ U! −TS! = A
W ! + ST!
applied loads:
• the equality applies to a reversible process = “Elastic”
• the inequality applies to an irreversible process 21
 Elastic deformation of a polymer
at constant temperature
We define a material as elastic if deformation under applied
loads is a reversible process (thus “=” sign).

! =A
W ! + ST!
But at constant temperature T! = 0
Power supplied
to the system ! !
W =A
Rate of change in
the free energy

To complete a model, we need to express W ! in terms of σ,

and A! in terms of B. HOW to establish such relationships?
22
 Linking Cauchy stress σ to B
... By reading handouts P.16 to P.20, we can show that:

Cauchy stress ρ ⎛ ∂A ⎞
tensor (symmetric, σ =2 ⎜ B⎟ (3.7)
σij = σji) ρ0 ⎝ ∂B ⎠
density ratio
• Note that we have not yet made any restriction to small deformations
• Eqn. (3.7) is valid for hyperelastic material (large elastic deformations)

• Isotropic invariants of B applicable to isotropic materials

(3.11)
1⎡ 2
I1 = trB ; I2 = ⎢( trB ) − trB 2 ⎤⎥ ; I3 = det B
2⎣ ⎦
23
Essentially the 3 eigen values of B
Handout P.23
3.3 Hyperelasticity:
Phenomenological modelling
• A material that is elastic even at large deformations is
called “hyperelastic”, i.e. rubbery

What happens to the volume during deformation?

Since mass is constant, the density ratio after deformation is

ρ old volume 1 1 1
= = = = (3.21)
ρ0 new volume λ1λ2λ3 det B I3

3rd isotropic invariant

Principal new length
λ=
stretch: old length I3 = det B = λ12λ22λ32
24
Handout P.23

Hyperelastic “rubbery” polymers

Experimentally, hyperelasticity is seen in polymers only
under certain conditions:
• at temperatures above the glass transition, T >Tg (rubbery)
• when the polymer is amorphous
• when there is connectivity between the molecules
(chemical crosslinks or entanglements providing ‘physical
crosslinks’).

- Under these conditions, the molecules move very

freely past each other, but they are reluctant to
move apart from each other (volume conserved!).

25
Handout P.24

G << K and ν ~0.5

• In hyperelastic polymers, the change of volume (K) meets
much more resistance than change of shape (G), i.e.

• e.g. Typically G ~ 1 MPa, while K ~ 1 GPa

• Hence ν ~ 0.5, i.e. material is:
G≪K
⎛ 2G ⎞
INCOMPRESSIBLE ⎜1− ⎟
(constant volume) 1 ⎝ 3K ⎠ 1
ν= ! (3.20)
2⎛ G ⎞ 2
⎜1+ ⎟
⎝ 3K ⎠

In light of these observations, there are two

consequences for modelling hyperelasticity... 26
Handout P.24

Modelling rubbery polymers

(a) I3 =1 (see eqn (3.21)), since incompressible λ1λ2λ3 = 1
A depends on only I1 and I2 and the density ratio is unity

(b) The hydrostatic component of stress (–p) is

indeterminate from the constitutive model alone, and
must be found from the boundary conditions (BCs).

Thus for a known deformation state, the Cauchy stress

is obtained from either:
(3.22)

∂A (I1, I2 ) ∂A ( λ1, λ2, λ3 )

σ =2 B − pI or σ =2 B − pI
∂B ∂B
27
Handout P.25

Cauchy stress expressed in terms of

the isotropic invariants
For the models expressed in terms of the invariants, it is

helpful to evaluate first their differentials with respect to B

⎡⎛ ∂A ∂A ⎞ ∂A 2 ⎤
σ = 2 ⎢⎜ + I1 ⎟B − B ⎥ − pI (3.25)

⎣⎝ ∂I1 ∂I2 ⎠ ∂I2 ⎦

Hydrostatic stress -p (pressure) to be found

from the boundary conditions (BCs)

28
Handout P.25

Phenomenological hyperelastic models

The parameters C0, C1, C2, CG, Im, µn and αn are material parameters,
to be found empirically by fitting the models to experimental data.

29
Handouts P.26-27

Example problems

30
Handout P.28

neo-Hookean model vs. experiments

• Relatively simple to use as it
involves only one material
expt. parameter C0
A = C0 (I1 − 3 ) Table P.25
• Predicts that the Cauchy
stress is a linear function of B
λ ~ 1.2
σ = 2C0B − pI (3.26)

Tension • Limitations?
- It cannot fit real data for
Compression rubbery polymers to within
accuracy acceptable for many
purposes, beyond a nominal
strain of about 20% (λ ~ 1.2)
31
Handout P.29

neo-Hookean vs. uniaxial tensile data

Nominal stress s is found from Data are plotted as nominal stress (s)
true stress σ by exploiting the versus nominal strain (λ ─ 1)
incompressibility: 7

6
Natural rubber
σ 11

nominal stress (MPa)

s11 = (3.30)
5 Strain Exp.
λ stiffening
4
There are two systematic errors: Strain
3
(1) over-estimate of stress at softening
intermediate values of λ, and 2
(2) under-estimate of stress at 1
highest value of λ.
0
Potential solutions...? 0 1 2 3 4 5 6 7
1) Mooney-Rivlin; 2) Gent
nominal strain λ − 1
32
Handout P.30

Example problem 3

33
Handout P.31

Hyperelasticity: Physically-based model

Chemical structures of some common polymers that are
hyperelastic at room temperature (RT)
H CH3 H H H H H H CH3 H Cl H H

C C C C * C C C C * * O Si * * C C C C *
* * n
n n n
H H H H CH3 H H

1,4 polyisoprene 1,4 polybutadiene poly(dimethyl siloxane) 1,4 polychloroprene

(‘natural rubber’) (PDMS ‘silicone rubber’) (‘neoprene’)

• These are all flexible molecules with Tg < RT

• They are liquids at RT until they are chemically crosslinked
(e.g. using sulphur) at discrete points,
• ... creating a 3D molecular network that cannot flow, and
hence forms a soft solid.
34
Handout P.32
Entropic elasticity
Rapid changes
of shape
Crosslink
Some resistance to
deformation comes
entirely from the
resistance of the system
to reduction of its entropy
• But even though crosslinked, there is
vigorous Brownian motion above Tg,
as they would in the liquid state
• Between crosslinks, the molecular
chains move freely relative to each
other
• The network changes shape with
negligible cost in its internal energy U
U ≪TS
• The free energy of deformation A
can be considered entirely entropic
in origin (recall: A = U – TS)
A ! −TS 35
Handout P.33

Physical model assumptions

nK
LK Kuhn segment • Model derived following the
e3 approach first suggested by Kuhn

a Kuhn • Imagine each long molecule has a

chain r
set of rigid segments (nK, LK)

e2 • ... connected together by universal

e1
Crosslink
(or entanglement)
• For such an “imaginary” chain
LK = length of each Kuhn segment
nK = no. Kuhn segments per chain
r = instantaneous end-to-end
separation (a broad distribution);
the directions of r will be random.
joints as if it were a freely-jointed
chain à “Kuhn chain”
molecule, the direction of each
segment is, by definition, totally
uncorrelated to the directions of the
other segments.
36
Handout P.34

3D random walk
• In the absence of external forces (subscript 0) acting on the network,
there will be equal numbers of bonds pointing in all directions

• The path r from one end of the chain to the other is a 3D “random
walk”. This allows us to do some statistics on it, c.f. handout P.34...

NB. <…> denotes avg. over the

whole population of chains
r2 = nKL2K (3.34)
0

• It is of interest to compare the unperturbed (i.e. coiled up &

compact configuration) root-mean-square end-to-end distance rrms
to the stretched-out length of the chain L:

2 L
rrms ≡ r = LK nK ; L = LK nK ; hence = nK (3.35)
0 rrms

λmax 37
Handout P.35

..with externally applied stresses..

i-th Kuhn segment • When stresses are applied to the
External polymer in the macro-world, in the
e3 micro-world this corresponds to
forces
microscopic forces being applied
f
to each individual chain via the
r crosslinks.
Crosslink -or-
-f e2
Entanglement • The forces perturb the chain from
e1 its random shape, and thereby
reduce its entropy

• Derivation of the relationships between magnitude of applied force f

and end-to-end distance r, and the final link to continuum Cauchy
stress σ ... See full details in your Handouts P.36-40!

38
Handout P.41

“Inverse Langevin” Model

NkT ⎛ λmax ⎞ −1 ⎛ λ ⎞
σ= ⎜ ⎟ L ⎜ ⎟ B − pI (3.49)
3 ⎝ λ ⎠ ⎝ λmax ⎠
N chains per Hydrostatic stress -p (due to intermolecular
unit volume potentials; vdW) to be found from BCs, as this
model assumes incompressibility (K>>G)

= 1I = trB
where λ is the rms stretch
3 1 3
(3.46)

λ
Maximum stretch max = nK = 1I (3.35)
3 max
⎛ 1− 1 x 2 ⎞ is the Langevin function (3.48)
L −1 ( x ) ! 3x ⎜ 3 ⎟ approximated using the
⎜ 1− x 2 ⎟
⎝ ⎠ Padé expression 39
Handout P.43

Relating to real molecular dimensions

• To use the Inverse Langevin model to predict the behaviour of a
particular rubbery polymer, we need to relate the parameters of the
freely-jointed Kuhn chain to the real molecular dimensions
• ... by equating the (a) unperturbed rms end-to-end distance, and (b)
stretched-out length, respectively:
Characteristic ratio Bond length Total # bonds between crosslinks

rrms 2 = r 2 = nK L2K = C∞nbb 2; L = nKLK = nbbp (3.53)

0
(a) Kuhn Real (b) Kuhn Real
• If we call Mc the molar mass of the chains between crosslinks, then
from the definition of Avagadro’s number NA, N and nb are both
determined by Mc thus

ρ Mc crosslink
N = NA ; nb = nb0 (3.54)
# chains per
unit volume
Mc M0 monomer 40
Handout P.7 – RECAP

Stretched out lengths of polymers

Molecule b
a polyethylene
axis
(PE) molecule
bp
Let the number of main-chain bonds per monomer be nb0

Then the total number of such bonds per molecule is

nb = n nb0 n
Let the mean main-chain bond length, as projected onto the axis of the
molecule when stretched out straight, be bp

Then the stretched out length of one monomer, L0, and of the
whole molecule, L, are respectively

L0 = nb0bp ; L = nL0 = nbbp (1.1)

41
Handout P.45

Example problem 4

42
Handout P.46

4. Linear Viscoelasticity of Polymers

• Polymers frequently deviate from purely elastic behaviour.

• They are then said to be “inelastic”, in which deformation

under applied loads is an irreversible process and hence the
net work done is positive
! dt = 0
ΔW > "∫ A (4.1)

• Net work done is dissipated as heat, or locked inside the

material as a permanent change in (micro)structure.

• One form of inelasticity is linear viscoelasticity, occuring at

small strains of typically < ~0.005 (0.5%)

• It behaves as a time-dependent, but linear system obeying the

Boltzmann Superposition Principle (similar to an electric circuit
containing only resistors, inductors and capacitors, see B5/B8).
43
Handout P.20-22 – RECAP (B5, A3)

! dt = 0)
3D linear elasticity ( "∫ A
• Let the Infinitesimal strain tensor ε becomes vanishingly
small relative to unity

• Dilation (vol. strain) Δ = tr ε = ε11 + ε 22 + ε 33 (3.16)

Δ
• Deviatoric strain tensor e≡ε− I (3.18)
3
Mean vol. strain
3D equation of
linear elasticity
for modelling σ = K Δ I + 2G e (3.19)

small strains
Bulk modulus Shear modulus
Cauchy stress Mean stress Deviatoric stress 44
Handout P.47

3D equation of linear viscoelasticity

For an isotropic linear viscoelastic polymer, the 3D equation
of linear elasticity - eqn (3.19) – is then replaced by the
corresponding 3D equation of linear viscoelasticity

σ = K Δ I + 2G e (3.19)

Convolution integral:
t dΔ t de
σ (t ) = ∫ K (t − u ) du I + 2 ∫ G (t − u ) du (4.2)
0 du time step 0 du
time-dependent
Cauchy stress
K(x) G(x)
3D stress response is fully determined by 2 material time-dependent functions:
K(x) = Bulk stress relaxation modulus function
G(x) = Shear stress relaxation modulus function
45
Handout P.47

Expressing σ(t) in terms of σm & σ’

Using * to represent a convolution integral, we can express
equation (4.2) in terms of the mean stress σm and
deviatoric stress σ’ :

t dΔ t de
σ (t ) = ∫ K (t − u ) du I + 2 ∫ G (t − u ) du (4.2)

0 du 0 du
Hydrostatic Deviatoric
response response (4.3)

s = σ mI+ s ʹ where σ m = K ∗ Δ! ; s ʹ = 2G ∗ e!
Experimental results for polymers always show much less time-
dependence in the hydrostatic response than in the deviatoric response
(e.g. a factor of ~1000 in G(x) relaxation vs factor ~2 in K(x) at Tg). 46
Handout P.49-50

Spring-dashpot analogy
Zener model: Linear viscoelastic behaviour similar to that of a polymer is
exhibited by Hookean springs (k) and Newtonian dashpots (c) coupled
suitably in series or parallel.
... Revise your B5 & B8 notes ...

Zener model captures qualitatively all the key features

1 2 typical of polymers, e.g. in tensile tests (creep, stress
relaxation, dynamic tests/cyclic deformation ...)
!k
kR The response of this model is defined by three equations:
c dy F1 1 dF1
F = F1 + F2; = + ; F2 = kR y (4.8)
dt c Δk dt
y Defining the two time constants of the system:
F
Relaxation time Retardation time
c ( Δk + kR ) ; τF > τ y (4.9)
τy = ; τF = τ y
Δk kR 47
(input = y) (input = F)
Handout P.50

Zener model
Combining equations (4.8), and employing time-constant
definitions (4.9), gives a single differential equation governing
the response of the Zener model:

dF ⎛ dy ⎞
F +τy = kR ⎜ y + τ F ⎟ (4.10)
dt ⎝ dt ⎠
Relaxation time Retardation time

Now we consider three example cases.

48
Handout P.51-53

Case 1: Creep and Recovery

A step load F0 is applied at time 0 and removed at time tu
F0 = 1 N; y 0 = 1 mm; tu = 2 s.

displacement y (mm)
2.0
The resulting displacement is: creep
1.5 Response
⎧⎪ ⎛ t ⎞ ⎤ ⎫⎪
y =⎨
1 ⎛ 1
+⎜ −
1 ⎞⎡
⎟ 1 − exp ⎜ − ⎟ ⎥ ⎬ F0 ( 0 < t ≤ tu ) 1.0 y0 recovery
( Δk k ) ⎜k ( Δk k ) ⎟⎢
⎩⎪ + R ⎝ R + R ⎠⎣ ⎝ τ F ⎠ ⎦ ⎪⎭
0.5
tu
⎧⎪⎛ 1 1 ⎞⎡ ⎛ ( t − tu ) ⎞ ⎛ t ⎞ ⎤ ⎫⎪ 0.0
y = ⎨⎜ − exp ⎜ − − exp ⎜ − ⎟ ⎥ ⎬ F0 ( t > tu ) .
⎜ ⎟
⎟ ⎢
⎪⎩⎝ kR ( Δk + kR ) ⎠ ⎢⎣ τF ⎠
⎟
⎝ τF 1.0 F0

force F (N)
⎠ ⎥⎦ ⎪⎭
⎝
Input
0.5
(4.11)
0.0
0 1 2 3 4 5 6 7
Let: time t (s)

kR = 1 N/mm; Δk = 1 N/mm; c = 1 Ns/mm ∴ τ y = 1 s; τ F = 2 s.

49
Handout P.52

Case 2: Stress Relaxation

A step in displacement of magnitude y0, is applied at time 0
and reversed at time tu. y = 1 mm; t = 2 s.
0 u
The required force is: 3
⎡ ⎛ t ⎞⎤ 2 relaxation Response

force F (N)
F = ⎢kR + Δk exp ⎜ − ⎟⎟ ⎥ y 0 ( 0 < t ≤ tu )
⎢⎣ ⎜ τy 1
⎝ ⎠ ⎥⎦
0
⎡ ⎛ t ⎞ ⎛ ( t − tu ) ⎞ ⎤
recovery
-1 tu
F = Δk ⎢exp ⎜ − ⎟⎟ − exp ⎜⎜ − ⎟⎟ ⎥ y 0 ( t > tu ) .

displacement
⎢⎣ ⎜ τy τ y ⎠ ⎥⎦ 1.5
⎝ ⎠ ⎝

y (mm)
1.0 y0
0.5 Input
(4.12)
0.0
0 1 2 3 4 5 6 7
time t (s)

kR = 1 N/mm; Δk = 1 N/mm; c = 1 Ns/mm ∴ τ y = 1 s; τ F = 2 s.

50
Handout P.52

Case 3: Dynamic Mechanical Response

A continuous oscillating displacement y * = y 0 exp i ωt is applied

The required oscillating force is F * = F '+ iF " where

⎛ Δkω 2τ y2 ⎞ ⎛ Δkωτ y ⎞ F*
F"
F ʹ = ⎜ kR + 2 2
⎟ y *; F ʹʹ = ⎜ 2 2
⎟y * (4.13) δ
⎜ 1 + ω τ y ⎟⎠ ⎜ 1+ ω τ y ⎟
⎝ ⎝ ⎠
F' y*
2.0 2.0
Response
1.5 1.5
y (mm)

1.0 F 1.0

Force F (N)
0.5 F Input 0.5
0.0 y 0.0
or

y
-0.5 -0.5
F (N)

-1.0 -1.0
-1.5
ω = 1 rad/s -1.5
-2.0 -2.0
0 1 2 3 4 5 6 7 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
time t (s) displacement y (mm)

ΔW = !∫ F dy = area enclosed by the ellipse > 0 51

Handout P.53-54

Case 3b - Variation with frequency ω

2.5 0.4

2.0
τy 1/τF 1/τy
0.3
|F'| Loss
force F (N)

1.5

tan δ
0.2 tangent
1.0
0.1 plot
0.5
|F"|
0.0 0.0
0.01 0.1 1 10 100 0.01 0.1 1 10 100
frequency ω (rad/s) (log scale) frequency ω (rad /s) (log scale)

The tangent of the phase angle tan δ = F ʹʹ / F ʹ peaks at a

frequency ω = 1/ τ yτ F
• The patterns of response seen here are similar to those of polymers
in the linear viscoelastic regime (but only qualitatively)
• The one systematic discrepancy between the response of the
particular spring-dashpot Zener model and the observed response of
polymers is that it is too localised in the time/frequency domains.
52
Handout P.55

Generalised Maxwell model

A truer representation of the viscoelasticity of
k1 k2 kj a polymer would be given by a spring-dashpot
kR
model with multiple arms, such as
c1 c2 cj N + 1 arms (j = 1..N)
1 2 N N+1
F y

Replacing equation (4.8), the equations of the model are then:

N dy Fj 1 dFj
F = ∑ Fj + FN +1; = + (j = 1..N ); FN +1 = kR y (4.14)
j =1 dt c j k j dt
The j-th relaxation time is now τ j = cj / kj
The relaxation spectrum of the model: ⎡τ j , k j ⎤ (j = 1..N )
⎣ ⎦ 53
Handout P.55-57

Shear relaxation spectrum

The shear “relaxation spectrum” of such a generalised Maxwell
model is given by the set of pairs of:
⎡τ j ,G j ⎤ (j = 1..N )
⎣ ⎦
Example data: shear stress relaxation modulus G(t) for an amorphous
polystyrene at 100°C, near, but just above its Tg~95°C

The discrepancy is very clear ... we need a shear relaxation spectrum

0.30
o
0.25 polystyrene at 100 C
GU Zener
PS model 0.20 13 elements

Gj (GPa)
0.15
0.10
0.05

GR~1MPa 0.00
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4
log10 τj (s)
54
Handout P.56

Shear stress relaxation modulus, G(t)

The equations governing the shear response would then be
(by analogy with equations (4.14) and replacing y by a
shear angle γ (= 2 x tensorial shear strain e) and F by the
corresponding shear stress σ’ and k by shear modulus G):

N dσ ʹj σ ʹj dγ
σ ʹ = ∑ σ ʹj + GRγ where + = Gj (j = 1..N ) (4.15)
j =1 dt τj dt

Applying the model to the case of shear stress relaxation (γ(t) is a step
function at t = 0, of height γ0), the shear stress relaxation modulus is then

by analogy σʹ N ⎛ t ⎞
with eqn G( t ) = = GR + ∑ G j exp ⎜ − ⎟⎟ (4.16)
(4.12) γ0 j =1 ⎜ τj
⎝ ⎠
55
Prony series
Handout P.59

Dynamic shear modulus G*

1.4

Complex shear modulus (GPa)

The shear relaxation spectrum given 1.2 Storage modulus
above for polystyrene at 100°C could 1.0
G'
be used to predict the dynamic shear 0.8
modulus G* governing the oscillatory 0.6
response of polystyrene in shear, via 0.4
the analogue of equations (4.13) 0.2 Loss modulus
G"
0.0

σ* -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
G* = = Gʹ + iGʹʹ log10ω (rad/s)
γ*
by analogy with eqn (4.13)

N G j ω 2τ 2j N G j ωτ j
where Gʹ = GR + ∑ ; Gʹʹ = ∑ (4.21)
j =1 1 + ω 2τ 2j
( ) j =1 1 + ω 2τ 2j
( )
56
Handout P.52 (RECAP)

Case 3: Dynamic mechanical response

A continuous oscillating displacement y * = y 0 exp i ωt is applied

The required oscillating force is F * = F '+ iF " where

⎛ Δkω 2τ y2 ⎞ ⎛ Δkωτ y ⎞ F*
F"
F ʹ = ⎜ kR + 2 2
⎟ y *; F ʹʹ = ⎜ 2 2
⎟y * (4.13) δ
⎜ 1 + ω τ y ⎟⎠ ⎜ 1+ ω τ y ⎟
⎝ ⎝ ⎠
F' y*
2.0 2.0
Response
1.5 1.5
y (mm)

1.0 F 1.0

Force F (N)
0.5 F Input 0.5
0.0 y 0.0
or

y
-0.5 -0.5
F (N)

-1.0 -1.0
-1.5
ω = 1 rad/s -1.5
-2.0 -2.0
0 1 2 3 4 5 6 7 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
time t (s) displacement y (mm)

ΔW = !∫ F dy = area enclosed by the ellipse > 0 57

Handout P.61

Example problem 5

58
Handout P.62

How are viscoelastic data obtained?

• ... spanning such a broad range of times; there is no single technique
capable of achieving this.
• We may exploit the phenomenon of time-temperature equivalence (TTE)

4.2 The role of temperature

• Some of the earliest data for an uncrosslinked
Glassy sample of the amorphous polymer
<< Tg polyisobutylene (PIB): H CH3
• Tg ~ -20°C * C C *
n
H CH3
• Tensile stress relaxation modulus E(t) (in
dynes/cm2 = 0.1 Pa) vs time for a wide range
of T (log scales on both axes)
Rubbery • The data exhibit time-temperature
~Tg superposition: with suitable horizontal shifts,
they all superimpose to form a single “master
Elasto-viscous fluid curve”
59
>> Tg
Handout P.63

Time-temperature superposition (TTS)

The master curve for PIB obtained by time-temperature superposition
of these curves (in previous slide), onto the curve for 25°C (reference
temperature T0).

The factor by which the time

Glassy scale for a particular curve has
<< Tg to be reduced/expanded to
give a fit to the reference
temperature curve (here T0
=25°C) is these days called aT
Rubbery
(but called K298 in the graph)
~Tg

Elasto-viscous
aT fluid >> Tg
60
Handout P.64

Linking E(t) to G(t)

• In and above the glass transition Tg region, K >> G
• As such the Poisson’s ratio is ν ~ 0.5 (i.e. incompressible)
• and hence E = 2G(1 + ν ) ≈ 3G
• It follows that under these conditions, the tensile relaxation
modulus function E(t) is related simply to the shear relaxation
modulus function G(t):

E(t ) ≈ 3G(t )
61
Handout P.64

Time-temperature equivalence (TTE)

• The observation of time-temperature superposition (TTS) in the
E(t) data shows, to a good approximation that,
• ...not only E(t) but also the shear relaxation modulus G(t) at any
temperature T is related to that at a reference temperature T0 by a
simple expansion or contraction of the time-scale.

In terms of equations, this means: (4.23)

⎛ t ⎞
G ( t ,T ) = G ⎜ ,T0 ⎟ or G (log t,T ) = G (log t − log aT ,T0 )
⎝ aT ⎠
Expansion / contraction factor Reference temperature

Experimental data further suggest that the same aT function also applies
to all τj :
Relaxation time
at any temp T τ j (T ) = aT τ j (T0 ) for all j (4.24)
62
Handout P.65

Vogel-Tamman-Fulcher (VTF) equation

35 This graph shows lnaT plotted versus
30 calculated
1/T (absolute). Two points are notable:
25 experiment
(a) The data do not give a straight line
20
PIB relation (i.e. do not fit the
lnaT

15 Arrhenius relation, when T > Tg).

10
5 (b) The data are fitted by the alternative
0 2-parameter calculated line included
3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 in the graph.
1000/T
(4.25)
e.g. the best-fit values of
material parameters in PIB C C
are: ln aT = −
C = 1446 K ; T∞ = −120.5°C (T − T∞ ) (T0 − T∞ )
Reference temperature 63
Handout P.67

Glass transition temperature Tg

Recall: the glass transition temperature Tg of
an amorphous material is the temperature Cooling rates
below which it cannot maintain structural q1 > q2
equilibrium, on a given experimental time-scale

• As shown in plot, faster cooling

leads to a higher measured Tg.

• Because glass is not in

thermodynamic equilibrium, its
structure gradually evolves with time
(e.g. polymer at BàC).
• ‘Fictive temperature’ Tf =
temperature at which the glass is
believed to have the same structure
in equilibrium
Equilibrium
line
The value of Tg is defined in
terms of the volume V or
enthalpy H measured during
heating or cooling, and its
values are rate dependent
64
Handout P.68

Physical ageing in glassy polymers

• The viscoelastic properties also vary with time in the glassy state (<Tg)

Shown here are tensile creep

curves for PVC (Tg = 83°C)
obtained after rapid cooling
from 90°C to 20°C, and
waiting (‘ageing’) for varying
times te before testing.

• Note how the creep curves shift to the right with increasing ageing time
• ...there appears to be time – ageing-time (te) superposition going on.

τ j ( te ) = aeτ j ( te0 ) for all j (4.27)

Each relaxation time after ageing for a time te is related by the

same factor ae to its corresponding value after ageing for a time te0 65
Handout P.69

Secondary relaxations of polymers

• The viscoelastic process visible in the glass transition (Tg) region
is often labelled the ‘primary relaxation’ of the polymer.
• But viscoelasticity is also prominent in other regions of the
temperature scale. These are called the ‘secondary relaxations’

They differ from the primary relaxation in that:

(i) they are not usually associated with evidence of ‘physical ageing’
(i.e. not linked to microstructural evolution).
(ii) They are usually found, to reasonable accuracy, to show time-
temperature superposition, with the relaxation times conforming to
the Arrhenius relation as follows

⎛c⎞ c c
τ ∝ exp ⎜ ⎟ hence ln aT = − (4.28)
⎝T ⎠ T T0
Δ H0 Activation energy (also denoted as Q)
where c =
R 66
Handout P.70

Example problem 6

67
Handout P.71

Non-linear viscoelasticity
Exp. data from tensile • With increase in stress, the
creep and shear creep viscoelasticity shown by a polymer
tests on PMMA at 30°C becomes increasingly non-linear
• i.e. the relation between stress
and strain is time-dependent, but
the material ceases to behave as
a linear system.
• It no longer obeys the Boltzmann
Superposition Principle
• Many of the questions of
(i) how to model them in the
macro-world, and
(ii) how to explain them in terms of
what is going on in the micro-
world form part of the ongoing
research activities... (DPhil)
68