Methods of Improving Constitutive Equations

Maxwell Model
We can improve with new time derivatives or new strain measures.

W O

wW wt

K0 J
t t ) K  ( t  O  0 e J (t , t )dt 2 f O

W (t )

We can also change the basic equation: linear modifications non-linear modifications
Faith A. Morrison, Michigan Tech U.

Other Constitutive Approaches

Simple Maxwell Model, shear Upper-Convected Maxwell Model, general

W 21  O

wW 21 wt

K0J 21

W  OW

K0J
retardation time

Simple Jeffreys Model, shear Upper-Convected Jeffreys Model, general (Oldroyd B Fluid)

W 21  O1

wW 21 wt

wJ K0 J 21  O2 21 wt

W  O1W

K 0 J  O 2 J

Faith A. Morrison, Michigan Tech U.

Maxwell Model - Mechanical Analog

W 21  O wW 21 wt K0J 21

Jeffreys Model - Mechanical Analog

W 21  O1 wW 21 wt wJ K0 J 21  O2 21 wt

Faith A. Morrison, Michigan Tech U.

Unfortunately, this change only modifies G(t-t); the Jeffreys Model is a GLVE model Simple Jeffreys Model
(not frame-invariant)

W  O1

wW wt

wJ K0 J  O2 w t

Now, solving for W21 explicitly we obtain,

W (t )

t c K0 O2  tO 2K 0 O2 1 c     1 e G ( t t ) J (t c) dt c O1 O1 O1  f
t

G (t  t c)
Other linear modifications of the Maxwell model motivated by springs and dashpots in series and parallel modify G(t-t) but do not otherwise introduce new behavior.

(Might as well use the Generalized Maxwell model)

Faith A. Morrison, Michigan Tech U.

Non-linear modifications of the Maxwell Model

White-Metzner Model Oldroyd 8-Constant Model

W

K (J ) W G0

K (J )J

W  O1W 

1 O1  P1  J W  W J  1 P 0 trW J  1 Q1 W : J I 2 2 2 1 K0 J  O2 J  O2  P 2  J : J  Q 2 J : J I 2

The Oldroyd 8-constant contains many other constitutive equations as special cases.

UCM UCM 
terms

UCJ

Faith A. Morrison, Michigan Tech U.

The Oldroyd 8-Constant model contains all terms linear in stress tensor and at most quadratic in rate-of-deformation tensor that are also consistent with frame invariance.
W  O1W  1 O1  P1  J W  W J  1 P 0 trW J  1 Q1 W : J I 2 2 2 1 K0 J  O2 J  O2  P 2  J : J  Q 2 J : J I 2

Giesekus Model

W  OW 

DO W :W K0
quadratic in stress

K0 J

The only way to choose among these nonlinear models is to compare predictions.
Faith A. Morrison, Michigan Tech U.

We can also modify integral models to add non-linearity and thus produce new constitutive equations.
Factorized Rivlin-Sawyers Model

W (t )

 M (t  t c) ) 2 ( I1 , I 2 )C  )1 ( I1 , I 2 )C
f

1

dtc

Factorized K-BKZ Model

W (t )
I1, I2 are the invariants of the Finger or Cauchy strain tensors (these are related).

t wU wU 1  M (t  t c) dt c 2 wI C  2 wI C 2 1 f

Again, the only way to choose among these nonlinear models is to compare predictions
(see R. G. Larson, Constitutive Equations for Polymer Melts). Faith A. Morrison, Michigan Tech U.

Choosing Constitutive Equations

We have fixed all the obvious flaws in our constitutive equations, and now we have too many choices! We could make predictions and compare with experimental data, but some of the models (Rivlin Sawyer, K-BKZ) have undefined functions that must be specified.

How to proceed?

We need some guidance.

All along we have taken a continuum-mechanics approach. We have run that course all the way through. Now we must go back and seek some insight from molecular ideas of relaxation and polymer dynamics.

Faith A. Morrison, Michigan Tech U.

Molecular Approach to Polymer Constitutive Modeling

molecular tension force on arbitrary surface

~ f

W  dA n

stress tensor

We now attempt to calculate molecular forces by considering molecular models.

Polymer Dynamics

end-to-end vector, R

polymers may be modeled as random walks.

Faith A. Morrison, Michigan Tech U.

Polymer coil responds to deformation

A polymer chain adopts the most random configuration at equilibrium.

end-to-end vector, R

When deformed, the chain tries to recover that most random configuration, giving rise to a spring-like restoring force.

spring of equilibrium length and orientation R

We will model the chain dynamics with a random walk.

Faith A. Morrison, Michigan Tech U.

Gaussian Springs
Equilibrium configuration distribution function - probability a walk has end-to-end distance R

E  E 2 RcRc \ 0 ( R) e S

From an entropy calculation on a random walk we can calculate the force needed to deform a Gaussian spring If we can relate this force to the arbitrary force on a surface, we can connect these two

3kT R Na 2

molecular tension force on arbitrary surface

~ f = dA n

stress tensor

Faith A. Morrison, Michigan Tech U.

Molecular force generated by deforming chain

~ f

Tension force on dA

Force on surface dA due to chains of ETE R

dR1dR2 dR3

Probability chain of ETE R crosses surface dA

Probability chain has ETE R

Force exerted by chain w/ ETE R

R Q 3 n
(see text)

dA Q

2 3

\ ( R )dR1dR2 dR3

3kT R Na 2

Faith A. Morrison, Michigan Tech U.

Probability chain of ETE R crosses surface dA

dA
a
1

n
c
Q1 3

intersection with dA

a
1

b
R n
n

d
1

13

Q1 3

Probability chain of ETE R crosses surface dA

R Q n Q 1

1 3

c
Faith A. Morrison, Michigan Tech U.

volume Q 1

Molecular force generated by deforming chain

~ f

3kTQ Na 2

1 3

RR

R R { R R \ ( R )dR1dR2 dR3
BUT, from before . . .

~ f
Comparing these two we conclude,

W  dA n

molecular tension force on arbitrary surface in terms of

W

2 3

3kTQ RR Na 2

(dA Q

Molecular force generated by deforming chain

Faith A. Morrison, Michigan Tech U.

How can we convert this equation,

3kTQ RR Na 2

Molecular force generated by deforming chain

which relates molecular ETE vector and stress, into a constitutive equation, which relates stress and deformation? We need a idea that connects ETE vector motion to macroscopic deformation of a polymer network or melt.

Faith A. Morrison, Michigan Tech U.

Elastic (Crosslinked) Solid

Between every two crosslinks there is a polymer strand that follows a random walk of N steps of length a.

R2
Distribution of ETE vectors

ETE = end-to-end vector

R1
Faith A. Morrison, Michigan Tech U.

How can we relate changes in end-to-end vector to macroscopic deformation?

ANSWER:

affine-motion assumption: the macroscopic dimension changes are proportional to the microscopic dimension changes after

before

Faith A. Morrison, Michigan Tech U.

Consider a general elongational deformation:

O1 0 0 O2 0 0 0 0 O3 123

1

For affine motion we can relate the components of the initial and final ETE vectors as, ETE after

O1

R1 R1c

O2

R2 c R2

O3

R3 c R3

ETE before

O1 R1c c R (t ) O2 R2 O Rc 3 3 123
Faith A. Morrison, Michigan Tech U.

We are attempting to calculate the stress tensor with this equation:

3kTQ RR Na 2

R R { R R \ ( R )dR1dR2 dR3

O1 R1c c R (t ) O2 R2 O Rc 3 3 123

But, where do we get this?

Faith A. Morrison, Michigan Tech U.

Probability chain has ETE between R and R+dR:

\ ( R )dR1dR2 dR3

Configuration distribution function Equilibrium configuration distribution function:

E  E 2 RcRc \ 0 ( R) e S

But, if the deformation is affine, then the number of ETE vectors between R and R+dR at time t is equal to the number of vectors with ETE between R and R+dR at t

E  E 2 RcRc Conclusion: \ ( R ) \ 0 ( R ) e S c
Faith A. Morrison, Michigan Tech U.

10

Now we are ready to calculate the stress tensor.

3kTQ RR Na 2
Ric Ri Oi

O1 R1c c R (t ) O2 R2 O Rc 3 3 123

R R { R R \ ( R )dR1dR2 dR3
c
3

E  E 2 RcRc \ ( R) \ 0 ( R ) e S

Final solution: W

i e i Q kTOi2 e
Faith A. Morrison, Michigan Tech U.

Final solution for stress:

i e i Q kTOi2 e

Compare this solution with the Finger Strain Tensor for this flow. 2 O1 0 0 1 1 T 1 C (t c, t ) F F 0 0 O2 2 0 0 O2 3

123

Since the Finger tensor for any deformation may be written in diagonal form (symmetric tensor) our derivation is valid for all deformations.

Q kT C

1

Which is the same as the finite-strain Hookes law discussed earlier, with G=QkT.
Faith A. Morrison, Michigan Tech U.

11