Rubber elasticity

Marc R. Roussel
Department of Chemistry and Biochemistry
University of Lethbridge
February 21, 2009

A rubber is a material that can undergo large deformations (e.g. stretching to five or ten
times its original length) and then return to its original shape and size. Many polymeric
materials, and not just rubber itself, display rubber elasticity in some range of temperatures.
When we characterize these materials, we find that they are made of long chains, with
occasional cross-links between the chains. In the resting state, these chains will adopt random
coil configurations. As we apply a stress by pulling on a sample of rubber, the chains will
tend to align along the stress, which allows the material to stretch. The cross-links pull the
material back to its original shape once the stress is removed.
In this note, we will look at a statistical treatment of rubber elasticity. This will allow
us to relate the elastic behavior to properties of the polymer.

1 Cross-linking statistics
Figure 1 shows two primary polymer molecules which have formed some cross links. There
are both internal cross-links (within a single molecule) and external cross-links (between
molecules). Both types can be important to rubber elasticity. The important statistic for
rubber elasticity is in fact the number of chains bounded by two cross-link junctions, a
quantity called the number of active chains, denoted e . Free ends are uninteresting
because they can be displaced more-or-less freely. If we have N primary polymers, there
will be 2N free ends. Let be the number of monomers involved in cross-links (two per
cross-link). Each cross-link takes two chains and divides them into four, so the number of
chains added by making /2 cross-links is equal to . From this number, we have to subtract
the number of chains that end freely. Thus,

e = 2N = (1 2N/) .

Now let M be the average molar mass of the primary polymers, and Mc be the molar mass
per cross-linked monomer, i.e. Mc = N M/. Another way to look at Mc is that its the
average molar mass of the active chains. We can then rewrite the last equation

e = (1 2Mc /M ) . (1)

1
Figure 1: Two primary polymer molecules (i.e. two polymer chains) which have formed some
cross-links, shown as bold dots. The chains are colored differently for ease of visualization.

2 Thermodynamics of elasticity
When we write down the differential of U in courses in chemical thermodynamics, we nor-
mally only consider pressure-volume work. Here, we need to also consider the extension
work. Suppose that f is the externally imposed (equilibrium, i.e. reversible) tension on an
elastic body. Then
dw = P dV + f dL,
where dL represents a change in the length of the body. You will recognize the new term as
a special case of the general definition of mechanical work as force times distance. Then we
have
dU = dw + dq = P dV + f dL + T dS.
In this form, we have a differential of U in terms of the variables (S, V, L). Physically, these
are not the most convenient variables. We would like to rewrite U as a function of (T, V, L).
To do this, we need to rewrite the differential of S in terms of those variables:
  
S  S  S 
dS = dT + dV + dL.
T V,L V T,L L T,V

If we now substitute this equation into the differential of U , we get

 !  ! 
S  S  S 
dU = T P dV + T + f dL + T dT.
V T,L L T,V T V,L
 
U  S 
= T + f,
L T,V L T,V
 
U  S 
or f = T . (2)
L T,V L T,V

2
If we can develop a theory for the thermodynamic properties of rubbers, then equation 2
will give us an equation for the force which must be applied to get a certain extension L
from the equilibrium length L.
The elasticity of rubbers is largely due to conformational rearrangements of the chains,
which are obtained largely by rotations around bonds of the framework. These rotations
should take very little energy. If we assume that they take no energy at all, then we get the
ideal rubber model. In terms of the quantities in equation 2, an ideal rubber is one for
which 
U 
= 0.
L T,V
The elastic force is then due to an entropic effect alone:

S 
f = T . (3)
L T,V

It is worthwhile spending a few minutes thinking about what equation 3 tells us in relation
to rubber elasticity. When we pull on a piece of rubber, we straighten out the polymer chains.
This decreases the entropy, so S/L|T,V < 0. Another way to put this is that there are
more microstates associated with the collapsed state than with the stretched state. The
random thermal motions of the chains in a stretched state are therefore more likely to be
in a direction which would tend to bring the polymer back toward its relaxed state, which
generates the resistive force f . This is remarkable: Entropy, which we normally think of as
a somewhat abstract statistical quantity, is directly responsible for a macroscopic force!

3 Stress and strain

In materials science, we dont normally talk about the force vs extension relationship, which
is what equation 3 would give us. There are two problems:

1. The force depends on the geometry of the sample, in particular its cross-sectional area
perpendicular to the applied force. To solve this problem, we define the stress

= F/A.

2. The actual extension (L) for a given stress depends on the initial length of the sample,
L0 . On the other hand, the strain

= L/L0 = (L L0 )/L0

only depends on the stress. Equivalently, we can use the stretch ratio

= L/L0 = + 1.

We will use in these notes, although clearly and are interchangeable quantities.

3
Materials scientists thus usually discuss deformation using the stress-strain relationship.
(The term stress-strain relationship is used even when the strain is described in terms of the
stretch ratio.)
From the definition of , we have L = L0 . Therefore
1
= .
L L0
Equation 3 therefore becomes 
T S 
f = . (4)
L0 T,V

4 Statistical theory of rubber elasticity

To calculate the entropy of a rubber, we will use the Boltzmann equation:

S = k ln W,

where W is the number of microstates. The active chains may have different lengths, which
affects W . We can compute W by
Y
W = (Wn )n ,
n

where n is the number of chains containing n bonds. This gives

X
S=k n ln Wn .
n

We need to reinterpret this equation a little: In our case, we cant count the microstates.
Suppose that I give you the probability density pn () that a given n-mer has end-to-end is
stretched by a factor from its resting length. The number of microstates between and
+d is proportional to pn () d. Since we are going to take a logarithm, the proportionality
constant, as well as the factor of d, end up in an additive constants:
X
S=k (n ln pn () + constant) .
n

Our next step will be to take the derivative in equation 4, so the constant will vanish:

kT X ln pn () 
f = n  . (5)
L0 n T,V

Cross-links occur where two primary polymers have come sufficiently close together in
the right orientation. Cross-linking should therefore have only a small effect on the locations
of the monomers which became cross-linked. The distribution of relative positions of the two

4
cross-link junctions defining an active chain should therefore obey the Gaussian probability
density derived in the last lecture:
2 2 2 2
pn (x, y, z) = (q/ )3 eq (x +y +z ) , (6)

where q 2 = 3/(2n`2 ). We also had hr02 i = n`2 = 3/(2q 2 ), where hr02 i is the mean squared
end-to-end distance of the relaxed polymer.
We now introduce an affine transformation: We assume that the deformation of a
sample of rubber stretches all the coordinates homogeneously by factors x , y and z in
each of the three Cartesian directions. Thus, (x, y, z) = (x x0 , y y0 , z z0 ), where (x0 , y0 , z0 )
are the initial relative coordinates of the end of a chain. Taking a logarithm of equation 6
and introducing the affine transformation, we get

ln pn = q 2 (x x0 )2 + (y y0 )2 + (z z0 )2 + constant,
 

where again we wont be too concerned with the constants since they wont contribute to
the force.
Rubbers are essentially incompressible, so we must have x y z = 1. Suppose that we
z .1
stretch a sample of rubber along the x axis. Let x = . We will typically find that y =
Substituting this relationship into the incompressibility condition, we get y = z = 1/ .
This gives
ln pn () = q 2 2 x20 + (y02 + z02 )/ + constant.
 

There is just one more thing: If we have n polymers of length n, then we should average
this quantity over all the initial end-to-end distances:

ln pn () = q 2 2 hx20 i + (hy02 i + hz02 i)/ + constant.

 

For the relaxed polymer, there is nothing special about the x axis, i.e. hx20 i = hy02 i = hz02 i.
Since hr02 i = hx20 i + hy02 i + hz02 i, this gives hx20 i = hy02 i = hz02 i = hr02 i/3. We therefore obtain

q 2 hr02 i 2 1
+ 2/ + constant = 2 + 2/ + constant.



ln pn () =
3 2
Note that q and hr02 i have both dropped out of the non-constant part of this expression. The
derivative in equation 5 therefore doesnt depend on n:

ln pn () 
= 1/2 .


L 
T,V
P
It can therefore be pulled out of the sum, leaving n n = e , the number of active chains.
Putting it all together, we get the elastic force
kT e
1/2 .


f=
L0
1
Some polymers do deform differently in different directions, but then we would have to orient the sample
correctly with respect to any special axes to observe this phenomenon.

5
To get the strain, we divide this equation by the cross-sectional area. There are actually two
slightly different conventions in use here. One is to divide by the cross-sectional area of the
relaxed sample, A0 , giving
kT e
1/2 .


0 = (7)
V0
0 is called the nominal stress. This is the convention normally used by polymer scientists.
The other is to divide by the area corresponding
to stretch ratio . Since the y and z are
transformed by the ratio y = z = 1/ , A = A0 /. The true stress is therefore

kT e 2

= 1/ .
V0
We will pursue our analysis with equation 7. Using equation 1, the stress becomes
kT
(1 2Mc /M ) 1/2 .


0 =
V0
If we define the number of moles of cross-linked monomers nc = /NA , we get

nc RT
(1 2Mc /M ) 1/2 .


0 =
V0
Almost the entire volume is occupied by active chains. The mass of the sample is therefore
approximately nc Mc . If we denote by v the specific volume of the relaxed sample, then
V0 = nc Mc v, and therefore
 
RT 1 2
1/2 .


0 =
v Mc M

This is our final expression. It predicts two interesting properties of the stress-strain rela-
tionship:

1. For a given strain, the stress increases linearly with temperature. Alternatively, if we
fix the stress (e.g. by suspending a mass from a piece of rubber), then the extension
should decrease as we increase the temperature. This is a well known property of
rubber which the statistical theory correctly predicts. Note that this behavior is the
opposite of what we observe for normal materials: Typically, heating something up
causes it to expand. Putting a stress on normal materials (e.g. metals) as they are
being heated does not change this behavior.

2. Increasing the number of cross-links, keeping all other properties constant, decreases
Mc , which in turn increases the stress at fixed strain. This is probably intuitively
obvious, but its nice that the statistical theory actually predicts this behavior correctly.