Sie sind auf Seite 1von 25

Contemporary Physics

ISSN: 0010-7514 (Print) 1366-5812 (Online) Journal homepage: https://www.tandfonline.com/loi/tcph20

Rubber elasticity

L. R. G. Treloar

To cite this article: L. R. G. Treloar (1971) Rubber elasticity, Contemporary Physics, 12:1, 33-56,
DOI: 10.1080/00107517108205104

To link to this article: https://doi.org/10.1080/00107517108205104

Published online: 20 Aug 2006.

Submit your article to this journal

Article views: 109

Citing articles: 8 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tcph20
CONTEMP. PHYS., 1971, VOL. 12, NO. 1, 33-56

Rubber Elasticity
L. R. G. TRELOAR
Department of Polymer and Fibre Science
University of ManChester Institute of Science and Technology
SUMMARY. The elasticity of rubber arises not from the forces between molecules in
the ordinary sense, but from the thermal energy of the constituent atoms in the
long-chain molecule. It is associated with a reduction of entropy on deformation
rather than with an increase of internal energy, such as occurs in an ordinary solid.
This mechanism explains the unusual thermoelastic properties of rubber, originally
studied by Gough and Joule.
The basic concepts of this kinetic theory of rubber elasticity are explained, a s well
as the subsequent developments which have taken place in the quantitative treatment
of the mechanical properties of rubber under the most general type of strain
Attention is drawn to the unusual ‘ normal-stress ’ effects which occur only in materials
which are capable of undergoing large elastic deformations. A modification of the
basic theory which enables the photoelastic properties of a rubber-like material to be
interpreted in terms of the polarizability of the long-chain molecule is examined and
shown t o yield information on the statistical form of the molecule.
Finally, consideration is given to recent developments in the thermodynamics of
elasticity, and to the problem of interpreting the small but significant changes in
internal energy which are found to occur on deformation of an actual rubber.

1. The problem of rubber elasticity


Considered as materials, the rubbers are of peculiar interest, particularly to
the physicist. There are two main reasons for this. The first is that they
possess a number of remarkable physical properties, of which the most
important is their very high elastic deformability, which distinguishes them
from most of the other materials which are normally regarded as solids. The
second is that, among the materials of high-molecular constitution, they are
structurally the simplest, with the result that their main physical properties
are capable of quantitative interpretation in terms of their molecular structure
to an extent which is not yet possible for other types of polymeric materials.
In approaching the problem of rubber elasticity many, if not most, of the
ingrained habits of thought which the physicist has acquired during the course
of his training are either unhelpful or positively disadvantageous. To give
some examples:
( 1 ) The basic concept of the classical theory of elasticity, Hooke’s law, does
not apply even theoretically to a rubber, though the material may be perfectly
elastic.
(2) The deformation of a rubber under constant temperature conditions
takes place in such a way that its internal energy is substantially unchanged in
the process.
(3) I n large-deformation theory a shear stress does not produce a simple
shear strain, and vice versa.
These, and many other related properties, indicate that, in addition to its
practical importance, the problem of rubber elasticity is not lacking in
intellectual fascination.
Nevertheless, it would be true to say that, broadly speaking, the problem of
the elasticity of rubber has attracted comparatively little attention from the
C.P. C
34 L . R. G . Treloar

physicist, and particularly from the more academic type of physicist. Possibly
one reason for this is that the subject is one which draws its inspiration from
both chemical and physical sources, and in which, therefore, the physics
content cannot be entirely isolated from the chemical background. However
that may be, there is no doubt that, as the emphasis on the utilization of rubbers
in the engineering field strengthens, the further development of the subject
will rely heavily on the services of the physicist in both the experimental and
theoretical fields.
The aim of this review is t o set out briefly the key problems in the physics
of rubber elasticity, to examine the theoretical concepts involved in the
elucidation of these problems, and t o indicate the extent t o which the various
physical properties may be quantitatively accounted for in terms of a relatively
simple model of the molecular structure of rubberlike materials.

2. Elastic and thenno-elastic effects


During the whole of the nineteenth century natural rubber was regarded
as a very peculiar material, with properties quite different from those of any
other known substance. Although today we have a considerable number of
synthetic rubbers, it is still true that the rubbers, as a class, possess properties
of peculiar scientific interest.
Fig. 1 shows the stress-strain curve of a typical vulcanized rubber. This
is markedly non-linear: i t cannot be represented in terms of a constant value of
Young's modulus, except for quite small strains. I n this small-strain region a
typical value of Young's modulus is about 1 MN m-2 (10' dyn cm-2). Such
a value is lower than that for an ordinary solid (e.g. a metal) by a factor of
the order lo6. The corresponding elastic extensibility, typically in the range

EXTENSION, "10

Fig. 1. Typical stress-strain curve for vulcanized rubber.


Rubber Elasticity 35

500 to 1000 per cent, is higher than that for a normal solid by a factor of about
1000.
It was discovered by Gough (1805) that a strip of rubber extended by a
constant force contracted on raising the temperature. Alternatively, if held
at a constant stretched length, the tensile force increased on heating. This
effect was later coniirmed by Joule (1859), and is known as the Gough-Joule
effect. Another effect discovered by Gough, and also confirmed by Joule, was
that the adiabatic extension of a rubber strip was accompanied by a rise in
temperature, while an adiabatic retraction produced cooling. These two
effects-the reversible heat of extension and the contraction on heating-have
important thermodynamic implications, and were in fact shown by Kelvin to
be thermodynamically related to each other.
These peculiar thermo-elastic effects, like the purely mechanical properties
of rubber, clearly cannot be accounted for in terms of interatomic forces of the
type responsible for the mechanical properties of an ordinary crystalline solid.
In such a material the individual atoms are held in fixed mean positions by
well-defined interatomic forces. These forces vary rapidly with the distance
between atomic centres. Application of a stress changes the interatomic
distances in such a way that the internal forces are brought into equilibrium
with the external stress. Theoretically the maximum deformation which
this type of structure can withstand is about 20 or 30 per cent; in practice it
is very much lower.

3. The statistical theory of rubber elasticity


Basic mechanism
The solution to the problem of rubber elasticity, and the interpretation of
the Gough-Joule effect, had to await the gradual development of chemical
methods for dealing with structures of very high molecular weight. Although
Faraday (1826) had shown more than a century earlier that rubber is a simple
hydrocarbon, with an empirical formula corresponding to (C5H&, it was not
until about 1930 that the true nature of the rubber molecule, and in particular
its very great length, became satisfactorily established. It was shown that
the rubber molecule was built up by the successive addition of a single
' repeating unit '-the isoprene unit:
4Ha-C=CH-CH2-
I
CH3
to form a very long chain. A typical chain would contain about 5000 of such
units, corresponding to a molecular weight in the neighbourhood of 350 000.
It soon came to be realized that a long-chain molecule of this kind will not
be rigid, like a stiff rod, but will have a considerable degree of flexibility, owing
to the relative ease with which any given C-C bond can rotate about neigh-
bowing bonds in the chain. It was Meyer who first clearly recognized the
significance of this property for the interpretation of rubber elasticity (Meyer
et al. 1932). The principle is illustrated in fig. 2, which represents a simple
paraffin or polyethylene chain (CH,),. Each bond is assumed to be capable
of free rotation about the preceding bond, thus leading to a randomly-kinked
form for the chain, in which the distance between the ends will, on the average,
36 L. R.G. Tretoar

be very much less than the fully outstretched length (fig. 2 ( c ) ) . If such a
molecule is forcibly extended and then released it will in the course of time
return t o some more highly-kinked state such as that shown in fig. 2 (b) under
the influence of the random bombardments of surrounding molecules. The
molecule is thus inherently elastic, but its elasticity is kinetic, not static, and
depends upon the statistical configurations of the chain.

\
I
I
I
I
I
I I
I
I
I
I
I
I

(b)
z (c)

Fig. 2. (a)Rotation about bonds in chain. ( b ) Resultant random conformation.


(c) Extended zig-zag structure.

Some idea of the statistical form of the long-chain molecule may be gained
from the model represented in fig. 3. This represents a paraffin chain con-
taining a thousand C-C bonds, constructed on the basis of random rotation
about successive bonds.

Thermodynamic consequences
The contrast between the basic concept of rubber elasticity and the concept
of intermolecular force fields could hardly be more absolute. The elasticity
of a rubber is due not t o any attractive forces between molecules in the
ordinary sense of the term, but t o the thermal motion of the parts of the
long-chain molecules themselves-sometimes called the micro-Brownian motion.
I n many respects-and particularly in its thermodynamic aspects-the
elasticity of rubber is more akin to that of a gas than to that of an ordinary
solid. The extended molecule retracts to a less extended length not because
the shorter length has a lower energy, but because the shorter length has a
higher probability, or in thermodynamic terms, a higher entropy.
Just as in the case of a perfect gas we make the assumption that the internal
energy of the system (ie. the kinetic energy of the molecules) depends only on
Rubber Elasticity 37

Fig. 3. Model of polyethylene chain containing 1000 C-C bonds.

the temperature, and not on the volume, so, by analogy, we may make the
assumption that the internal energy of a rubber is a function only of the
temperature, and not of the state of strain (e.g. length). This assumption
leads directly to the following important conclusions:
1. The tensile force in a rubber held at constant stretched length is propor-
tional to the absolute temperature.
2. The heat evolved in an isothermal extension is equal to the mechanical
work done by the stretching force.
The statistical theory thus provides an immediate explanation of the long-
established thermo-elastic effects discovered by Gough and confirmed by Joule.

1.75

1.50-

1.25 -

1.00 ' I I I I I

Fig. 4. Effect of temperature on tensile force on stretched rubber. (Meyer and


Ferri 1936.)
38 L. R.G. Treloar

An example of the variation of the tensile force at a constant length


corresponding to an extension of 350% with temperature, taken from
the work of Meyer and Ferri (1935), is reproduced in fig. 4. A t this extension
the expected theoretical relationship is approximately satisfied.
More recent work, which will be referred to later, has been concerned with
certain rather small but significant deviations from strict proportionality
between tensile force and temperature. These deviations, however, do not
substantially affect the general conclusions to be drawn from the original
experimental data of Meyer and Ferri.

4. Quantitative development
Large and small strains
The classical theory of elasticity is essentially a small-strain theory. It
provides a complete description of the mechanical properties of an elastic
solid, in the region of small strains, in terms of two independent physical
constants. Any homogeneous state of strain can be expressed as the sum of
a change of volume (dilatation) together with two (or three) shear strains.
The resultant stress is derived on the basis that each component of strain
produces a corresponding component of stress, and that these stresses are
additive.
In rubbers, the strains are large. Under these conditions not only are the
stress-strain relations for particular types of strain non-linear, but, more
importantly, individual stresses and strains can no longer be simply super-
imposed. This second consideration arises from the fact that the properties
of the material are actually changed by the deformation, and that in the
strained state it is no longer isotropic, in contrast to the classical elastic solid,
whose properties are assumed to be isotropic and independent of strain.
This more complex strain dependence necessitates a quite different approach
to the method of representation of the elastic behaviour. The statistical
theory has been remarkably successful in providing a basis for such a new
approach.

Chain statistics
In any real molecule, the geometry of the chain, i.e. the succession of bond
lengths and valence angles, is a function of its chemical constitution, and hence
varies from one type of molecule t o another. The statistical properties of the

Fig. 5. Function representing the probability of a given end-to-end distance for


long-chain molecule.
Rubber Elasticity 39

chain, however, are not very sensitive to the details of the chain structure,
and may be expressed in terms of a general function representing the proba-
bility P(r)dr that the chain ends shall be separated by a distance lying between
r and r + dr. This function has the form shown in fig. 5, and is represented by
the equation
4b3
P(r)dr = -r2 exp [ - b2r2]dr
7T42

in which the parameter b depends on the detailed chain geometry and also on
the chain length. The root-mean-square (r.m.s.) value of r , which is a
convenient average statistical parameter to employ, is given by
-
(r2)l/2= (Q)’l2/b.
For the hypothetical case of a randomly-jointed chain of n links each of length
1, which is frequently used for mathematical convenience,

For any other chain geometry the parameter b and the r.m.s. length differ
only by a numerical factor from the values (3). The important conclusion
is that the r.m.s. length is proportional to the square root of the number of
links in the chain, i.e. to the square root of the chain length itself.
The first step in the application of the statistical treatment of the single
chain to the problem of the mechanical properties of a rubber is the derivation
of the entropy of the chain as a function of the distance between its ends.
By following the standard processes of statistical thermodynamics, according
to which the entropy associated with a given state is proportional to the
logarithm of its probability, the entropy s for the chain may be shown to be
of the form
s = c - kb2r2, (4)
where k is Boltzmann’s constant and c is an arbitrary constant. The entropy
is thus a maximum when the two ends of the chain are coincident ( r = 0) and
diminishes progressively with increasing extension.

Network properties
I n order to apply this result for the entropy of the single chain to the problem
of the elasticity of rubber it is necessary to introduce certain assumptions
concerning the structure of the rubber and the relation between the deformation
of the molecules and the strain in the bulk material. The first of such assump-
tions is that the forces between the chains are negligibly small. This means
that on deformation of the material no work is done against inter-molecular
forces, the only change being a change in the entropy of the individual chains.
However, if there were no forces between molecules the system would behave
as a liquid; there would be no restriction on the slippage of one molecule past
another. To overcome this difficulty the further assumption is made that a
small number of permanent connections or cross-linkages are introduced
between the chains. Since the molecules are very long the number of such
cross-linkages required may be quite small, and these will therefore not
seriously interfere with the freedom of motion of the chains (the micro-Brownian
motion) except in the immediate vicinity of the cross-linkages.
40 L. R. G. Trebar

This concept of cross-linking is basic t o the theory of rubber elasticity.


I n practice these cross-linkages are introduced in the process of vulcanization,
originally discovered by Charles Goodyear in 1839. This process, which
consists essentially of a chemical reaction with sulphur, greatly enhances the
elastic properties of the original raw rubber, and reduces irreversible creep and
flow effects. By cross-linking, the original assembly of individually separate
chains (fig. 6 ( a ) )is converted into a coherent 3-dimensional network structure
(fig. 6 ( b ) ) which will revert t o its original state after deformation.

Fig. 6. Model of rubber (a) before cross-linking, ( b ) after cross-linking, ( c ) on


subsequently straining.

Afine deformation of chains


Let us now consider a network, such as that represented in fig. 6 ( b ) ,t o be
deformed, as in fig. 6 ( c ) . We define the ' chain ' in the network structure as
the segment of the molecule between successive points of cross-linkage. When
the structure is deformed, the individual chains are likewise deformed in the
sense that their end-to-end distances (defined by the points of cross-linkage)
are changed. The problem is to calculate the changes in the entropy of the
individual chains, and hence by summation the change in the total entropy of
the system, corresponding to a specified deformation applied to the boundaries
of the network. We thus have the problem of relating the molecular deforma-
tion to the deformation of the material in bulk.
The mathematical examination of this problem (James and Guth 1943)
leads to a very simple result, namely that the junction points in the network
move as if embedded in an elastic continuum. The deformation of the chain
end-to-end vectors is thus identical t o the deformation of lines drawn in
corresponding directions on the bulk material. This is called an afine (i.e.
similar) deformation of chains.
Application of the affine deformation principle enables the change in the
entropy of the system for any given state of strain t o be derived. Since there
is no change in the internal energy (A U = 0 ) )this leads at once to an expression
Rubber Elasticity 41

for the change AA in Helmholtz free energy, which is equal to the work W
done by the applied forces in producing the deformation, i.e.
W =AA = A U - TAX = - T A S , (5)
where A S is the entropy of deformation and T the absolute temperature.
From the value of W so calculated, it is a simple matter to derive the stress-
strain relation corresponding to the particular type of strain considered.

5. Stress-strain relations for vulcanized rubber


Qeneral strain
The most general type of strain may be defined in terms of the strain
ellipsoid. A sphere of unit radius in the unstrained state is deformed to an
ellipsoid with three unequal principal semi-axes A,, A, and A,, corresponding
to extensions in the ratio A,, A, and A, in three mutually perpendicular
directions. I n general, the axes of the strain ellipsoid are rotated with respect
to the corresponding diameters of the unstrained sphere; the general strain is
therefore equivalent to a pure strain, in which the axes do not rotate, together
with a ' rigid-body ' rotation.
Since no work is done by the applied forces in a rotation of the deformed
body, the elastic properties of the material may be completely specified in
terms of a pure homogeneous strain without rotation.
It is convenient to consider the material in the unstrained state to be in the
form of a cube of unit edge length. Under a pure homogeneous strain this is
transformed to a rectangular block of dimensions A,, A, and A, (fig. 7). It is
usual to assume (in agreement with experiment) that the deformation takes
place without any appreciable change of volume, i.e.
A,A,A, = 1. (6)
This implies that only two of the three principal extension ratios are inde-
pendently variable.

Fig. 7. Pure homogeneous strain. ( a )Unstrained state, ( b ) strained state.

The principal stresses t,, t 2 and t , corresponding to this state of strain act
normally to the surfaces of the deformed cube in the directions of the principal
strains (fig. 7 ) . The problem is to find the relation between these principal
stresses and the principal extension ratios A,, A, and A,.
The statistical theory of the network yields the following expression for the
work of deformation (per unit volume):
W=$NkT(A12+ A , 2 + A , 2 - - 3 ) , (7)
42 L. R. G. Treloar

where N is the number of ' chains ' (segments between junction points) per
unit volume. From this it is easy to obtain the strees-strain relations.
Putting
NkT = G, (7 a )
these are of the form
tl-t,=G(h,2- A,'),
t , - t , = G( - As2).
Four important features of these equations are particularly noteworthy.
These are
(1) The expression for the work of deformation, and the corresponding
stress-strain relations, involve only a single physical parameter relating
to the material, i.e. N , the number of chains per unit volume. This,
in turn, is determined by the number of cross-linkages introduced.
(2) For a given state of strain, only the differences between principal stresses
are obtainable. This is equivalent to the statement that the principal
stresses are indeterminate to the extent of an arbitrary hydrostatic
pressure, and is a direct consequence of the assumption of constancy of
volume or volume incompressibility.
(3) The stress-strain relations are non-linear, involving the squares of the
principal extension ratios. Hooke's law is not, in general, obeyed.
(4) As a corollary to ( l ) ,all rubbers, whatever their chemical constitution,
should have the same form of stress-strain relations, subject only t o a
scale factor or modulus.

Particulars types of strain


( a ) Simple extension
The indeterminacy of the principal stresses does not usually introduce any
difficulty in the application of the theory to practical problems, since in most
cases the boundary conditions eliminate the uncertainty. I n the case of a
simple extension, for example (fig. 8 b), the principal stress t acts in the
direction of the extension, and the lateral stresses are zero,
Hence
tl=t; t,=t,=O. (9)

-A

Fig. 8. (a) Unstrained state, ( b ) simple extension, (c) uniaxial compression.


Rubber Elasticity 43

If h is the extension ratio, the condition (6) for constancy of volume yields:
A , = A; A, = A, = A-"2. (10)
The first of eqns. (6) then gives the stress:
t = NkT(A 2 - l/A). (11)
In this equation t is the true stress, or stress per unit cross-sectional area,
measured in the deformed state. In practice, it is more convenient to introduce
the nominal stress F , or force per unit unstrained area. Putting t = F/area =
AF (from (10)) eqn. (11) becomes:
F=NkT(A-l/h2). (12)
The form of this equation is shown in fig. 9.

( b ) Uniaxial compression
This state of strain (fig. 8 c) is formally equivalent to simple extension, but
with A (the compression ratio) less than 1. The same formulae (11) and (12)
apply as for simple extension, but t and F are in this case negative, i.e.
compressive.

( c ) Simple shear
For a simple shear in the ( x , y ) plane (fig. 10) the principal axes of the
strain ellipsoid are inclined to the direction of shearing. Their magnitudes
are related by the equations

EXTENSION

Fig. 9. Combined stress-strain curves for rubber in extension and uniaxial com-
pression. Circles and points, experimental. Continuous curve, theoretical
(eqn.:(W
44 L. R. Q. Tretoar

where A, is the major axis of the strain ellipsoid in the plane of the shear.
If y is the amount of the shear (Love, 1927)
y=tan + = A - l / A (14)
Insertion of the values (13) in the expression (7) for the work deformation
yields, with ( 1 4 ) ,
W =-;NkT(hz+l/h2-2) = i N k T y 2 (15)
The shear stress t,, is obtained by differentiation of W, i.e.
t z y = dW/dy = NkTy = Gy. (16)
Thus for simple shear Hooke’s law (stress proportional to strain) is obeyed.
This is the only type of deformation for which Hooke’s law applies.
It is seen from eqn. ( 1 2 ) that the quantity N k T which occurs in the stress-
strain relation for simple extension is equivalent to the shear modulus.

Fig. 10. Geometry of shear strain. ( a ) Unstrained state, ( b ) strained state.

Experimental examincztion
Space does not permit a detailed presentation of the experimental evidence
which has been obtained in support of the foregoing theoretical conclusions.
It is only possible to sketch out the lines which have been followed and to give
a few illustrative examples. Fuller accounts have been given elsewhere
(Treloar 1958, 1970).
Typical data for the first part of the extension curve and for uniaxial com-
pression are shown in a combined plot in fig. 9. It is seen that these data lie
on a single continuous curve which is approximately of the expected form,
though there are significant deviations at high extensions. A similar degree
of agreement is obtained for simple shear (fig. 1 1 ) ’ the stress-strain curve being
approximately linear up to very high values of the shear strain. Moreover,
the value of G required to fit the data for extension and compression for a
particular rubber (0.39 MN m-2) also fit the shear data for the same rubber.
Experiments have also been made for a pure homogeneous strain of the most
general type, obtained by stretching a rubber sheet by different amounts in
two perpendicular directions and measuring the corresponding stresses. Fig. 1 2
represents the relation between the difference of the two principal stresses in
the plane of the sheet, t , - t,, and the corresponding values of h I 2- A 2 2 . It is
seen that the results are in good agreement with the theoretical relation (8).
These experiments, and many others, have demonstrated that the statistical
theory gives a reasonably accurate description of the elastic properties of
rubber. Not only are the various types of stress-strain relations satisfactorily
Rubber Elasticity 45

SHEAR STRAIN

Fig. 1 I . Stress-strain relation for rubber in simple shear. (a) Experimental,


( b ) theoretical (eqn. 16)).

/ I I I I I
0 2 4 6 8 10

Fig. 12. Pure homogeneous strain. Difference of principal stresses plotted against
difference of squares of principal extension ratios in plane of sheet.
46 L. R. Q. Treloar
reproduced, but-what is more important-these particular stress-strain
curves are shown t o be all related to a single general function (eqn. (7)) which
contains only one material parameter or elastic constant. Furthermore, by
making use of cross-linking agents which react in a quantitative manner with
the rubber hydrocarbon chains, i t has been possible t o estimate chemically the
number of cross-linkages introduced, and hence also the number of chains per
unit volume, N . This is found to agree reasonably closely (e.g. to within
about 25%) with the value of N calculated from the measured modulus of
the cross-linked rubber. Since the relevant equations (e.g. eqn. (12)) contain
no adjustable parameter, this agreement is highly significant.

6. Normal stress effects


Simple shear
The theoretical conclusions discussed above bring out some of the more
remarkable differences between the large-strain behaviour of a rubber and the
small-strain behaviour treated by the classical theory of elasticity. Even
more remarkable, however, are the conclusions derived from an examination
of the normal stresses generated in simple shear, and in allied types of deforma-
tion such as torsion. Problems of this kind have been considered particularly
by Rivlin (1948).

tYX i YY
Fig. 13. Tangential and normal components of stress in simple shear.
Consider a material subjected to a large simple shear in the (2,y) plane, as
in fig. 10. Fig. 13 represents an element cut from the deformed material with
it5 surfaces normal t o the x,y and z directions. We denote the normal stress
components by t x Z , t,, and t Z Z , and the tangential stress components by tzy
and tvx. Rivlin has shown that for a material whose stored-energy function
has the form given by eqn. (7),the components of stress are given by
(tangential),
t x , = tyx = Gy
1
t x x - t,, = G y 2 ; t,, - t z z = 0 (normal).
These equations imply that a large simple shear cannot be maintained by shear
(17)

stresses alone; it is necessary in addition t o apply a normal stress (tensile or


compressive) t o a t least one pair of opposite faces. If we choose the case when
there is no stress in the direction normal to the plane of shear (t = O ) , we have
tYy=tzz=O; tzz=Gy2. (17
Rubber Elasticity 47

The stress component tZz represents a tensile stress acting in the direction of
shearing. If this stress is not applied, i.e. if only the tangential stress com-
ponents t,,, t,,, are applied, the resultant deformation will not be a simple
shear; there will be, in addition, a contraction in the x direction, and an
expansion in the y and z directions.
It is interesting to note that these normal-stress components are proportional
to the square of the shear strain. Components of this type have no analogue
on the classical theory, in which all the components of stress are simply
proportional to the corresponding components of strain.

Torsion of a cylinder
I n the torsion of a cylinder (fig. 14) the element of the material is subjected
to shear strain. This gives rise to normal stress components of the type con-
sidered, but in a slightly different form. If we take the case where the
cylindrical surface is stress-free, it is found that at any point within the
cylinder there are normal components of stress acting in both the radial and

Fig. 14. Couple M about axis and normal thrust N acting on cylinder in torsion.

axial directions, in addition to the ordinary tangential components. I n this


case the normal stresses are compressive. The axial component leads to a
total compressive force or thrust N , acting on the end surface of the cylinder,
of amount
N = &rrGa+h2a4 (18)
where i,h is the torsion in radians per unit axial length of the cylinder, and a is
the radius of the cylinder. The couple M generated by the tangential com-
ponent of the stress is given by
M =4~ G t , h a ~ . (19)
This couple is identical in form to that given by the classical theory for a
material of shear modulus CT (as would be expected, since Hooke’s law applies
in simple shear). The normal thrust, however, is proportional to the square
48 L. R.G. TTeloar

Fig. 15. Relation between couple and torsion for cylinder of 2-54 cm diameter
(Rivlin and Saunders 1951.)

of the torsion, and has no counterpart on the classical theory. If the com-
pressive force is not applied in the axial direction, the cylinder will increase in
length when twisted. This effect is easily demonstrated.
The experiments of Rivlin and Saunders (1951) have confirmed that the
normal thrust on a cylinder in torsion is proportional to the square of the twist
(fig. 16) while the couple is proportional to the first power of the twist (fig. 15).
There was, however, some discrepancy between the values of G derived from
these two measurements. For a quantitative representation of all the data a
2-constant form of stored-energy function of a semi-empirical type (known as
the Mooney-Rivlin function) was found t o be necessary. The molecular
significance of this form of deviation from the statistical theory is still not clear.

Fig. 16. Compressive force N on end face of cylinder in torsion plotted against
square of torsion, for cylinder of 2.54 cm diameter. (Rivlin and Saunders
1951.)
Rubber Elasticity 49

Weissenberg effects
Normal-stress effects may be observed not only in vulcanized rubbers, but
also in polymer solutions and polymer melts subjected to shear flow, torsion,
etc. Such effects in liquids were demonstrated and studied particularly by
Weissenberg. Since they have been the subject of a recent review article
(Trevana 1968) they are only referred to here in passing. The point which it
is desired to emphasize is that polymer solutions and polymer melts display
many of the properties of a rubber-like solid (e.g. elastic recoil), the chief
difference being that the stresses generated by a deformation rapidly relax.
But in continuous steady flow, the stresses may be considered to be con-
tinuously regenerated, and the relations between the normal and tangential
components of stress are of a similar form to those for a solid rubber.

7. Photo-elastic properties of rubber


One of the outstanding merits of the statistical theory is its ability to deal
with a number of other phenomena in addition to the purely mechanical
properties of rubber. By suitable modifications, or by the introduction of
additional features, it has been successfully applied to various aspects of
swelling and solution properties, to the relation between the equilibrium swelling
in organic liquids and the applied stresses, and to the phenomena of photo-
elasticity (Treloar 1958). We shall deal here only with this last effect.
I n its normal unstrained state rubber is both mechanically and optically
isotropic. When strained, however, it develops anisotropic properties, that is
to say, its properties are different in different directions.
The optical anisotropy leads to the phenomenon of a strain-induced double
refraction, or birefringence-a phenomenon already well-known in transparent
solids (i.e. glasses). In the case of a rubber the optical anisotropy is associated
with the difference of polarizability of the chain molecules in the direction of the
chain axis and in the direction normal to the chain axis. I n the unstrained
state the molecules are arranged in all directions at random, with the result
that the system as a whole is optically isotropic, but on the application of a
stress the molecules become preferentially oriented or aligned, and the system
is no longer isotropic.
The optical properties of the molecule may conveniently be represented in
terms of a chain of randomly-jointed links (fig. 17) in which the individual
links are themselves optically anisotropic, having polarizabilities a1 for an
electric field acting in the direction of the length of the link and a2 for a field
in the transverse direction. It was shown by Kuhn and Griin (1942) that the
resultant chain has an optical anisotropy which is a function of the distance

Fig. 17. Randomly-jointed chain with optically anisotropic links.


C.P. D
50 L. R. G. Treloar

between its ends. Knowing this relationship, it is possible to treat the photo-
elastic properties of a network of such chains in a manner analogous to the
treatment of the mechanical properties of the network, the total polarizability
of the network being derived both for the direction of the extension, and for
the transverse direction. Applying the standard Lorentz-Lorenz relation
between refractive index n and polarizability P per unit volume,
n2-1 4~r
-=-p
n2+2 3
to each of these directions separately, Kuhn and Griin were thus able to obtain
the corresponding difference of refractive indices n , - n 2 for light polarized in
these respective directions. The resulting relation between birefringence and
stress is
n,-n, - 2 ~ r( n , 2 + 2 ) 2
45kT no (011 - 01217
t
where t is the tensile stress and no the mean refractive index (i.e. the refractive
index in the unstrained state). Since all the terms on the right-hand side of
this equation are constant, this means that the birefringence is proportional
to the applied tensile stress.

Fig. 18. Relation between birefringence and stress for natural rubber. 0 Stress
increasing. Stress decreasing.
Rubber Elasticity 51

Fig. 18 shows the experimentally observed relation between birefringence


and stress for vulcanized rubber at 25" and 75°C. At the higher temperature
the expected linear relationship is observed, but at 25°C an anomaly appears
at a stress of about 2 MN m--2(corresponding to an extension of about 250%).
This is due to the onset of crystallization in the rubber at high extensions.
The crystallites so formed are themselves birefringent, and their presence
therefore introduces an additional effect not envisaged by the simple theory
for an amorphous cross-linked network. At the higher temperature crystalliza-
tion does not occur to a signscant extent. It is noteworthy that the curves
are reversible so long as crystallization is absent, but that crystallization gives
rise to a large hysteresis loop, which means that the crystallites persist until
the stress is reduced to a level well below that required for their initiation.
From these experiments it is possible to derive, with the help of eqn. (21),
~ as,the optical anisotropy of the ' random link '
the value of the quantity 0 1 -
in the chain. This has a characteristic value for any given polymer chain.
It is an important parameter, for it enables us to estimate the length of the poetu-
lated random link. It is not possible to go into the details of this calculation;
it depends on the prior calculation of the longitudinal and transverse polariza-
bilities of the repeating unit in the chain structure using independently
determined data for the polarizabilities of individual bonds in the structure.
The important point is that if we can estimate the length of the ' random link ',
and hence the number of ' random links ' in the chain, we have a method of
estimating the statistical length, e.g. the r.m.s. length (eqn. (2)) of the chain.
This can then be compared with the statistical length calculated on the basis
of free rotation about bonds in the actual chain structure. Estimates of this
kind have been made for a number of polymers, including polyisoprenes
(natural rubber and gutta-percha) (Saunders 1957) polyethylene (above the
crystal melting point) (Gent and Vickroy 1967, Saunders et al. 1968) and
silicone rubbers (Mills and Saunders 1968).

8. Thermodynamics of elasticity
In Section 3 reference was made t o the basic thermodynamic consequences
of the statistical theory, and to the temperature dependence of the stress
(fig. 4) associated with the reduction of entropy on extension. I n the presenta-
tion given there the purpose was to convey the essential ideas in the simplest
form, and certain complications which arise in a more critical examination
were omitted. Having now reviewed the general development of the statistical
theory it is appropriate to reconsider the fundamental thermodynamics more
closely.
Although at high extensions the tensile force, at constant length, increases
approximately in proportion to the absolute temperature, this is not true for
rather small extensions. This is shown in fig. 19, from which it is seen that at
small extensions the tension actually falls instead of rising, as the temperature
is increased. By interpolation the inversion of slope is found to occur at about
10% extension.
In thermodynamic terms this means that there is a significant increase in
internal energy on deformation in contrast to the postulate of the statistical
theory, according to which the change in internal energy should be zero.
52 L . R. G . T'relour

_.

EL
~~ O0 20 TEMPERATURE,
40 'C
60 80

Fig. 19. Dependence of force at constant length on temperature for extensions


indicated. (Anthony et ul. 1942.)

The explanation of the reversal of slope or ' thermo-elastic inversion ' at about
10% extension is in fact quite simple. It arises from the ordinary thermal
expansion of the rubber in the unstrained state. This will be clear from fig. 20,
in which Zoo represents the uiistrained length at temperature To and lo, the
unstrained length a t temperature T . If I is the strained length, which is
maintained constant while the temperatnre is varied, it is clear that the strain
(1 - Z,)/ actually decreases as the temperature is raised. Denoting the strains
at temperatures To and T by e, and eT, respectively. we have:
eT=eO- -P( T - T o ) ,
3 (22)

+F
Fig. 20.
Rubber Elasticity 53

where /3 is the volume expansion coefficient. It follows that if the modulus


(NkT in eqn. (12)) is proportional to the absolute temperature, the force a t
constant length will not be proportional to the absolute temperature, since
constant length does not correspond to constant strain. If we assume the
strains to be sufficiently small for Hooke's law to apply, the condition for
inversion, namely (lJf/lJT)z=O, is found from (22) to be

eo= B
- (2T-To). (23)
3

Putting /?= 6.6 x (for natural rubber) and taking a typical range of T as
in fig. 19, i.e. T0=273', T=353", we obtain from (23)
e, = 0.095 = 9.5%, (24)
which is sufficiently near t o the experimental 10% strain.
The significance of the above analysis is that the internal energy change
which accompanies the extension of rubber is not essentially connected with
the mechanism of the elasticity itself, but is a secondary effect associated with
small changes in the volume of the material. The volume of the rubber is
determined by the forces between the chain molecules; these forces are of the
same kind as the forces between the molecules of an ordinary solid If we
consider the variation of the tensile force on rubber at constant strain (i.e.
constant extension ratio A ) rather than at constant length, this is found to be
approximately proportional to the absolute temperature for all values of the
strain, and the inversion effect does not enter into the argument.
This was the conclusion originally arrived at by Gee (1946), who derived the
further result that if the stress-temperature relations could be measured at
constant length and constant volume (by the application of an overall hydrostatic
pressure to counteract the thermal expansivity) then the force would be
strictly proportional to the absolute temperature, and the internal energy
change (associated with the change of volume under the normal constant
pressure conditions) would be zero.
This conclusion is now known to be not quite as exact as was believed at
the time. Later very careful experiments by Allen et al. (1963, 1968), in
Gee's laboratory on the stress-temperature coefficient at constant volume
have shown that even under these conditions the change of internal energy
on extension is still not quite zero. Its magnitude depends on the type of
rubber; for natural rubber the internal energy change accounts for about 13%
of the total stress (the remainder being due to the change in entropy). This
small residual effect may be accounted for on the basis of a more rigorous
theoretical analysis of the network problem by Flory (1961) which takes into
account the possibility of internal energy effects within the single long-chain
molecule itself. In the simpler treatments of the network it had previously
been assumed that the internal rotations about bonds take place quite freely.
In a real molecule, however, there is some restriction of rotation due to the
interferences between neighbouring atoms along the chain, as a result of which
different conformations of the chain will have different internal energies. For
such a chain the statistical properties, as represented by eqn. ( l ) ,will no longer
be independent of temperature. Flory's analysis relates the internal energy
54 L. R. G. Treloar

contribution t o the stress, a t constant volume, t o the temperature dependence


of the statistical length of the free molecule.
Thermo-elastic studies of extension a t constant volume involve rather
refined methods, and are not easy to carry out. It has recently been shown,
however (Boyce and Treloar 1970), that an almost identical result may be
obtained by working with torsion, under ordinary constant pressure conditions.
The experiment consists in measuring the variation with temperature of the
couple M required to maintain a constant torsional strain # in a. twisted cylinder.
If the statistical theory applied, this should be proportional to the absolute
temperature. A small correction has to be applied t o take account of the
expansion of radius on heating, but this expansion (unlike the case of simple

0 0

- 2 -0-297
A -
I I 1 I I
20 30 40 50 60
TEMPERATURE, "C

Fig. 21. Dependence of couple on temperature for cylinder in torsion, for values of
#ao shown. Temperaturc increasing. 0 Temperature decreasing. (Boyce
and Treloar 1970.)
Rubber Elasticity 55

extension) does not have a major effect on the strain. I n particular, there
should be no stress-temperature inversion effect. The experimental data
(fig. 21) bear out these expectations; the relative rates of increase of the couple
are independent of the amount of torsion, and nearly proportional to the
absolute temperature. The small deviation from linearity yields the result
that about 13% of the couple is due to the internal energy change (associated
with the internal energy of the single molecule), in precise agreement with the
results of Allen et al. for simple extension at constant volume.

9. Conclusion
This review has concentrated primarily on the purely elastic properties of
rubber and on their molecular interpretation. It has been shown that the
statistical theory is capable of providing a satisfactory general account of the
most important of these properties and of predicting results which have
subsequently been confirmed experimentally. There are indeed certain
significant deviations from the simple theory as expounded, but except for
those revealed by thermodynamic studies these have not been discussed.
Outstanding among such deviations are deviations in the form of the stress-
strain relations (cf. fig. 9)) which though partially described by a more complex
strain-energy function have so far not received any satisfactory or conclusive
explanation in terms of molecular mechanisms.
Many of the practically significant properties of rubbers, however, are
related more closely to the imperfections of the elasticity rather than t o the
elasticity itself, i.e. to the visco-elastic properties of the material. I n contrast
to the purely elastic properties, which are similar for all types of rubber, the
visco-elastic properties are extremely dependent on the chemical constitution
of the material. It is these properties which determine the dissipation of heat
in a cyclic deformation and the range of temperature over which rubber-like
properties are retained. They also play an important part in determining the
resistance to tearing and fatigue. Nevertheless, although an understanding of
the basic mechanism of the purely elastic behaviour is not related directly to
these more practical situations, it does provide the general framework within
which their interpretation has to be discussed. I n this sense the statisitcal
theory must be regarded as an important development not only in our know-
ledge of the structure and properties of rubber-like polymers, but also in our
ability to cope with the practical needs of the rubber industry.

REFERENCES
ALLEN,BIANCHI and PRICE,1963, Trans. Faraday SOC.,59, 2493.
ALLEN,KIRRHAM, PADGET and PRICE, 1968, Polymer Systems (Macmillan)p. 51.
ANTHONY, CASTONAND GUTH,1942, J . phys. Chem., 46, 826.
BOYCEand TRELOAR, 1970, Polymer, i 1 , 2 1 .
FARADAY, 1826, Quart. J. Sci., 21, 19.
RORY, 1961, Trans. Faraday Soc., 57, 829.
GEE, 1946, Trans. Faraday SOC.,42, 585.
GENT and VICKROY, 1967, J . Polym. Sci.,A-2,5, 17.
GOUGH,1805, Mem. Lit.phil. SOC.Manchester, 1, 288.
JAMESand GUTH, 1943, J. chem. Phys., 11,455.
JOULE, 1859, Phi!. Trans., 149, 91.
KUHNand G R ~ N1942,
, Kolloidzeits, 101, 248.
56 Rubber Elasticity

Love, 1927, The Mathematical Theory of Elasticity (Cambridge).


MEYERand FERRI,1835, Helv. chim. Acta, 18, 570.
MEYER, v. SUSICH and VALKO,1932, Kolloidzeits, 59, 208.
MLS and SAUNDERS, 1968, J. Makromol. ScLPhys.
RIVLIN,1948, Phil. Trans. R. SOC.A, 241, 379.
RIVLINand SAUNDERS, 1951, Phil. Trans. R . Soc. A, 243, 251.
SAUNDERS, 1957, Trans. Faraday SOC.,53, 860.
SAUNDERS, LIGHTFOOT and PARSONS, 1968, J. Polym. Sci., A--2.
TRELOAR, 1958, The Physics of Rubber Elasticity (Oxford), 1970, Introduction to
Polymer Science (Wykeham).
TREVENA, 1968, Contemp. Phys., 9, 537.

The Author:
L. R. 0. Treloar, 0.I3.E.. B.Sc., Ph.l)., L).Sc., F.InsL.P., Professor of Polymer and Fibre Science,
University of Manchestrr Institate of Science and Tcchnology. Previously on tho research staff
of the General Electric Company ( 1 929-38), the British Rubber Producers’ Research Associatiolr
(1938-19), the British Rayon Rescarch Associat,ion(1049-61) and thc Cotton, Silk and Man-hiatlo
Fihres Research Association (1961-63).

Das könnte Ihnen auch gefallen