Sunteți pe pagina 1din 21

JOURNAL OF POLYMER SCIENCE: PAKT C NO. 12, PP.

89-109 (1966)

Statistical Mechanics of Polymers

EUGENE GUTH, Oak Ridge National Laboratory, *


Oak Ridge, Tenneasee

1. INTRODUCTION
Statistical mechanics of polymers originated 30 years ago.’ Since then
it has grown to such an extent that a recently published book2 of 562 pages
could not cover it completely. I n view of this phenomenal growth of our
subject, we shall deal in this paper mostly with comments on one central
topic : the statistical mechanics of the long-range elasticity of network
polymers (e.g., rubber, etc.). This topic has been called “the one of the
greatest importance.”3
I n our discussion we shall use the fundamental principles of statistical
mechanics4 and avoid special and/or unrealistic models. It is well known
that the pressure exerted by an ideal gas can be approximately computed
using a quite unrealistic model. I n this model all molecules have the same
velocity and a third of their total number moves in the direction of each of
three Cartesian coordinate axes. Still, from the essential correctness of
the result one can not conclude that the model used is realistic. A realistic
treatment has to be based on the Maxwellian distribution of the velocities.
Furthermore, any statistical-mechanical theory has to be self-consistent.
Our exposition will be largely explanatory and descriptive. However,
we shall touch, in some detail, on some points which are perhaps novel or
a t least not well known.
Polymer molecules consist of a large number (one hundred or more)
identical units, the monomers, and form flexible chains. These flexible
polymer chains are large enough to constitute (small) “ensembles” (with
possibly size-dependent properties). Furthermore, the total number of
polymer chains in a solid or in a solution also represent a statistical en-
semble. The behavior of such an ensemble of (small) ensembles is a topic
of statistical physics.
I n contrast to liquids, there i s a common property of the chemical struc-
ture of polymers : their macromolecular state and their flexible chain struc-
ture. This leads to a general theory of polymers.
The main stumbling block to a general theory of liquids is the impossibil-
ity of a general theory of intermolecular interactions, which vary consider-
ably for different substances. I n contrast, many important properties of
polymers depend mainly on intramolecular interactions, which vary much
* Operated by Union Carbide Corporation for the U. S. Atomic Energy Commission.
89
90 E. GUTH

less for different substances, e.g., loosely speaking on size and shape of
the macromolecules. For example, viscosity, thermal conductivity, dif-
fusion, and light scattering can be treated in a general manner on the basis
of intramolecular statistics. Osmotic pressure and other solution properties,
on the other hand, are determined not only by the intramolecular, but to a
large extent by the intermolecular, interactions. T h u s we shall distinguish
between intramolecular and intermolecular properties of polymers and the
corresponding statistics. Here, we shall restrict ourselves mostly to intra-
molecular statistics.
Owing to the possibility of using statistics, the theory of physics of poly-
mers-at least in certain fundamental aspects-is simpler than that of low
molecular materials. For example, the equation of state of even solid
hydrogen is neither universal nor simple. On the other hand, the equation
of state, e.g., the stress-strain relation for rubberlike polymers, is universal
and simple. Loosely speaking, such a polymer behaves as an (ideal) gas
whose molecules are connected by flexible strings or “worms.” Moreover,
the stress-strain relation will be a universal function of the extension, up
to several hundred per cent elongation. It might be of interest to point out
that a rubberlike solid has, simultaneously, physical properties of a solid,
liquid, and a gas. It has a fixed shape, as does a solid; it is incompressible,
as is a liquid (first established by Riintgen) ; and the temperature depend-
ence and the order of magnitude of its Young’s modulus is like that of a gas.
It heats up when stretched adiabatically (Gough-Joule effect), similarly to
the heating up of a gas if compressed adiabatically.
The peculiar thermal behavior of stretched rubber shows up strikingly
in the phenomenon of thermoelastic inversion. At small elongations (10%
or less), rubber expands on heating as does any other solid. For larger
elongations it contracts on heating. Thus, a t a certain critical elongation,
called the thermoelastic inversion, the slope of the stress-temperature curve
changes sign.
The principal biologically important materials, like proteins and nucleic
acids, are (bio) polymers. Thus, biophysics is a branch of polymer phys-
ics-perhaps the most important. Similarly, biochemistry is a branch of
polymer chemistry-perhaps the most important.
Thus, the problem of rubberlike elasticity is deeply interconnected with
all those sciences which make a fundamental inquiry into organic life.
These sciences are biochemistry, biophysics, physiology, pathology, and
genetics, and they deal with nerves, muscles, vitamins, proteins, and
chromosomes.
Rubber belongs to a class of substances which nature uses as a framework
for all living things. The similarity betwceri rubber arid rcstirig muscle
has loiig been recognized. I t is well brought out by the f w t that rcstirig
niuscle develops heat, wheii stretc.liet1 rapidly, just as rulher does. Of
course, the behavior of muscle is, in gcncral, much more complicated than
that of rubber. Nevertheless, in certain essential aspects, rubber may be
considered as a simple prototype or model of muscle.
STATISTICAL MECHANICS 91

From this point of view, the study of rubber may be considered to be an


important link in the monumental attempt of the human mind to unveil the
mystery of organic life. I n the tree, Hevea brasiliensis, nature is the manu-
facturer of rubber. After the discovery of rubber latex and its trans-
formation into solid crude rubber by the colloidal process of coagulation,
the next important development in rubber technology was the discovery
of vulcanization by Goodyear. Though unknown to Goodyear, vulcaniza-
tion is an imitation of an artifice of nature to preserve an organic material
from decay and make it more enduring. Wool protein or keratin contains
sulfur as a protective agent, just as vulcanized rubber does. Consequently,
it may last a long time after the “crude” (i.e., unprotected) proteins of an
animal body have disintegrated. The stereospecific synthesis of rubberlike
materials is a successful attempt of man to replace nature as a manufac-
turer of rubber or to find substitutes for this material.
It is of interest to point out that one of the earliest investigators or
rubber elasticity was a physiologist (Schmulewitsch, 1870). In recent
years, the quantitative study of muscle has required delicate and accurate
instruments. Work of this type has been carried out chiefly by physiolo-
gists with a background in physics and mathematics. The physics of
rubber was studied many years ago by Joule, Kelvin, Stefan, Boltzmann,
and Rontgen.

2. STATISTICAL MECHANICS OF POLYMER ELASTICITY


General Concepts
Network polymers, like rubbers, resemble gases very strikingly in their
thermoelastic behavior. The proportionality of stress to absolute tem-
perature under constant strain suggests that the tension in stretched rub-
bers, like the pressure in gases, is associated with a change in entropy of
the material when it is deformed, rather than with a change in internal
energy. Quantitatively, the behavior of stretched rubbers, or of gases
under pressure, is determined by their free energy, F, which in the case of
simple tension may be written as
F(L,T) = U(L,T) - TS(L,T) (1)
where L represents the length of the stretched material, and the other
symbols have their customary significance. The tension is

The same formulae hold for gases, if L is replaced by the volume V ,


and 2 by the negative pressure ( - p ) . The simple and uniform thermo-
elastic behavior of gases, indepeiiclent of their chemical composition, is due
to two important characteristics: ( 1 ) the internal energy is in all cases
nearly independent of the volume; ( 2 ) the entropy consists of two parts,
one associated with the heat capacity, h, of the gas,dependent on the chemi-
cal composition, but inalcing no contribution to (bS/bV),, and the second
92 E. GUTI-I

associated with the number of positional configurations available


to the system, independent of the composition, and making a contribution
to (dS/dV)T,which is independent of T . The uniform and simple thermo-
elastic behavior of “ideal” rubberlike materials, independent of their com-
position, may be understood in a similar way. The internal energy of the
liquidlike mass of molecules will be practically independent of its form, such
strong chemical bonds as do exist being but little stretched by the relatively
small forces which already produce considerable deformations of the ma-
terial; (bU/bL)Twill be small. As will be indicated later in more detail
the entropy again consists of two parts, one associated with the heat capac-
ity and not contributing to (bS/’dL)T,the other associated with the number
of configurations available to the molecules within the material. The
number of these configurations will depend on the external form of the
material but not on the temperature, on the general structure of the polymer
molecules but not on their particular chemical composition. (bX/dL),
will, accordingly, be independent of temperature, and we will have ap-
proximately the form

Z(L,T) = Z,(L)T (3)

By keeping in view all these characteristics, one arrives a t a fairly good


understanding of the existence of rubber elasticity, and even can proceed
to the establishment of mathematical expressions which are in rather
satisfactory quantitative agreement with many experimental facts.
The present development of statistical mechanics makes it possible to
develop the theory of gases and solids by the application of first principles
of physics only. Practically, however, such a method, while highly de-
sirable, would be quite intractable mathematically. Therefore, another
approach is usually adopted. One starts with a clear-cut model of a gas
or solid. Such a model is always, admittedly, an approximation. It may,
however, be an excellent approximation if it neglects only unessentials and
takes care of the essential characteristics of the system under consideration.
For a clear understanding of the situation, it is desirable to put all necessary
approximations into the model, and then treat it rigorously as far as the
quantitative mathematical-physical development is concerned.
I n the present state of development, the most complete theory of rubber
elasticity is roughly analogous to the theory of an ideal gas. The char-
acteristic features of rubberlike elasticity are made comprehensible, but
further elaboration is required if one is to explain quantitatively the char-
acteristics of individual elastomers and the differences between them.
However, taking summary account of (a) hindered rotation and (b) thermal
expansion leads to an equation of state which is roughly analogous to van
der Waals’ equation, for real gases, but still neglects excluded volume
effects.
I n the case of solids, particularly crystals, the forces between the atoms
determine the properties of the system, the kinetic mobility of the atoms
SI’ATIY’I’ICAL MECIIAN ICS 93

being greatly limited. Recently it has been shown, however, that mole-
cules or groups of molecules may have a kinetic mobility even in the solid
state. For rubberlike materials, particularly, kinetic mobility of the con-
stituents is well established. This consideration has led to a statistical
mechanics of rubber elasticity.
The statistical mechanics of rubber elasticity may be developed in four
stages:
1. Treatment of isolated long-chain molecules of given chemical char-
acteristics, such as chain length, valence angle, and internal mobility.
2. Treatment of bulk rubber, considered as a network formed by firmly
linking together such molecules, with the interaction between the molecules
taken into account only in the simplest possible manner.
3. Treatment of the process by which isolated long-chain molecules arc,
during cure, linked together into a coherent and (relatively) nonplastic
mass.
4. Introduction into the treatment of a more detailed account of the
forces acting between molecular chains.
The first step leads to results applicable only to very dilute solutions of
isolated rubber molecules; the second leads to an approximate theory of
bulk rubber the quality of which depends upon the postulated characteris-
tics of the individual chains. The simplest assumption is that all chain
segments in the network are still comparatively long (soft rubbers). It
permits the detailed working out of a theory corresponding closely to that
of an ideal gas. The third step is necessary if one is to give a quantitative
connection between the characteristics of the isolated molecules and those
of the cured inaterial in bulk. The fourth step, finally, corresponds to the
passage from the treatment of ideal gases to that of real ones.

3. STATISTICS OF SINGLE CHAINS

The statistical problem for an ideal (e.g., freely jointed) chain is identical
with the problem of random walk.’ The general expression for the prob-
ability C ( r ) of a chain of N segments of fixed length 1 is given by Rayleigh’s
integral (r = Ljl)

C ( r ) = (1/2a”3r ) L m s sin rs (sin s/s)~cis (4)

The general expression (4) is exact, but not immediately applicable to


practical problems. However, all cases of interest can be obtained3 by
different kinds of series developments for the integral (4) both for large and
for small N . I n the limit of large N we consider the two cases r << N
and r - N .
94 E. GUTH

Series Development for N >> 1 and r << N .


Here we shall just quote the result

exp (-3r2/2N)X 1 - 3(5 - 10r2/N

+ +
+3r4/N2)/20N (1015 - 9660r2/N 13734r4/N2- 5364r6/N3
+567r8/N4)/5600N2- 9(315 - 1260r2/N 1134r4/N2 +
-324r6/N3 + .
27r8/N4. .)/1400N3.. .] ( 5 )
The first factor is the well-known Gaussian formula.

Series Development for N >> 1 and r - N.


Again we just quote the result

x[?
.e-l(t)

-}
2t
{ l - t2 - .e-l(t)
-112

. ][1 +
9 (t)
.. .] (6)
where t = r / N , with q(t) = (k4 - 3Nk22)/8k22.The above development
is valid for the entire range 0 5 t < 1 , while the development (5) is good
only for t << 1 .

Development for Small N (Finite Series)


The integral (4) can be also evaluated in form of a series, which is just
the three-dimensional analog of the (one-dimensional) Bernoulli distribu-

(r)
tion. The result is
k < (N--r)/2
C(r) = [l/ZN+l(N- 2 ) ! d 3 r ] (-1)* (N - 2k - r ) N - 2 (7)
k=O

This series contains, as special cases, the expressions given by Rayleigh


and Chandrasekhar for the first few values of N . For large N , this series is
not useful because of severe cancellations.

4. STATISTICAL MECHANICS ON SINGLE CHAINS


We wish to emphasize here that the existence of a unique equation of
state is a limit property15just as is the existence of phase transitions. This
can be proved in the context of chain molecules by deriving the equation
of state using two different ensembles: (I) a length-temperature ( L , T )
ensemble, (11) a force-temperature (2,T) ensemble. In the first case we
calculate the average force acting between the ends of the chain for a given
fixed extension

z= Z(L) (8)
STATISTICAL MECHANICS 95

In the second case, we calculate the averagc extcnsion of thc chain when
it is stretched by a constant force.
L = I(Z) (9)
For chains of a few segments, the instantaneous value of L may differ
markedly from 1and the same is true for 2 and z.
Under these circum-
stances eqs. (8) and (9) are by no means converse relations; only in the limit
of very long chains will (8 and 9) lead to the same equation of state.
The probability distributions C ( L ) given in Section 3 are, essentially,
partition functions or thermodynamic probabilities in the terminology of
Planck. The entropy can be defined using Boltemann’s principle
s = klogC(L) (10)
The Helmholtz free energy will be then, according to eq. (1)
F = U - TS = U - kT log C(L) = -kT log C(L) (for U = 0) ( 1 1 )
The equation of state is then determined by eq. (2)
z = (W/dL), (12)
All this assumes ensemble (I).
For ensemble (11) one uses the Gibbs function, which, for our system,
becames
G(Z,T) = U - ZL - TS (13)
From this one determines the extension of the system.
L = -(dG/bZ), (14)
In statistical mechanics the Gibbs function is related to the sum over all
accessible states of the statistical ensemble, weighted with the factor
eZL/kT
. I n analogy to (11) we have
G(2,T) = -kT log D ( Z ) (U = 0); (15)
where D ( 2 ) is a generalized partition function
D ( 2 ) = f C(L)eZL’kTdL, (16)
which is, essentially, the Laplace transform of the partition function C ( L ) .
For an ideal chain of N segments each of length 1 we obtain*
t = L/N = cot h 12 - 1/12 = C(ZZ), (17)
where S ( x ) is the Langevin function.6 For small forces (ZI << kT),we
obtain the equation
- NP
t=-Z or 3kT
Z = - t
kT 3 N12
already obtained in the first paper on statistical mechanics of polymers.’
* t* = It; t = L / N , as used in eq. (6).
96 E. GUTH

The appearance of the Langevin function is a consequence of the physical


analogy between the orientation of the dipoles by an (external) field and of
the orientation of the segments by an (external) force.
For real chains, in contrast to ideal chains, there is no free rotation but
instead we have hindered rotation. I n particular, we can have rotational
isomerism.' In any case then in the Gaussian approximation
-
L2 = '/3 NIZ = '/3 ?(T) (19)
becomes temperature dependent.

5. STATISTICAL MECHANICS OF NETWORK


POLYMER ELASTICITY
In this secton we shall discuss briefly some aspects of networks of Gaus-
sian chains, for which the number of configurations is a Gaussian exponen-
tial function of the distance between their ends. We have to distinguish
between two related, but still different problems. In the first problem8
we concern ourselves entirely with the dependence of the entropy of the
cured and thus uniquely fixed network on its external form. For this prob-
lem it is not appropriate to consider structures, that could have been formed
instead of the particular $xed network. In the second problemg we consider
the increase in rigidity duiing cure. For this problem we compute the
shape-dependent term in the entiopy for all possible structures, and then
average the results. weighting each with the probability that the corresponding
structure i s the existing m e .
We shall distinguish between fixed points and junctions. The former are
points at which the network is immobilized by external constraints.
Specification of the fixed points corresponds exactly to the specification
of the volume in which a gas is contained. In the Gaussian approximation
each junction can take on any position in space, whatever the positions of
the other junctions are. The junctions are not fixed in space, but exhibit
Brownian motion compatible with the constraints in the network. lo
It is important to realize that the Gaussian network is a model, used to
calculate the shape-dependent term in the entropy due to the change of
the number of configurations available to the real molecular network as the
macroscopic dimensions change. The free energy of a moderately cured
network polymer contains, in addition, a term corresponding to a liquid-like
mixture of disconnected chains. Thus such a polymer occupies volume
because of the space-filling character of its molecular chains.
We shall distinguish between active and passive portions of the network.
The former stretches between the fixed points, contributes to the forces
exerted by the network on these points, and is part of some non-retracing
path along the chains from one fixed point to another. The latter consists
of unattached material and loose ends, attached to the network. A portion
STATISTICAL MECHANICS 97

of the molecular chain that extends between two adjacent junctions of the
network, we shall call a segment.
Only the active portion of the network will contribute to the entropy
a term that depends upon the positions of the fixed points, e.g., upon the
external form of the material. Let us denote the fixed points b y subscripts
a, 0 . . . the junctions by numerical or letter subscripts, and the points of
both types by subscripts 7 , v . . . . The positions of these points are in-
dicated by vectors rr. The partition function for the number of configura-
tions of the network is equal to the (double) product of the number of con-
figurations possible for the individual segments
C(r,, rB, . . . ; rl, r 2 . . .) = IITIc(/r, - r,l) (20)
r > v

C(Jr, - rd) is the configuration function for a segment of N,, links of equal
length 1.
The total number of configurations of the network, consistent with given
positions of the fixed points only, is obtained by summing, e.g., integrating,
C(ra, rs, . . . ; rll r2, . . .) over all possible sets of positions of the junctions
C(ra, rB, . . .) = f drl, f dr2, . . . C(ral rB, . . . ; rll rz, . . .) (21)
We have now to distinguish between Gaussian and non-Gaussian segments
and networks, respectively.
Gaussian Networks
For a Gaussian segment joining the points r, and ru and having N , , links
each of equal length I, we have the configuration function

With (22) the integral in (21) can be completely evaluated.‘l However,


such a complete evaluation is not necessary for the purpose of just obtaining
the number of configurations for the network as a function of its external
dimensions. For a Gaussian segment configuration function (22), integra-
tion over all junctions in eq. (21) eliminates all variables describing the
junctions, but retains a Gaussian function of the fixed points only

The form of this result is independent of the topology of the network,


which affects only the constants C and NlaB; the latter has the character
of the number of links in a chain and, in general, may depend on all the
constants N , , in eq. ( 2 2 ) . Equation (23) corresponds exactly to a fictitious
network of independent chains of N f a 6links, extending between the fixed
points CY and p. For the computation of entropy and of entropy forces,
then, an arbitrary Gaussian network can be replaced by a corresponding
set of independent Gaussian chains running between each pair of fixed
9n E. GIJTH

I)oint,s. This simplified eqiiiv:dent network is very convenient in cert:lin


t,ypts of wlculations.
We consicler T I O W the sl)cc:iul (::LSC of a unil (.ubc of n I)olynicr, suhjcc:tetl
lo uniform strct.ohing par:dIcl 1.0 l,hc edgcs until t.licsc havc lcllgths I,,,
I,,, and 11,. If :dl external forces w e nppliod normal to the surfaces, the
fixed p0int.s mill lie on these surfaces in positions which change in proportion
to the corresponding dimensions of t h e material. Taking t.hree of the
faces of the cuboid as the c0ordinat.e planes
2, = z,‘O’L,; ya = ya(0’L,; 2, = 2,(0)1,, (24)
x,(O), ya(0), z,(O) being the coordinates of the ath fixed point when the
polymer is unstretched and at a st.andard temperature. We wish to em-
phasize that these relations can be derived from the assumption that all
forces are normal to the surfaces.
The number of configurations possible for the network, now a function
of L,, I,,, L,, then becomes simply
C(LZ, L,, L,) = KI exp { - (K,LZ24-K,L,2 + K,LZ2)} (25)
whcrc

and K , and K , have similar forms. If the polymer is isotropic on the aver-
age, then K , = K , = K , = K / 2 , and equation (25) simplifies to

Equation (27) and, more generally, eq. (23) state a very simple and interest-
ing result: when a n arbitrary network of Gaussian chains is subjected to uni-
form stretching in three coordinate directions, by corresponding displacement
of its fixed points, the number of conJigurations possible for the network i s a
Gaussian function of its external dimnsions. Loose ends or unattached
material do not change this result, nor will they affect K or the entropy
forces exerted by the network.
For uniform stretch in the z direction of an incompressible polymer
L, = L,; v = L,L,L, = LZ2L, = 1 (28)
Introducing the entropy by appropriate generalization of eq. (10) and
using eqs. (11) and (12), we obtain the equation of state

where in the first approximation


STATISTICAL MECIIAKICS 99

111 c y . c29) t.lic first, tmii in thc twacltet- rcl)rcscnty thc inward pull due to
the cvitrol)y folws csc~rtc.tl1)y i.110 tict\vork, wliile tlio sc~c*oritl ~ tlio
t ( T J I11:~s
c1i:w:wt.w of :ti1 iiit,cnitLl 1iytlrost.atic:prcssriix! tluc to t.lic t,cii(lctic*yof t,hc
tictwork to cotit,r:tct. I ) , in cq. (XU), is t.lic tot.:~l(cxtcrii:iI : i i d in1~crri:tI)
pressure.
If one makes the assumption that all network junctions are $xed in
space,12each segment of the network will have a fixed extension det.ermined
by the external form of the mat,crial alone. For a unit cube stretched into
a cuboid with edges L,, I,,, I,,, the ext,cnsions of the T V segment will be

The number of configurations for the network is simply the product of the
numbers of configurations possible for the component segments

C(L,, L,, I,,)

The corresponding entropy is then


= Co IIII exp
r > Y
3
-- r?,
1
1Squations (24) are assunled now to be valid for both the fixed points and
the junctions
x r = x,(O)L,; ?JT = y,‘O’L,; 2, = Z,(O)I+ (34)
Equations (24) are assumed now to be valid for both the fixed points and the
junctions
n

The physical reason for obtaining the same form of the equation of state
(29) for such an unrealistic niodcl assuming fixed junctions was explained
in detail in references 8a and b. Briefly, it is possible to treat the junctions
as fixed in deriving results that deprnd on the statistical behavior of the
links as components of th r segments, but not on the statistical behavior of
segments as components of the network.
In reality, it has been proved in great detail in reference 8b that the junc-
tions have a Brownian nzotion comparable to that of other elenienls of the nel-
work. More particularly, the mean fluctuation of the junction coordinates
differs from that of the midpoints of the segments by less than 20%. Thus,
a model with fixed junctions is just as unrealistic as a model of an ideal gas
in which all molecules have the same velocity. Such an iinrcalistic model
should not be used in discussing the detailed behavior of the network. The
fact that in both cases essentially correct equations of state can be derived
from unrealistic models dors not justify a genrral use of such models,
which should be avoided.
100 E. GTJTH

Non -Gaussian Networks *


We start from eq. (6). Since we are interested essentially in log C ( r ) ,
it is clear that the first square bracket in (6) gives the main contribution,
and the contribution from each succcssive bracket is only of the order 1/N
of the cont,ributiori from the preceding bracket,. For small t , we can use a
series development of the first bracket of (6)
3NP
(36)
k=2

where a2 = 9/20, etc.


Taking only the first correction term for non-Gaussian behavior, we
can (1) carry out the integrations indicated in (U), (2) treat the junctions
as restricted to their most probable positions, arid (S), assuming that all
junctions are fixed, we can use the complete series (36). Thus, we obtain,
taking the first correction term in ey. (86)

For case ( 1 ) we obtain on the right side of eq. (87) the additional terms:
-61 - 62(L,? + L,2 + I’,2), (38)
where for a regular tetrahedral network
8 K 1
61 = 76,
16N
2 = -
2
x 4N
(39)

For case ( 2 ) , H and G in c‘q. (38) are replaced by quantities H ’ and G’,
which, however, for both cubic and tetrahedral regular networks, reduce
to ZZ and G.
&, &, H , G, IZ’, G’ are all constants of similar character as K . Their
geiieral definition is given i n eqs. (3.9, 3.24a, and 3.24b) of rcf.l2 The
additional terms (37) for case ( 1 ) reflect the fact that affine deformation is
not assumed here for the junvtions, but only for the fixed points.
With H = 3G, the q u atio n of statc for unilateral stretch becomes
(neglecting thernial expansion and c.ompressibility)

In the derivation of (37) and of (40) we assumed isotropy of the network


for all orientations of the c.oordinatc axc’s. If we relax this assumption and
assume isotropy just for n special orientation of the axes, we can obtain
STATISTICAL MECHANICS 101

siniple closed expressions for the regular cubic and the regular tetrahedral
model. For the former, we obtain
= Kc) I
(KkT/3Kc)[e-'(LK,) - L-3'2~c-1(L-1'2 (42)
For the latter, we have

For low extensions c-' (x) 3x and both eqs. (42 and 43) reduce to the
Gaussian eq. (29) as they should. The parameters K , and K~ are related to
the maximum extensibility of the chains and, hence, of the network
(Kt d5Kc).
6. Determination of the Modulus from its Change during Cure
I n the equation of state (29) the quantity K enters. It is defined by
eq. (26) or alternatively by eq. (35). To arrive at a definite value of K it
is necessary to introduce further assumptions into the theory. It, follows
from eq. (3.5), and the assumption of isotropy that

K = (2/3) (K, + K , + K J f c c IT,~(")I~/N,~


=
s > Y
(44)

where rrV(")is the extension of the T V segment when the material is un-
stretched; can be explicitly expressed in terms of the N,, and r,,
the vector for the positions of the fixed points. Thus K can be explicitly
expressed for a completely specified network as a sum from the contribu-
tions of the component segments.
Equation (44) is of interest theoretically. Moreover, it suggests a more
practical statistical specification of the network for the determination of
K . We define
A(O, ) = y r V ( " ) / l NTY 1 (45)
the fractional extension of the TY segment in the unstretched network.
Then (45) may be written

This result shows that K does not depend on the detailed structure of the
network, but only on the number of segments with a given number of links,
and on their fractional extensions when the material is unstretched. If
F(N)dN is the number of segments per unit volume for which the number
of links lies between N and N +
d N , and G(N,X)dX is the fraction of these
segments which have fractional extensions between X and X dX, then +
102 E. GUT13

If one assumes that the extensions of the molecules have the same distribu-
tion as would be found in a systcni of uncoristrairicd rnoleculcs subject to
thermal motion, tlieii

Substitution into (47) gives K =


Lrn
F(N)dN = G, the total number of
segments per unit volume, independently of the form of F ( N ) .
Convincing arguments can be presented against the assumption of the
distribution (48). First, in reality the network is formed by processes
which involve segments constrained by their inclusions into the network
segments to which the distribution (48) does not apply. Second, the mean
extension of a segment in the network must decrease on the average, since
the segment is linked to a flexible, yielding and not to a rigid structure.
Finally, subdivision of the segments, with progressing cure as new bonds
are formed, will change the distribution.
If we do not make assumptions about the form of G(N,A),but assume
inst.ead that all the N , are equal
N,, = N (49)
then we can define

Remembering that according to eq. (19) for free, unconstrained segments

(51)

we can write

This is, according to (19), also possible if we have hindered rotation, in


which case will depend on the temperature. If F ( N ) is a narrow dis-
tribution, the preceding splitting of the double sum (44) into the two fac-
tors [To12 and l/[ro12 is still feasible, but only approximately. For a
broadly distributed F ( N ) the splitting is not permissible at all. Using
(52) it is possible to correlate K with solution properties (viscosity) of
unconstrained, unperturbed molecules. However, clearly, eq. (52) does
__.

not help
-
in the a priori determination of G(N,r(O))and thus of
If [rolZdepends on T,the “front factor” will differ from unity. We
shall prove, however, that even if this is not the case, and even if the
splitting is not feasible, still f, in general, will not be equal to unity.
STATlY'I'lCAL MECJIANlCS 103

Any realistic theory of the magnitude of the modulus must follow, so to


say historically, the gradual growth and reinforceiiient of the network by
the process of cure. Such a theory has been developed. The main as-
sumptions are: (a) the network is Gaussian and (b) the probability of
bonding between two elements of the system is proportional to the prob-
ability of their presence in the same small volume in space. These two
assumptions, together with an assumption as to the lengths of the chains
the material before cure, should determine K as a function of the total
number of bonds formed in the material during cure. Owing to mathe-
matical difficulties connected with the division of the network into active
part and loose ends we obtain a relation between K and the number of
bonds formed between elements of the network as it exists a t the time of
bond formation. This is a fraction of the total number of bonds that we
can only estimate. The theory is carried out in three steps:
1. One computes the relative probability that the next bond to be formed
within the network will be formed between any two given points of the
network.
2. Next, one computes the change in entropy of the system which would
result from the formation of such a bond.
3. Finally, one computes the average change in K per bond formed
between elements of the network, a t this stage of the cure. For the averag-
ing a distribution has to be assumed giving the relative number of bondable
element pairs in the whole network. The relation of the two elements is
described by two parameters: the separation of their mean positions in
the material and another parameter, which determines the variability of
their separation. Fortunately, special properties of the Gaussian network
lead for a large class of distribution functions to the same result of the
averaging process.
The main conclusion is that the average change in K , per bond formed
within the network, is equal to 1, independently of the detailed structure
of the network or the state of cure of the material. Then

K = B, (53)

where B, is the total number of bonds, which, during the process of cure,
were established between pairs of elements in the network as it then existed,
but exclusive of those established between elements of the same loose end.
We shall indicate now in detail, how B, can be related to G, the number
of chains in a unit volume. If B bonds are introduced into a system of N
unbranched chains, then we have the relation

G = 2 B f N (W
This is so because each bond introduced into the system increases the
number of chains by two (three) if it connects points of two different chains
(the same chain). Let an active network, stripped of all loose cnds, contain
10.1. E. GUTH

G, segments. Let there be J junctions in this network. Each of these


junctions will form the end of three or four segments. Clearly, then

Each chain attached to the network will include a t least two (terminal)
segments, which are not part of the active network. If all chains are
attached to the network, one has also
GU<G-2N=2B-N (56)
The degree of inequality rises with increasing number of loose ends. G, is
a structural quantity, determined by observation of the completed network.
I n contrast B, is a historical quantity; its determination implies the knowl-
edge of the history of the network. B, is the number of bonds introduced
as intranetwork bonds when they were formed. The same final network
can be formed by alternative processes for which B, has different values.
Thus between B, and G, there can be only a statistical relation.
A bond formed between different free ends increases B, by 1 and G,L
by 1 or a greater value-if the free ends have a complex structure. A bond
inside one chain of the active network increases B, by 1, but leaves G,
unchanged. A bond between two different chains of the active network
or between a free end and a chain of the active network increases B, by 1,
and G, by 2 or more. Since these latter processes predominate in the later
stages of the cure, we have
G, Z 2B,, 6 (57)
and, consequently
K g G,/2 G B,.
Thus the modulus according to this theory differs from that based on use
of eq. (48)by a factor 1/2. The resulting equation of state may be written
2 = B,RT[L - 1/L2] (59)
or for the increase in tension as the cure proceeds
AZ = AB,RT[L - l / L 2 ] (60)
In experimental investigation16 of the elasticity of irradiated poly-
dimethylsiloxane, it was found that the modulus, considered as a function
of the number of crosslinks per unit volume introduced by the irradiation,
is only half of the value predicted by a theory based on eq. (48). It was
observed that, in the later stages of cure, the increase in tension in a sample
of unit original cross section, stretched to relative length L, is given by
AZ = ACRTIL - 1/L2] (61)
where AC is the increase in number of moles of crosslinks per cubic centi-
meter. Now, during the early stages of cure Mais considerably less than
STATISTICAL MEClIANlCS 105

AC, since many of the crosslinks being formed involve molecules not pre-
viously incorporated into the network. In the later stages of cure, most of
the molecules have been incorporated into the network, and Ioose ends
make up only a minor part of the network. Under these conditions A B
= AC, and our theoretical result reduces to the observed form.
We presented this comparison with experiment, though the importance
of the theory stands with its logical consistency even if the assumed random
bonding would not occur in polymers. The theory presents the simplest
model for the increase of the modulus during cure. Needless to say, if
crosslinking can occur only at a limited number of points along the chains,
then the relation (53) is not valid. Furthermore, if these bondable points
are far apart, K will be considerably greater than B,.

7. MODEL FOR REAL POLYMERS WITH


MOLECULAR INTERACTIONS16
In the theory of gases the first step from ideal to real systems was van der
Waals’ equation. Now eq. (29) can be written in a form which brings out
the analogy to van der WaaIsl equation (neglecting the small term PP)

Zu is the contribution to Z due to internal energy, while Z s is the contribu-


tion due to entropy. (If we take account of change of volume on stretch-
ing, then the enthalpy replaces the internal energy.)
A model is briefly described in which the shape dependence of the internal
energy, and its effect on the thermoelastic behavior of rubber, are taken
into account to some extent. Intermolecular forces having the general
character of van der Wads’ forces are explicitly introduced in this model.
The model is admittedly crude, and the discussion of it is intended pri-
marily as an initial exploration of a difficult field. The theoretical treat-
ment of the model, however, is self-consistent.
There are several factors that can contribute to the change of internal
energy as rubberlike materials are stretched :
a. Change in the average extension of the component molecules, with
consequent change in the average intramolecular energy.
b. Change in the relative orientations of adjacent segments of different
molecules, with consequent change in the average potential of van der
Waals’ forces between the molecules.
c. Change in the fraction of the material that is crystallized.
Factor (a) will be neglected entirely, by treating the individual chains as
consisting of links of fixed length that take on their orientations independ-
ently; (b) and (c) are intimately related and will be taken into account.
Assume that the forces between adjacent links of neighboring chains can
be resolved into two parts-radial forces, which keep the molecules apart,,
106 E. GUTH

and torques, which tend to determine their relative orientations. The


radial forces will determine the average separation of neighboring links and
thus the volume, thermal expansion, and compressibility of the material.
Consideration in detail is not given to these forces, which are taken into
account as in the earlier theory, but attention is directed to the forces
which tend to establish definite orientations of next-neighbor links with
respect to each other. It is assumed that each link has the same number,
z, of next-neighbors, and any dependence of orienting forces on the distances
between next-neighbors is neglected; in effect use is thus made of a quasi-
crystalline model of the local structure of rubber. The crystallization of
rubber indicates the existence of forces tending to make the molecular
chains parallel, but one would expect that in nonpolar materials an anti-
parallel orientation of neighboring links would be energetically about as
favorable. An attempt is made to represent the interaction of neighboring
links by an interaction potential:

where e12 is the angle between the links. This assumption has been made
only because it seems to be the simplest assumption of generally suitable
character; one would expect that the actual potential changes more rapidly
with increasing deviation from parallelism or antiparallelism of the molecu-
lar chains.
The effect of these interactions has been treated in what is essentially
the internal (Weiss) field approximation in the theory of ferromagnetism.
The interaction of each link with its neighboring link is taken to determine
an effective potential energy depending on the orientation of this link.
This effective potential energy is computed, for any orientation of the link
in question, as the time-average of its energy of interaction with all adjacent
links-not only the links adjacent to it a t one instant, but all links which
may become adjacent to it during the motion of the molecules. This
effective potential is of the form

where e is the angle between the axis of the link and the direction of stretch,
and A is a constant that is the same for every link. This effective poten-
tial plays much the same role in this theory as the Weiss field plays in the
theory of ferromagnetism.
This model shows a first-order transition, an orieritational melting with
finite latent heat. This is not discussed her(>, nor are the details of the
rnatheniatical treatment of the model,which has considerable methodological
interest. Rather, the resulting stress-strain curve for ternpertrturcs well
above the melting range is treated. We obtain

z = D ~ [-L 1 p 1 + a[i- - 2 0 ~ 1 ~-
1 / ~ 3 1 3 I/L~I (65)
S1'ATISI'ICAL MECIIAN ICS 107

where

1
KkTX
D2 =
5[(5kT/Bz) + 11 > O
I n equation (66) M is the number of links per unit volume, [h(O)I2 is the
mean-square fractional extension of the chains (between junctions in the
network) in the unstretched state. B is the interaction constant of equa-
tion (63) and z is the number of next neighbors of each link. The quantity
+
[(5kT/Bz) 11 is negative and riot too small, so that Dz > 0 as stated in
eq. (66).
It is worth noting that the entropy force 2, = T ( b Z / b T ) , and the in-
ternal energy force Zu = 2 - 2 s corresponding to eqs. (65 and 66) are
both affected by the link interactions:

2s = K k T [ L - 1/L2] = KkT >( ~ [X(0)lz


5
[(ZkTIBz) + 1Ip2
x 3 11~31- [I -
( 2 [ ~- 1 / ~ 3 ] } (67)
516 T
Zu = K k T [ h ( 0 ) ]X2 ~ [(5kT/Bz)
BZ
+ l]-2(2[L3- l/L3] - [l - l/L3]1 (68)
Thus i t is not entirely justifiable to analyze the stress of real rubbers into
entropy and internal energy terms, and to compare the entropy part with
the predictions of a theory in which link interactions are entirely neglected.
It should also be mentioned that both imperfect flexibility of the chains and
change of volume of the material with stretch will contribute to the ob-
served Zu.
The first term in eq. (65) comes from the Gaussian distribution function
for the chains; the second and third terms are new. Terms with the
same dependence on I, arise in the non-Gaussian theory.
Thus, the main difference between eqs. (65) and (40) lies in the different
temperature dependence of the coefficients and in the opposite sign of the
coefficients of the term [L3 - l/L3]. For application to real rubbers,
eqs. (65) and (40) should be combined to give the following stress-strain
relation:
z = D J L - I/LZ]+ (02'+ 2/3Dz)[1- 1 / ~ 3 1
+ (Dz' - 202)[7,3 - 1/L3] (69)
If accidentally at a certain teniperature D,' = 2D2then ey. (40) reduces to
z = D ~ [ L- 1/1,21 + ( D ~ +
' 2/302)[1- 1 / ~ 3 1 (70)
While such an exact cancellation is unlikely, the possibility of a partial
cancellation secins to be of interest.
108 E. GUTH

I t is a characteristic of both the non-Gaussian theory arid the present


theory that the coefficients of the second and third terms are interrelated
and their dependence upon temperature is fixed. This is not the case for
phenomenological approaches.
We would Iike to refer to a recent interesting non-Gaussian cubic model
taking account of molecular interactions.l7
Space limitations do not permit a detailed comparison of the theory of
network elasticity, as outlined in this article with experiment. We have to
refer to some recent literature.ls We wish to emphasize, however, that in
all recent work our “front factor” in the form of eq. (52) is used.

References

1. Among the almost innumerable contributions of Hermann Mark to polymer


science his contribution to establish the foundations of polymer physical chemistry and
physics stands out. After Staudinger established the foundation of polymer chemistry
by his proof of the existence the long-chain molecule, Meyer and Mark contributed
greatly to the clarification and development of polymer chemistry. The basic founda-
tions of polymer physical chemistry and physics were laid down in the independent
contributions by E. Guth and H. Mark, Monatsh. Chem., 65, 93 (1934) and W. Kuhn,
Kolloid-Z., 68, 2 (1934). Both of these papers developed the statistics of long-chain
molecules. Guth’s and Mark’s chief emphasis was on the clarification of the mechanism
of rubber elasticity and they worked out the first quantitative theory which accounted,
in a general way, for some of the peculiar properties of rubber. Connection was estab-
lished with Rayleigh’s [Phil. Mag., 37, 321 (1919) or Scienti3c Papers, Vol. VI, p. 604;
cf. also S. Chandrasekhar, Rev. Mod. Phys., 15, 1 (1943); M. C. Wang and E. Guth, loc.
cit., ref. 51 work on random walk. Kuhn’s main application was on the viscosity of
solutions. For earlier history of qualitative ideas about rubber elasticity cf. E. Guth,
H. M. James, and H. Mark, Advances in Colloid Science, Vol. 11, Interscience, New York,
1946.
2. Volkenstein, M. V., Conjigurational Statistics of Polymeric Chains, Interscience.
New York, 1963. We refer to this comprehensive book for a more complete set of
references than we can quote in this article; cf. also D. ter Haar, Rinehart, New York
(1954), loc. cit. Chap. XV; A. H. Wilson, Thermodynamics and Statistical Mechanics,
Cambridge University Press (1957), Zoc. cit. Chap. XIV. A. J. Staverman, Thermody-
namics of Polymers, Encyclopedia of Physics, Vol. XIII, ed. S. Flugge, Springer, 1962.
L. R. G. Treloar, The Physics of Rubber Elasticity, 2nd ed., Oxford, Clareridon Press, 1958.
3. Volkenstein, M. V., ref. 2, p. 2.
4. Uhlenbeck, G. E., and G. W. Ford, Lectures in Statistical Mechanics, Amer.
Math. Soc., Providence, Rhode Island, 1963, gives a brief and lucid presentation of funda-
mental principles and applications.
5. Wang, M. C., and E. Guth, J . Chem. Phys., 20, 1144 (1952); cf. also, A. Isihara,
N. Hashitsume, and M. Tatibana, J. Chem. Phys., 19, 1508 (1951).
6. James, H. M., and E. Guth, J. Chem. Phys., 11, 455 (1943). The Langevjn
function appears in this context first in: H. M. Jamesand E. Guth, Phys. Rev., 59, 111
(1941).
7. This is discussed in great detail in Volkenstein, ref.2 Ch. 3 and 6.
8. (a) James, H. M., and E. Guth, J . Chem. Phys., 11,455 (1943); (b) H. M. James
and E. Guth, J. Polymer Sei., 4, 153 (1949); (c) H. M. James and E. Guth, J. Chem.
Phys., 21, 1039 (1953). We follow mostly ref. (b). The Appendix to ref. (a) contains,
however, a detailed discussion of the average forces exerted by the network segments.
Equation (29) appears first: E. Guth and H. M. James, Znd. Eng. Chem., 33,624 (1941)
STATISTlCAL MEClIANlCS 109

for V = 1, and in H. M. James slid E. Gnth, Ind. Eng. Chem., 34, 1365 (1942) for 1. = 1
+ ol(T - To).
9. James, H. M., aiid E. Guth, J . Chew&.Phys., 15, 669 (1947); eq. (46) has been
given first by H. M. James and E. Guth, J . A p p l . Phys., 15, 294 (1944) ey. (7).
10. cf. ref. 8 (b), part I11 and Appendix B.
11. James, H. M., J . Chem. Phys., 15, 651 (1947).
12. The derivation in Treloar’s book, ref. 2 pp. 66-70 is identical with the derivation
given here and originally in ref. 8 (b). However, here and originally in ref. 8 (b), this
derivation is preceded by a treatment not assuming the junctions to be fixed. Only
this way can the meaning of the assumptions of fixed junctions and a f h e deformation
be made physically clear.
13. We follow ref. 5 part 11.
14. We follow ref. 9.
15. Bueche, A. M., J . Polynier Sci.,19, 297 (1956); cf. also E. Guth arid H. hI.
James, J . Polymer Sci., 24, 479 (1957).
16. This summary is taken from Proceedings of Third Rubber Technology Conference,
Heffner, London (1954), p. 364. This reference seems to have been overlooked in the
literature.
17. Krigbaum, W. R., and M. Kaneko, J . Chem. Phys., 36,99 (1962).
18. Treloar, L. R. G., ref. 2; P. J. Flory, C. A. J., Hoeve, and A. Ciferri, J . Polymer
Sci.,34,337 (1959); A. Ciferri, J . Polymer Sci., 54, 149 (1961); A. V. Tobolsky, D. W.
Carlson, and N. Indictor, J . Polymer Sci., 54,175 (1961); R. J. Roe and W. R. Krigbaum,
J . Polymer Sci.,61, 167 (1962); K. J. Smith Jr., A. Greene, and A. Ciferri, Kolloid-Z.,
194, 49 (1964). In the last reference the formula for the entropy of a non-Gaussian
network contains a factor (1 - l/N), which is questionable: K. J. Smith, Jr., A. Ciferri,
and J. J. Hermans, J . Polymer Sci.,A2, 1025 (1964). A consistent approximation is our
eq. ( 3 i ) with or without terms (38), depending on whether the junctions are assumed
to be free or fixed.

S-ar putea să vă placă și