Chapter 4 Polymer Physics

Molecular Motion in Dilute Solution

Chapters 4 and 5 of Doi deal with Dynamics of dilute and semi-dilute to concentrated polymer
systems. Doi terms the latter two "Entangled" to emphasize the importance of topological states on
polymer dynamics at high concentrations. Dynamics of chains are involved in any experiment
where time is a factor. These include dynamic light scattering, elastic neutron scattering,
rheological measurements, and dynamic mechanical measurements. The properties of interest are
of a visco-elastic nature, with always involves some kind of time constant associated with a
relaxation.

The simplest approach to the dynamics of polymers is the Rouse, Bead and Spring model which
was already introduced in the discussion of rubber elasticity. Doi approaches this by first giving
an overview of Brownian motion which serves as a basis to address the Rouse Model.

Brownian Motion:

A small spherical particle suspended in a solvent will display random thermally driven motion such
that the average velocity for a number of particles (or for a single particle over time) as a function
of time is 0, <V(t)> = 0, since the velocity is random. The velocity in the x-direction can be
considered for simplicity. At two times, t1 and t2 , the product of two velocities of the same particle
will have the value V(t1 )V(t2 ) and the average of this function is not necessarily zero, <V(t1 )V(t2 )>
again, the average being over a single particle in time where only the difference
(t1 - t2 ) is of importance, or for a number of particles. We can write,

<V(t1 )V(t2 )>= C v (t1 - t2 )

where C v (t) is the velocity correlation function. The decay of Cv (t) is described by a velocity
correlation time, τV.

For time-invariant, random processes such as Brownian motion (i.e. for processes where there is
no physical meaning to t = 0 because all times are basically equivalent) the velocity correlation
function has a simpler expression,

<V(t1 )V(t2 )>= <V(t)V(0)> = Cv (t)

At short time differences Cv (t) will have the value of the mean square velocity, <V2 (t)>. For long
time differences this function will approach 0 (<V(t)> 2 ) as the time correlations between velocities
go to 0.

The mean square velocity, <V2 (t)>, is due to the thermal motion of the particles and is smaller for
larger mass particles,

<V2 (t)> = kT/m

where m is the mass and T is the temperature.

The spherical particle experiences a viscous force which dampens the Brownian fluctuations which
lead to its velocity. This viscous force is the product of a friction coefficient, ζ, and the velocity.
Stokes law defines the friction coefficient for a spherical particle as

ζ = 6πηsa

1
for a sphere of radius "a" and a solvent of viscosity ηs. This force balances with the
mass*acceleration of the particle,

m (dV/dt) = -6πηsa V

This equation describes an exponential dampening of the velocity,

V = V0 exp(-6πηsa (t1 - t2 )/m) = V0 exp(-(t1 - t2 )/ τV)

so τV = m/(6πηsa)

For a sphere of density ρ, m = (4π/3)a3 , and τV = (2π/9)ρ a2 /ηs.

The relaxation time, τV, is very small for colloidal scale objects such as polymer coils, O(10-10 s) for
normal solvent viscosity's and polymer sizes. This is many orders of magnitude smaller than
observable time scales in the experiments mentioned above, dynamic light scattering, rheology,
and dynamic mechanical experiments. This means that, for all accessible time scales, for a
polymer in dilute solution, it is a safe assumption to make the approximation τV ≈ 0.

We can consider the value of Cv (t) for accessible time scales for a colloidal scale particle
undergoing Brownian motion in solution under the assumption that τV ≈ 0. The displacement of
the particle, ξ(t), is given by the integral of the velocity over time,
t

ξ(t ) = ∫ V(t')dt'
0

The mean-square displacement, <ξ(t)2 >, for random diffusion is 2Dt (from the Central Limit
Theorem and a Gaussian distribution in velocities as shown by Doi, pp. 67),
t t t t

ξ (t ) = ∫ dt1 ∫ V (t1 )V (t2 ) dt2 = 2Dt = ∫ dt1 ∫ 2Dδ (t1 − t 2 )dt2

0 0 0 0

Which leads to the association,

<V(t1 )V(t2 )>= <V(t)V(0)> = 2Dδ(t)

where D is the diffusion coefficient, and δ(t) is the delta function.

Brownian Motion in a Potential Field:

In many dynamic measurements a particle is subjected to a potential field of some kind, i.e. in
rheology and in photon correlation spectroscopy. For this reason it becomes necessary to consider
the effect of a potential field, U(x), on the Brownian motion of a particle. For most purposes it is
convenient to consider U(x) as a "smoothly" varying function of position. The force experienced
by the particle due to this smoothly varying field is -dU/dx which is converted to an net average
velocity
2
<V> = -(1/ζ)dU/dx = <dx/dt>,

where ζ is the friction constant for the particle discussed above, after a time τV (≈ 0) has passed.

The mean velocity of a particle subjected to the potential field U(x) is subject to random variation
due to Brownian motion which can be assumed to linearly add to the mean velocity caused by the
field in the absence of inertia and non-linear effects,

dx/dt = -(1/ζ)dU/dx + g(t)

where g(t) is a Gaussian function identical to V(t) given above, <g(t)> = 0 and <g(t)g(t')> =
2Dδ(t-t'). The linear sum of the velocity due to a potential field and that due to Brownian motion is
referred to as the Langevin Equation.

Fluctuation-Dissipation Theorem:

The response of a system to an external force (constitutive parameter) can be used to describe the
dampening of thermal fluctuations. (i.e. thermal fluctuations are dissipated by the same
constitutive parameters which govern the response of the system to external forces.) Doi uses this
approach to describe the relationship between the diffusion coefficient and the friction coefficient,
ζ. We will take this as a general assumption of the fluctuation-dissipation theorem.

D = kT/ζ

This equation for the diffusion coefficient is known as the Einstein relationship.

The position of a particle undergoing Brownian motion, x(t+∆t), is given by discretizing the
Langevin equation given above,

x(t+∆t) = x(t) - (1/ζ)dU/dx ∆t + ∆G(t)

where ∆G(t) is a probability distribution function for the randomly varying probability function
g(t) which describes Brownian motion in the Langevin equation,
t +∆ t

∆G(t ) = ∫ dt' g( t' )

∆G(t) is a probability distribution function for the Gaussian term of the Langevin linear sum. The
position, x(t) which is a result of the potential U and this Brownian, Gaussian term also has a
probability distribution function, ϕ(x,t), which satisfies the equation,

∂ϕ ∂  1 ∂U ∂ϕ 
= ϕ + D
∂t ∂x ζ ∂x ∂x 

The Einstein relationship can be used in this expression for the probability distribution function of
position of a particle under a potential field to yield,
3
∂ϕ ∂  1 ∂U ∂ϕ 
= D ϕ+
∂t ∂x  kT ∂x ∂x 

Systems with Many Degrees of Freedom:

An isolated polymer molecule represents a system with many degrees of freedom. Consider the
position of each mer unit being described by a vector x i. Consider a system with many degrees of
freedom, {xi} = (x1 , x2 , x3 , ...x n ) subjected to a potential field, and ignoring Brownian
fluctuations,

dxi/dt = -Σ µ ij δU/δxi

where U is the potential energy, and µij are the components of the mobility matrix (similar to 1/ζ in
the equation for velocity of the sphere above) the summation is over "j". The mobility matrix
accounts for interactions between components of a system with many degrees of freedom.

Brownian motion can be accounted for in this approach through an expression for the probability
distribution function,

∂ϕ ∂  1 ∂U ∂ϕ 
∂t
= ∑ ∂x Dij  ϕ+
∂x j 
ij i  kT ∂x j

where

Dij = µijkT

The Langevin Equation becomes,

dxi/dt = - Σ µijdU/dxj + gi(t) + Σ δDij/δxj

where the summations are over "j".

The mean of gi(t) is

<gi(t)gj(t')> = 2Dijδ(t-t')

<gi(t)> = 0

Brownian Motion of a Harmonic Oscillator:

A harmonic oscillator has a potential described by a spring constant, "k",

U(x) = kx 2 /2

Using this potential in the Langevin equation we have,

dx/dt = -kx/ζ + g(t)

4
solving for x as a function of time,
t −k ( t −t' )
x( t) = ∫ dt' e
ζ
g(t')
−∞

The time correlation function of x(t) is given by,

x( t)x (0 ) = ∫ dt1 ∫ dt 2 exp[ −k( t − t1 − t2 ) / ζ ] g(t1 )g(t2 )

Using <g(t)g(t')> = 2Dδ(t-t') and D = kT/ζ,

<x(t)x(0)> = kTexp(-t/τ)/kspr

where

τ = ζ/kspr

For t=>0, <x2 > = kT/kspr , as predicted by a Boltzman distribution, Ψ equil α exp(-kspr x2 /2kT)

The mean square displacement after a time "t" is given by,

( x( t) − x(0)) = x (t )2 + x (0)2 − 2 x (t )x(0 )

= 2 x 2 − 2 x (t ) x( 0)
= 2kT / kspr (1− exp( −t /τ ))

For t=>0 this yields (2kT/ζ)t = 2Dt which is the expected result from the discussion of Brownian
motion above.

Rouse Theory:

As mentioned above, a polymer can be considered as a system with many degrees of freedom. In
Chapter 1 we mentioned that the potential energy for a bead and spring polymer model (the Rouse
model) is given by,
N
1
kspr ∑ ( Rn − Rn− 1)
2
U=
2 n=1

Which is a multibodied analogue of the Hookean spring model given in the previous section. The
beads in the Rouse model experience a drag force proportional to their velocity just as in the spring
model above, with a friction coefficient ζ for each bead. The Langevin equation for the Rouse
model is,

dRn 1 ∂U
=− + gn
dt ζ ∂Rn

5
The index for the chain goes from 0 to N with a special governing equations (dRi/dt) for beads 0
and N. For beads 1 to N-1 the governing equation is (combining the expression for U and the
Langevin equation),

dRn k spr
= ( Rn+1 + Rn− 1 − 2Rn ) + gn
dt ζ

For n=0 and n=N the same expression can apply if we use R-1 = R0 and RN+1 =RN.

If the chain is thought to be composed of infinitesimally small springs, then n can become a
continuous function. dR/dn corresponds to (Rn-1 -Rn+1 )/2 and d2 R/dn2 becomes (Rn+1+Rn-1 -2Rn ).
Then the governing equation becomes

dR k spr ∂ 2 R
= + gn
dt ζ ∂n2

and at n=0 and n=N dR/dn = 0 using the limits above. The second derivative of R with n (chain
coordinate) defines certain modes of motion for the chain, just as a guitar string has certain modes
of vibration.

Decomposition of the modes can be accomplished if the coordinate "n" is replaced with a set of
normalized coordinates based on the first order mode which is defined by the limit, dR/dn = 0
at N. This means one mode involves the length N. The normalized coordinates involve an integral
transform of R(n,t) into a Fourier equivalent, X p , where "p" is the Fourier equivalent of "n",

 πpn 
N
1
X p (t) = ∫ dn cos R(n,t)
N0 N 

where the modes "p" have integer values of 0,1,2,3,...

Then the governing equation becomes,

dX p k p,spr
=− Xp + gp
dt ζp

where ζp = 2Nζ, ζ0 = Nζ. For the p'th mode the spring constant is given by,

2 p2π 2k spr 6π 2kT 2

The relaxation time for the p'th mode of the Rouse chain is given by,

ζp ζN 2 b2
τp = =
k p p2 3π 2 kT

6
The relaxation time for the Rouse chain is predicted to follow N2 . Experimentally, the relaxation
time (as measured by NMR or dynamic light scattering for instance) is found to follow N3/df where
df is the mass-fractal dimension, 3/5 for good solvents. This is one of the failings of the Rouse
model and is due to hydrodynamic interactions which are not accounted for in the Rouse analysis.

Rouse Motion of the Center of Mass:

The zero'th mode, X 0 , describes the motion of the center of mass, RG(t)=(1/N)∫dn R(n,t), since
the cos term in the integral for X 0 goes to 1 for this mode. The mean square displacement for the
zero'th mode is given by,

< (X 0 (t)- X 0 (0))α (X 0 (t)- X 0 (0))β > = δαβ 2kT t/ζ0 = <(RG(t)- R G(0))2 > = 2kT t/(Nζ)

where α and β are different conformations of the chain or different identical chains.

From the first section the center of mean square displacement of the center of mass is related to the
diffusion coefficient for the center of mass, <(RG(t)- R G(0))2 > = 2Dt, so,

DG = kT/Nζ

The Rouse Model predicts the center of mass diffusion coefficient goes as the inverse of the
molecular weight. Experimentally, the center of mass diffusion coefficient is observed to go as
M-1/df , where df is the mass-fractal dimension of 3/5 in a good solvent.. This is a second of the
failings of the Rouse model and, again, is due to hydrodynamic interactions which are not
accounted for in the Rouse analysis.

Rouse Rotational Motion: