Prof.

Tibbitt Lecture 7 Networks & Gels

Lecture 7:
Boltzmann distribution & Thermodynamics of mixing
Prof. Mark W. Tibbitt – ETH Zürich – 14 März 2019

1 Suggested reading
• Molecular Driving Forces – Dill and Bromberg: Chapters 10, 15

2 Boltzmann distribution
Before continuing with the thermodynamics of polymer solutions, we will briefly discuss the Boltzmann distri-
bution that we have used occassionally already in the course and will continue to assume a working knowledge of
for future derivations. We introduced statistical mechanics earlier in the course, showing how a probabilistic
view of systems can predict macroscale observable behavior from the structures of the atoms and molecules that
compose them. At the core, we model probability distributions and relate them to thermodynamic equilibria.
We discussed macrostates, microstates, ensembles, ergodicity, and the microcanonical ensemble and used these
to predict the driving forces of systems as they move toward equilibrium or maximum entropy. In the beginning
section of the course, we discussed ideal behavior meaning that we assumed there was no energetic contribution
from interactions between atoms and molecules in our systems. In general, we are interested in modeling the
behavior of real (non-ideal) systems that accounts for energetic interactions between species of the system.
Here, we build on the introduction of statistical mechanics, as presented in Lecture 2, to consider how we can
develop mathematical tools that account for these energetic interactions in real systems. The central result is
the Boltzmann distribution, which provides probability distributions based on the underlying energy levels
of the system.
In this approach, we now want to consider simple models (often lattice models) that allow us to count
microstates but also assign an energy Ei for each microstate. We then use this information to calculate a
probability distribution for the likelihood that the macrostate assumes the proposed mircostates at equilibrium.
As an example we could consider the example below of a model, four-monomer ‘polymer’ chain constrained to
two-dimensional space that has four open conformations and one compact conformation.

To start, we must define the system and the associated energy levels. The system is a four-monomer chain
with two energy levels, where each energy level coresponds to the number of monomer–monomer interactions
that the chain can make. We will often take zero as the lowest energy state and refer to this as the ground
state. Disrupting the contact increases the energy of the system by  = 0 > 0. As an aside, we will use  for
intermolecular energies and E for macrostate energies, which are equivalent in this case but this is not the case
as we consider more complex systems. We refer to states other than the ground state as excited states. A
microstate refers to a specific conformation and each ground or excited state can be achieved by one or more
microstates. We thus seek to compute the set of proabilities {pi } = p1 , p2 , . . . that the model chain adopts

1
Prof. Tibbitt Lecture 7 Networks & Gels

the various possible conformations. We will come back to this example briefly after we derive the Boltzmann
distribution law below.
For this derivation we introduce the Canonical Ensemble, which assumes constant temperature T , volume
V , and components {Ni }. As before each system has many accessible microstates and observable macrostates.
The number of microstates in macrostate i is given as ni and with N total microstates we can write the
probability pi = nNi . From Lecture 2 Eq. 8, we know that:
S X
=− pi ln pi . (1)
kB i

As before, we are interested in predicting the macrostate at equilibrium, which we will determine when the
entropy is maximized. Stating this mathematically, we want to find the set of probabilities {pi } such that
dS = 0 with the following constraints:

X
pi = 1 (2)
i
X
pi Ei = hEi = Ē (3)
i

We can then use the method of Lagrange multipliers to solve for dS = 0 from Eq. 1 and the constraints in
Eqs. 2 and 3.
dS X X
=− [ln pi + 1] dpi = − [ln pi ] dpi , (4)
kB i i
X
dpi = 0, (5)
i
X
[pi dEi + Ei dpi ] = 0. (6)
i

In Eq. 6 the first term goes away as dEi ≈ 0, assuming that the energy levels have little dependence on T , V ,
{Ni }. Therefore, we can solve the set of equations with the Lagrange multiplier approach as follows:
X
[− ln pi − α − βEi ] = 0. (7)
i

At the global maximum, each term in the sum in Eq. 7 must be equal to zero so we have:

ln pi = −α − βEi , (8)
−α −βEi
pi = e e . (9)
P
Next, we can determine α with a normalization constraint by dividing by 1 = pi :
i

e−α e−βEi
pi = P −α −βEi (10)
e e
i
e−βEi e−βEi
= P −βEi = , (11)
e Q
i

e−βEi is defined as the partition function. Additionally, we can define β with the constraint
P
where Q ≡
P i
that pi Ei = Ē.
i

2
Prof. Tibbitt Lecture 7 Networks & Gels

X
dĒ = [Ei dpi + pi dEi ] (12)
i
X
dS = −kB ln pi dpi (13)
i
X
= −kB dpi [−βEi − ln Q] (14)
i
X
= βkB Ei dpi (15)
i
= βkB dĒ (16)
Therfore, we can connect with what we have shown about thermodynamic driving forces, namely:
 
∂S 1 dS 1
= ∴ = βkB ⇒ β = . (17)
∂U T,V,{Ni }} T dĒ kB T
This brings us to the common expression of the Boltzmann distribution law:

e−Ei /kB T
pi = . (18)
Q
As before, this distribution was calculated by maximizing entropy!
We return briefly to the model chain presented before. We can calulate the partition function as Q =
−0 /kB T
1 + 4e−0 /kB T and the probabilities as pc = Q
1
and po = 4e Q . What is expected to occur at high and low
T?

3 Thermodynamics of solution
As we move toward considering materials constructed from polymer chains, including solvent swollen networks
or gels, we need to consider how single polymer chains behave in solution. This can be polymer chains in a
‘solution’ of polymer chains or a more classic view of solution with polymer chains in a solvent. The approach,
as before in this course, will be to develop simple models that allow us to calculate the thermodynamics of
polymer solutions and, today, we start with simple solutions as we build toward this end.
To do this, we will use the Canonical Ensemble discussed above (constant T , V , {Ni }) and calculate the
Helmholtz free energy F = U − T S on a lattice model where S is the entropy of solution and U accounts for
the interaction energies between molecules on the lattice. Here, we will consider a lattice with N sites and nA
molecules of type A and nB molecules of type B. As before molecules A and B are equivalent in size and each
occupies one lattice site. We can visualize the problem as below:

First, we can calculate the entropy of solution as before by calculating the multiplicity as the total number
of spatial arrangements of the system W = nAN!n!B ! . Therefore, we can use the Boltzmann equation (Lecture 2)
and Stirling’s approximation:

3
Prof. Tibbitt Lecture 7 Networks & Gels

∆Ssol = kB (N ln N − N − nA ln nA + nA − nB ln nB + nB ) (19)
= kB (nA ln N + nB ln N − nA ln nA − nB ln nB ) (20)
hn nA nB nB i
A
= −N kB ln + ln (21)
N N N N
= −kB (nA ln xA + nB ln xB ). (22)

where xA = nA /N and xB = nB /N . This can also be written with the mole fraction of one species, say x = xA :

∆Ssol
= −x ln x − (1 − x) ln(1 − x) . (23)
N kB
We can then calculate the energy of solution. In our lattice model, this involves counting the sum of the
contact interactions between all of the nearest neighbors in the solution. In the system described above, we can
consider three types of interactions: AA, BB, and AB.

The total energy of the system is then:

U = mAA wAA + mBB wBB + mAB wAB , (24)

where mAA , mBB , and mAB are the number of AA, BB, and AB interactions, respectively, and wAA , wBB ,
and wAB are the interaction energies with the associated contacts. For this, we take the interaction energies to
be negative.
We then need to count the number of AA, BB, and AB interactions, which are not known in general. A
convenient approach is to express these in terms of nA and nB . Additionally, each lattice site has z contacts
available. Therefore,

znA = 2mAA + mAB , (25)

and similarly

znB = 2mBB + mAB . (26)

We can rearrange Eqs. 25 and 26 to solve for mAA and mBB :

4
Prof. Tibbitt Lecture 7 Networks & Gels

znA − mAB
mAA = , (27)
2
znB − mAB
mBB = . (28)
2
We can combine these results with Eq. 24 to arrive at:

   
znA − mAB znB − mAB
U= wAA + wBB + mAB wAB (29)
2 2
 zw   zw   
AA BB wAA + wBB
= nA + nB + wAB − mAB (30)
2 2 2

To compute U , we still need to calculate mAB and to do this we will employ a mean-field approximation
as before. We make the assumption that for any number of molecules the system is random and uniformly
dispersed, this allows us to more easily estimate mAB .
To do this, we consider a specific site on the lattic next to an A molecule. The probability that a B molecule
occupies the neighboring site can be calculated by assuming that B molecules are distributed randomly across
the lattice. Therefore, the probability that a given site is occupied by a B molecule pB is given by:
nB
pB = = xB = 1 − x. (31)
N
On this lattice, there are z nearest-neighbors so on average for each A molecule so the average number of AB
contacts on each A molecule is znB /N = z(1 − x). There are a total of nA A molecules, so
znA nB
mAB ≈ = zN x(1 − x). (32)
N
Therefore the total contact energy of the solution in terms of nA and nB can be written as:

 zw  zw  
AA

BB
 wAA + wBB
U= nA + nB + z wAB − N x(1 − x) (33)
2 2 2
 zw   zw 
AA BB
= nA + nB + kB T χAB N x(1 − x) (34)
2 2
where we define a dimensionless parameter χAB = kBz T wAB − wAA +w


2
BB
called the interaction parameter.
Then we can compute the full free energy of solution:

F (nA , nB ) nA nB zwAA zwBB

= nA ln + nB ln + nA + nB + χAB N x(1 − x). (35)
kB T N N 2kB T 2kB T
In practice, we are most interested in computing a free energy difference between the solution and the initial
pure states of A and B:

∆Fsol = F (nA , nB ) − F (nA , 0) − F (0, nB ). (36)

Thus, we arrive at the full free energy difference as a function of the mole fraction x and the interaction
parameter χAB :

∆Fsol
= x ln x + (1 − x) ln(1 − x) + χAB x(1 − x) . (37)
N kB T
This is the regular solution model originally developed by Hildebrand and demonstrates that while ideal
solutions are driven purely by entropy, we must also consider the interactions of molecules when computing the
free energies of real solutions. We will build on this in the next lecture as we work toward an understanding
of the thermodynamics of a polymer in solution, where one species in many times larger than the other species
around it.