You are on page 1of 79

Liquid State Theory

Statistical Mechanics of Liquids


and Complex Fluids
George Jackson

Department of Chemical Engineering


Imperial College London
South Kensington Campus
London SW7 2AZ

g.jackson@imperial.ac.uk
http://www3.imperial.ac.uk/people/g.jackson
Summary
A molecular description of matter using statistical mechanical theories and
computer simulation is the key to understanding and predicting the properties of
dense fluids and materials. Fluid systems form an integral part of our modern
lifestyle from the use of simple solvents in chemical processing to the design of
opto-electronic devices with liquid crystalline and polymeric materials.

As well as being of inherent scientific interest, the link between the microscopic
interactions between small numbers of molecules and the macroscopic
properties of the bulk system comprising them is of great industrial relevance.

In this part of the course the main theoretical achievements in liquid state theory
will be outlined. This with range from the original approach of van der Waals
(1873), through the use of correlation functions to describe the properties of
fluids (1950s-1980s), to the more sophisticated perturbation theories (1970s-
date). The most recent developments in statistical mechanical theories which are
used to probe the bulk or interfacial properties of fluids will be reviewed making
contact with the description of real systems wherever possible. The focus will be
on the structural and thermodynamic properties, with an emphasis on phase
equilibria.

2 G. Jackson – Imperial College London


Synopsis of Lectures
1. Equilibrium Structure:
1.1 particle densities and distribution functions
1.2 pair distribution function

2. Thermodynamics:
2.1 using the language of distribution functions
2.2 internal energy
2.3 pressure virial (as opposed to virial expansion)

3. Perturbation Theories:
3.1 van der Waals (1873)
3.2 Zwanzig (1954)
3.3 Carnahan and Starling and the hard-sphere model (1969)
3.4 Wertheim/SAFT (1980-1990s) for associating and non-spherical
molecules
3.5 Examples of the description of real complex fluids and mixtures
(from water to polystyrene solutions)

3 G. Jackson – Imperial College London


Recap of Statistical Mechanical Concepts
Total energy of classical system of N indistinguishable particles:

E (r , p ) = pi + U (r )
rN r N 1 N 2 rN

2m i =1
(1)
kinetic + potential energy
Canonical partition function:
...∫ exp(− β E (r , p ))d r d p where β = 1 /(kT )
1 rN r N rN r N
3N ∫
QNVT =
N!h
=
1
N!Λ 3N ∫
exp (− β U (rN
r )) d
rN
r integrating over momentum (2)

= NVT3 N configurational integral Z NVT = ∫ exp(− β U (r N ))d r N


Z r r
N!Λ

Helmholtz free energy: F = −kT ln QNVT (3)

r rr r r r r r r r
r N ≡ r1 r2 r3... rN p N ≡ p1 p2 p3... pN
4
positions momenta G. Jackson – Imperial College London
Ensemble average of variable in the canonical ensemble:
X NVT = ∑ X i Pi
i

...∫ X (r , p )exp(− β E (r , p ))d r d p


rN r N rN r N rN r N
=

∫ ...∫ exp(− β E (r , p ))d r d p
rN r N rN r N classically

= ∫ ...∫ X (r , p )P (r , p )d r d p
rN r N (N ) rN r N rN r N
(4)

Pi ∝ P (r , p ) probability of finding a system in a state i


(N ) rN r N

A probability density in the canonical ensemble can thus be


defined (cf. equation 2) as
exp(− β E (r , p ))
rN r N
P (r , p ) =
(N ) rN r N

∫ ...∫ exp(− β E (r , p ))d r d p


rN r N rN r N

exp(− β E (r N , p N ))
r r
1 (5)
= 3N
5
N ! h QNVT G. Jackson – Imperial College London
From the definition of the probability distribution function
given in equation 5 it follows that
P (r , p )d r d p = P (r , p )d r1 d p1... d rN d pN
(N ) rN r N rN r N (N ) rN r N r r r r

is the probability of finding the system in the phase space


rN r N rN r N
element d r d p about configuration r p , i.e., particle 1
r r
in volume element d r1 and momentum element d p ...
r 1
r
particle N in volume element d rN and momentum d pN .
r r r r
r2 , p2 r4 , p4

r r r r
r12 , p12 r7 , p7

r r
r10 , p10
System configuration
r r r r
r1 , p1 r9 , p9 rN r N
r p
r r r r
r8 , p8 r3 , p3
r r
(state i)
r r rn , pn
r11 , p11
r r r r
r6 , p6 rN , pN
r r
r5 , p5
6 G. Jackson – Imperial College London
The momentum can again be integrated out into a kinetic
contribution to define the probability density solely in terms
of the positions (cf. equations 2 and 5) as
exp(− β U (r ))
rN
P (r ) = ∫ P (r , p )d p =
(N ) rN (N ) rN r N rN
exp(− β U (r ))d r
rN rN (6)

P ( N ) (r N )d r N = P ( N ) (r N )d r1 ... d rN is now the probability of
r r r r r
rN rN
finding the system in the volume element d r about r ,
r r
i.e., particle 1 in d r1 ... particle N in d rN , irrespective of
momenta. r
r r4
r2
r r
r12 r7

r
System configuration
r10 r r
r1 r9
r
rN
r r
r8 r3
r
r rn
r11
r r
r6 rN
r
r5
7 G. Jackson – Imperial College London
For applications involving identical particles it is convenient
to define a generic probability density ρ (r ) such that
(N ) rN

ρ (r )d r = ρ (r )d r1 ... d rN is the probability of finding


(N ) rN rN (N ) rN r r
rN rN
the system in the volume element d r about r ,
irrespective of the particle labels. As there are N! possible
permutations
ρ (r ) = N ! P (r ) = N !
(N ) r N (N ) r N exp (− β U (rN
r )) (7)
∫ exp(− β U (r ))d r
rN rN

System configuration
r
rN
regardless of particle
labels.
8 G. Jackson – Imperial College London
1. Equilibrium Structure
1.1 Particle Densities and Distribution Functions
The full probability densities P (r , p ) and ρ (r ) are
(N ) rN r N (N ) rN

high-order functions ~ O(N). The properties of the system can


however be expressed in terms of lower-order or reduced
particle densities and distribution functions.

The reduced generic particle density ρ N (r ) is proportional


(n) n r
to the probability of observing any set of n particles in the
r
configuration r n and is defined by integrating out the
coordinates of the remaining N − n particles
ρ N (r ) =
(n) r n
P (r )d r
(N ) rN
N! r N −n
(N − n )! ∫ (8)
N ! ∫ exp(− β U (r ))d r
rN r N −n
=
(N − n )! ∫ exp(− β U (rr N ))d rr N
9 G. Jackson – Imperial College London
The factor N ! / (N − n )! is included in equation 8 to take
into account the number of ways that n particles can be
chosen from the total N in the system.

r
rn
r
r2 r4 Configuration r
of reduced set of n
particles.
r
r1

r
The other N − n
r3
r particle positions
… rn
are averaged out.

The reduced generic particle density is normalised such that

∫ρ
(n)
N (r )d r =
rn rn N!
(N − n )!
(9)
10 G. Jackson – Imperial College London
The single-particle density (n = 1) of a uniform fluid is simply
the number density
(1) r r N
ρ (r ) d r = = ρ (10)
V
(1) r r
and is normalised as ∫ ρ (r ) d r =N .
When the particles are uncorrelated (e.g., an ideal gas) the
n-body particle density is the product of the single-particle
densities, ρ (r ) = ∏ ρ (ri ) = ρ n , the last equality in the
(n) r n (1) r

case of a homogeneous fluid.


An n-body distribution function can then be defined in
terms of the reduced generic particle density to quantify the
extent by which the structure of the system deviates from
complete randomness (particle correlations):
ρ (r ) ρ N (r )
(n) r n (n) r n
g N( n ) (r n ) = n N
r
= (11)
(1) r ρ n

∏ i ρ ( r )
i =1
11 G. Jackson – Imperial College London
1.2 Pair Distribution Function
The pair distribution function (n = 2) of a uniform fluid can
be obtained at once from equations 8 and 11 as

ρ N (r ) N (N − 1) ∫ exp(− β U (r ))d r
( 2) r 2
rN r N −2
(2) r r
g N (r1 , r2 ) = = (12)
ρ 2
ρ 2
Z NVT

When the fluid is isotropic the pair distribution function


r r
depends only on the relative separation r12 = r2 − r1 . The
probability of finding any one of the N − 1 particles at a
relative distance from a given particle is proportional to
d r1 ∫ exp(− β U (r ))d r
r rN r N −2
(2) r (2) r r r
ρ N (r12 ) = ∫ ρ N (r1 , r12 ) d r1 = (N − 1)
1 ∫
N Z NVT
exp (− β U (r ))d r
rN r N −2 (13)
= (N − 1)V

12
Z NVT G. Jackson – Imperial College London
In the ideal (uncorrelated) limit
r N −1 N −1
ρ N( 2 ) (r12 ) = lim =ρ (14)
V N →∞ V

A radial pair distribution function can thus be defined as


(2) r
r ρ (r )
g (r12 ) = g N( 2 ) (r12 ) = N 12 = ideal 12
n( r )
(15)
ρ n ( r12 )
a form commonly employed in molecular simulation.
The forms of the function usually used in the theory of
radiation scattering are given in terms of appropriate δ
functions, which follow directly from equation (12):

g N (r1 , r2 ) = 2 ∑∑ δ (r1 − ri )δ (r2 − rj )


(2) r r 1 N N r r r r
ρ i =1 j =1
(16)

∑∑ δ (r12 − rj + ri )
1 1 N
r r r
N
g ( r12 ) =
ρ N i =1 j =1
13 G. Jackson – Imperial College London
The Fourier transform of the pair distribution function
( )
r r r r
S ( k ) = 1 + ρ ∫ g ( r12 ) exp − ik ⋅ r12 d r12 (17)
is the so-called static structure factor which can be
determined experimentally from neutron or x-ray scattering.
The inverse transform gives the distribution function:
r
g ( r12 ) =
1
[
ρ (2π ) 3 ∫
] (
r
)
r r r
S ( k ) − 1 exp ik ⋅ r12 d k (18)
1st shell

Structure of a fluid
Radial pair distribution function 2nd shell
of liquid argon from neutron 3rd shell

scattering. See Yarnell et al.,


PRA (1973); Hansen and
McDonald (2006).

14 G. Jackson – Imperial College London


Characterising the Structure of a Material

gas liquid solid

long-range order

short-range order

Radial pair distribution function of N = 500 Lennard-Jones


particles at different densities. See Glotzer et al. (2006).
15 G. Jackson – Imperial College London
2. Thermodynamic Properties
2.1 Using the language of the distribution functions
developed in Section 1.
Let us consider a uniform fluid for which the potential energy
can be expressed as a sum of pair-wise additive contributions
(e.g., hard-sphere, square-well, Lennard-Jones models etc.):

U (r ) = ∑∑ u2 ( rij )
rN N N
(19)
i =1 j >i

For simplicity we consider intermolecular potentials which


have radial symmetry.

When particles interact through pairwise-additive forces, the


thermodynamic properties can be expressed in terms of
integrals over the pair distribution function.
16 G. Jackson – Imperial College London
2.2 Internal Energy
The total energy of the system involves the kinetic (ideal) and
configurational contributions
E = U ideal + U (20)
3
U ideal = NkT kinetic energy
2
The configurational part which is due to the interparticle
interactions can be expressed as an ensemble average in
the usual way in terms of the N-body probability distribution
function:
U (r )exp(− β U (r ))d r
rN rN rN
U = U NVT =

∫ exp (− β U (rN
r )) d
rN
r
(21)
= ∫ U (r )P (r )d r
rN (N ) rN rN

17 G. Jackson – Imperial College London


In terms of the pair-wise intermolecular potentials
(cf. equation 19) the configurational energy (21) can be
written as ⎛ N N ⎞
u2 ( rij ) ⎟⎟ exp(− β U (r ))d r
rN rN
∫ ⎜⎜⎝ ∑∑
i =1 j >i ⎠
U= (22)
Z NVT
The double sum over i and j gives rise to N ( N − 1) / 2 pair
terms which yield the same results on integration. We
therefore focus on a given pair (in this case 12).
1 ∫ u (r
2 12 ) exp (− β U (rN
r ) ) d r
rN
U = N ( N − 1)
2 Z NVT

∫ exp (− β U (rN
r )) d
r N −2 ⎤
r r r
1
( )
= ∫∫ u2 r12 N ( N − 1)
⎢ ⎥ d r1 d r2 (23)
2 ⎢⎣ Z NVT ⎥⎦
2 (2) r r
ρ g N (r1 , r2 )
18 G. Jackson – Imperial College London
The integral in equation 23 over all N particles has been
split into integrations over 1 and 2 and over the other
N − 2 particles to introduce the previously defined pair
distribution function (cf. equation 12):

1 ∫ u (r
2 12 ) exp (− β U (rN
r )) d
rN
r
U = N ( N − 1) (24)
2 Z NVT
(2) r r r r
= ρ ∫∫ u2 (r12 )g N (r1 , r2 ) d r1 d r2
1 2
2
1 N2 (2) r r r r r r
= u (r )g N (r12 ) d r1 d r12
2 ∫∫ 2 12
where r12 = r2 − r1
2V
1 N2 (2) r r r
= ∫ u2 (r12 )g N (r12 ) d r12 as ∫ d r1 = V
2 V
r
= 2πNρ ∫ u2 (r12 )g (r12 ) r122 dr12 with d r12 = 4π r122 dr12

19 G. Jackson – Imperial College London


The final expression for the configurational energy per
particle in terms of the radial pair distribution function is
obtained from equation 24 as
u = = 2πρ ∫ u2 (r12 )g (r12 ) r122 dr12
U
(25)
N
This so-called energy equation can be interpreted easily
as follows: volume of shell

u=∫ 1
2 u2 (r12 ) ρ g (r12 ) 4π r122 dr12

mean number of
particles in shell
n( r12 )

mean energy of central particle


with particles in shell

sum over shells


20 G. Jackson – Imperial College London
2.3 Pressure
An expression for the pressure of the system in terms of
the radial pair distribution function can be obtained from
the thermodynamic definition of the pressure (volume
derivative of Helmholtz free energy):
⎛ ∂F ⎞
P = −⎜ ⎟ (26)
⎝ ∂V ⎠ NT
The free energy is related to the canonical partition
function through F = −kT ln QNVT (cf. equation 3) so
⎛ ∂ ln QNVT ⎞ ⎛ ∂ ln Z NVT ⎞
P = kT ⎜ ⎟ = kT ⎜ ⎟ (27)
⎝ ∂V ⎠ NT ⎝ ∂V ⎠ NT
The second equality of equation 27 stems from the fact
that the only the configurational contribution to the
partition function depends on the volume
Z NVT = ∫ exp(− β U (r N ))d r N
r r
21 G. Jackson – Imperial College London
We now employ the method of Green (1947) and re-
express the configurational integral1 in terms of coordinates
scaled by the box dimension L = V 3

Z NVT = ∫ exp(− β U (r ))d r


rN rN

= ∫ ...∫ exp(− β U (r ))d r1...d rN


rN r r

= ∫ ...∫ exp(− β U (r ))dx1dy1dz1...dx N dy N dz N


rN

= V N ∫ ...∫ exp(− β U (r N ))dx1*dy1*dz1*...dx*N dy *N dz *N


r

= V N ∫ exp(− β U (r N ))d r *N
r r
(28)
where
1 1 1
xi = xi / L = xi / V 3
*
y = yi / L = yi / V
*
i
3
z = zi / L = zi / V
*
i
3

dx = dxi / V
*
1
3
dy = dyi / V
*
1
3
dz = dzi / V
*
1
3 (29)
i i i
r* r
d ri = dxi dyi dzi = d ri / V
* * *

22 G. Jackson – Imperial College London


In the case of pairwise-additive potentials (cf. equation 19)
U (r ) = ∑∑ u2 ( rij )
rN N N

i =1 j >i

the configuration integral can be differentiated with respect


to volume as

⎟ = NV ∫ exp(− β U (r ))d r
1
⎛ NVT ⎞
Z N −1 rN r *N

⎝ ∂V ⎠ NT 0

⎛ ⎞ (30)
∂u2 ( rij ) ∂rij
exp(− β U (r )) ∑∑
N 1
V rN ⎜ N N
⎟ r *N
− ∫ ⎜ ⎟
dr
kT 0 ⎝ i =1 j >i ∂rij ∂V ⎠
where
rij = [
(xi − x j )2 + ( yi − y j )2 + (zi − z j )2 ]
=V
1
3
[(x − x ) + (y − y ) + (z − z ) ]
*
i
* 2
j
*
i
* 2
j
*
i
* 2
j (31)
1 ∂rij rij* rij
rij = V r
3 *
= =
∂V 3V
ij 2
3V 3

Note invariance of scaled distances to changes in volume.


23 G. Jackson – Imperial College London
The derivative of the configurational integral can then be
written as
⎛ ∂Z NVT ⎞
( ( ))
1
rN r *N
∫0 β
N −1
⎜ ⎟ = NV exp − U r d r
⎝ ∂V ⎠ NT
r N ⎛⎜ N N rij ∂u2 ( rij ) ⎞⎟ r *N
exp(− β U (r )) ∑∑
1
VN
− ∫ ⎜ ⎟
dr
kT 0 ⎝ i =1 j >i 3V ∂rij ⎠
= ∫ exp(− β U (r ))d r
N rN rN
V
r N ⎛⎜ N N ∂u2 ( rij ) ⎞⎟ r N
exp(− β U (r )) ∑∑ rij
1

3VkT ∫ ⎜ ∂rij ⎠ ⎟
dr
⎝ i =1 j >i
r N ⎛⎜ N N ∂u2 ( rij ) ⎞⎟ r N
exp(− β U (r )) ∑∑ rij
1
= ρ Z NVT −
3VkT ∫ ⎜ ∂rij ⎠ ⎟
dr
(32)
⎝ i =1 j >i
where we have transformed back to the original variables
(cf. equation 29).
24 G. Jackson – Imperial College London
We proceed to obtain an expression for the pressure from
equation 32:
⎛ ∂ ln Z NVT ⎞ 1 ⎛ ∂ ln Z NVT ⎞
⎜ ⎟ = ⎜ ⎟
⎝ ∂V ⎠ NT Z NVT ⎝ ∂V ⎠ NT
r N ⎛⎜ N N ∂u2 ( rij ) ⎞⎟ r N
exp(− β U (r )) ∑∑ rij
1 1
=ρ−
3VkT Z NVT ∫ ⎜ ∂rij ⎠ ⎟
dr
(33)
⎝ i =1 j >i
As for the internal energy the double sum over i and j gives
rise to N ( N − 1) / 2 pair terms which yield the same results on
integration so we can choose a given pair (say 12):
⎛ ∂ ln Z NVT ⎞
⎜ ⎟ =ρ
⎝ ∂V ⎠ NT
1 ∂u2 ( r12 ) ⎡
∫ exp (− β U (rN
r )) d
r N −2 ⎤
r r r (34)

6VkT ∫∫ r12
∂r12 ⎢
⎢ N ( N − 1)
Z NVT
⎥ d r1 d r2
⎥⎦

r r
ρ g (r1 , r2 )2 (2)
N
Note presence of pair distribution function (cf. equation 12)
25 G. Jackson – Imperial College London
In terms of the pair distribution function:
⎛ ∂ ln Z NVT ⎞ ∂u2 ( r12 ) ( 2 ) r r r r
g N (r1 , r2 ) d r1 d r2
1
⎜ ⎟ =ρ− ρ ∫∫ r12
2

⎝ ∂V ⎠ NT 6VkT ∂r12
∂u2 ( r12 ) ( 2 ) r r
g N (r12 ) d r12
1 2
=ρ− ρ ∫ r12
6kT ∂r12
2π 2 ∂u2 ( r12 )
=ρ− ρ ∫ r12 g (r12 ) r122 dr12 (35)
3kT ∂r12
The pressure is then given by
⎛ ∂ ln Z NVT ⎞ 2π 2 ∂u2 ( r12 )
P = kT ⎜ ⎟ = ρkT − ρ ∫ r12 g (r12 ) r122 dr12 (36)
⎝ ∂V ⎠ NT 3 ∂r12
This is usually referred to as the pressure equation or the
virial equation as it corresponds to an average of the
virial function r r ∂u ( r )
r1 ⋅ F12 = r12 2 12
∂r12
26 G. Jackson – Imperial College London
We have now demonstrated the usefulness of the pair
distribution function g (r12 ) in expressing the
thermodynamic properties such as the energy and
pressure of a many-particle system in terms of one-
dimensional integrals rather than the N-dimensional
integrals inherent in the full ensemble averages (cf.
equation 4). Though the expressions appear simpler in
form, the difficulty is now shifted to determining the
distribution function.

The other thermodynamic properties can be obtained from


the standard thermodynamic relations. For example, the
Helmholtz free energy can be obtained by integrating
equation 26.

27 G. Jackson – Imperial College London


3. Perturbation Theories
As we have seen in the previous sections the structural and
thermodynamic properties of fluids of particles interacting
through pair-wise forces can be determined from a knowledge
of two-body distribution functions. Unfortunately, an analytical
form of these functions is only available for a very limited
number of systems (e.g., particles interacting through a hard-
sphere pair potential). The problem is even more complicated
if there are many-body forces or if the intermolecular potential
is not spherically symmetrical.

Perturbation theories provide a means by which the


properties of a system (e.g., the distribution functions, free
energy or pressure) can be represented as a perturbation
from those of a reference system with known properties. Such
approaches therefore represent some of the most versatile,
accurate and powerful theories of the liquidG.state.
28 Jackson – Imperial College London
The intermolecular pair potential is often found to separate
naturally into a steep short-range repulsion and a smoothly
varying long-range attraction. The structure of most simple
liquids is largely determined by the molecular packing
which is dominated by the repulsive interactions.

Structure factor of a Lennard-Jones


fluid (curve) compared with that an
equivalent hard-sphere system
(points). See Verlet (1968); Hansen
and McDonald (2006).

The attractive interactions may thus be treated as a


uniform background potential that contributes to the
configurational energy of the fluid but does not affect its
structure.
29 G. Jackson – Imperial College London
3.1 van der Waals (1873)
Little was know of the form of intermolecular potentials at
the time that van der Waals developed his equation of
state for fluids. As Logan showed in his lectures the
assumptions made by van der Waals in developing his
theory of liquids involve the use of a mean-field
approximation (uniform background attractive potential).
Here, the total configurational energy is approximated as a
sum of one-body potentials which only depends on the
position of a given molecule:
N
r
U ≈ ∑ϕ i (ri ) (37)
i =1

The configurational integral can thus be expressed as


Z NVT = ∫ exp(− β U (r ))d r
rN rN

30
[ r r
≈ ∫ exp(− β ϕ (r )) d r ] N
≈ [(V − Vexc ) exp(− β u )]
N
(38)
G. Jackson – Imperial College London
In equation 38 Vexc is the average volume exclude to a
given particle by the other particles and u is the average
potential energy experienced by a particle which is
assumed to be constant (at a given density).

An expression for the pressure can be obtained directly


from the configurational integral (cf. equation 27)
⎛ ∂ ln Z NVT ⎞ NkT ∂u
P = kT ⎜ ⎟ = −N (39)
⎝ ∂V ⎠ NT V − Vexc ∂V

It is clear from the expression for the configurational


energy (cf. equation 25) that if one assumes g (r12 ) = 1, i.e.,
no correlations between particles,

u = −ρ a with the constant a = −2π ∫ u2 (r12 ) r122 dr12 (40)

31 G. Jackson – Imperial College London


The volume derivative of the one-particle potential is
thus given by
∂u N ρ
= 2a= a (41)
∂V V V
so that from equation 39 the pressure is simply
NkT
P= − ρ 2a (42)
V − Vexc
All that remains is to provide an expression for the
excluded volume. If one assumes that the particles are
hard-spheres with a diameter σ then the excluded volume
v
(shaded region below) for a pair of particles is exc 3 = 4
πσ 3
.
van der Waals approximated the total
excluded volume by
N
σ V exc ≈ vexc = Nb (43)
2
2
where the constant b = πσ 3
32 3 G. Jackson – Imperial College London
Inserting the approximate excluded volume (equation 43)
into equation 42 we obtain the familiar form of the
van der Waals equation of state in terms of his two
constants a and b:
NkT
P= − ρ 2a (44)
V − Nb
This can be considered as a perturbative approach in that
a uniform attractive background (perturbation) is added to
the hard-sphere (reference) term. We will come back to
this point in the following section.

The hard-sphere contribution to the pressure (or


compressibility factor Z) can be expressed in terms of the
packing fraction η (the fraction of the volume occupied by
the molecules) as
PV 1 1 πσ 3
Z= = = where η = ρ (45)
NkT 1 − ρ b 1 − 4η 6
33 G. Jackson – Imperial College London
3.2 Zwanzig (1954) – The High Temperature Expansion
The basis of modern perturbation theories of liquids can be
traced back to the work of Longuet-Higgins, Barker, Pople,
and Zwanzig in the early 1950s. Here we follow the
treatment of Zwanzig (1954).

Let us assume that we can separate the pair potential into


reference (0) and perturbative (1) contributions, i.e.,
u2 ( rij ) = u2( 0 ) ( rij ) + u2(1) ( rij ) (46)
The configurational energy can then can then also be
expressed in terms of the two contributions as

U (r ) = ∑∑ u2 ( rij ) + ∑∑ u2 ( rij ) = U 0 (r ) + U1 (r ) (47)


rN N N
(0)
N N
(1) rN rN
i =1 j >i i =1 j >i

34 G. Jackson – Imperial College London


The Helmholtz free energy of the system of particles
interacting via the full potential and the reference potential
can be expressed in terms of corresponding partition
functions (cf. equations 2 and 3):
F = − kT ln Q F0 = −kT ln Q0 (48)
The difference in free energy between the full and
reference system (i.e., the perturbation to the free energy)
is
Q Z
ΔF = F − F0 = − kT ln = − kT ln (49)
Q0 Z0
The free energy difference can be expressed in terms of
the ratios of the configurational integrals
Z = ∫ exp(− β U (r N ))d r N and Z 0 = ∫ exp(− β U 0 (r N ))d r N (50)
r r r r

as the kinetic (ideal) contributions cancel.

35 G. Jackson – Imperial College London


The ratio of the configurational integrals is

exp(− β U (r ))d r
rN rN
Z
=

Z 0 ∫ exp(− β U 0 (r ))d r
rN rN

exp(− β U 0 (r ))exp(− β U1 (r ))d r


rN rN rN
=

∫ exp (− β U 0 (rN
r )) d
rN
r

= ∫ P0 (U 0 (r ))exp(− β U1 (r ))d r
(N ) rN rN rN

= exp(− β U1 (r N ))
r
0
(51)
This corresponds to the average of the Boltzmann factor
of the perturbation contribution to the energy with respect
to configurations of the reference system; the probability
distribution of the reference P0( N ) appears in the integral
(cf. equation 6).
36 G. Jackson – Imperial College London
The perturbation to the free energy is obtained from
equation 49 as
ΔF = − kT ln exp(− β U1 (r ))
rN
0
(52)
This important result due to Zwanzig is the starting point for
the popular perturbation theories of Barker and Henderson
(1967), and of Weeks, Chandler and Andersen (1971).

To make the expression more tractable both the exponential


and the logarithm can be expanded in powers of the ∞
xi
perturbative energy. Starting with the exponential e = ∑
x

i =0 i!

⎛ ⎞
ΔF = − ln ⎜1 − β U1 + [β U1 ] − [β U1 ] + ... ⎟
1 1 2 1 3

β ⎝ 2! 3! ⎠ 0

1 ⎛ 1 2 2 1 3 3 ⎞
= − ln⎜1 − β U1 0 + β U1 − β U1 + ... ⎟ (53)
β ⎝ 2 0 6 0

37 G. Jackson – Imperial College London
We now expand the logarithm ln (1 + x ) = x − x + x − ...
1 2 1 3
2 3
1⎛ 1 3 3 ⎞
ΔF = − ⎜ − β U1 0 + β U1 − β U1 + ... ⎟
2 2

β⎝ 0 6 0

2
1 ⎛ 1 3 3 ⎞
+ ⎜ − β U1 0 + β U1 − β U1 + ... ⎟
2 2
(54)
2β ⎝ 0 6 0

3
1 ⎛ 1 ⎞
− ⎜ − β U1 0 + β U1 − β 3 U13 + ... ⎟ + ...
2 2

3β ⎝ 0 6 0

Collecting the terms in powers of the inverse temperature we
obtain the so-called high-temperature expansion:
1
2
(
ΔF = U 1 0 − β U 1 − U 1 0
2
0
2
)
1 2 3
6
(
+ β U1 − 3 U12 U1 0 + 2 U1
0 0
3
0
)+ ... (55)

38 G. Jackson – Imperial College London


When one chooses a hard-core reference system (0) and
an attractive perturbation (1), as is common practice, the
first-order term is the so-called mean-attractive energy.

An explicit expression for the mean-attractive energy can


be obtained in terms of the pair distribution function of the
repulsive reference system following a similar procedure to
that employed for the configurational energy in Section 2.2

U1 (r )exp(− β U 0 (r ))d r
rN rN rN
=

( ( ))
U1 rN rN
∫ − β
0
exp U 0 r d r
⎛ N N (1) ⎞
∫ ⎜⎝ ∑∑
⎜ u2 ( r )
ij ⎟
⎟ exp (− β U 0 (rN
r )) d
rN
r
=
i =1 j >i ⎠ (56)
Z0

39 G. Jackson – Imperial College London


The double sum over i and j again gives rise to N ( N − 1) / 2
pair terms which yield the same results on integration:

∫ u (r12 )exp(− β U 0 (r ))d r


(1) rN rN
1 2
U1 0
= N ( N − 1)
2 Z0
exp(− β U 0 (r ))d r
⎡ rN r N −2 ⎤
= ∫∫ u2 (r12 )⎢ N ( N − 1)
1 (1) ∫ r r
⎥ d r1 d r2 (57)
2 ⎢⎣ Z0 ⎥⎦

r r
ρ g (r1 , r2 )
2 (2)
0

where
N ( N − 1) ∫ exp(− β U 0 (r ))d r
rN r N −2
( 2) r r
g 0 (r1 , r2 ) = (58)
ρ2 Z0

is the pair distribution function of the repulsive reference.


40 G. Jackson – Imperial College London
Following a procedure equivalent to that of Section 2.2:
(2) r r r r
= ρ ∫∫ u2 (r12 )g0 (r1 , r2 ) d r1 d r2
1 2 (1)
U1 0
2
= 2πNρ ∫ u2(1) (r12 )g 0 (r12 ) r122 dr12 (59)
from which it is clear that the first-order term of the
perturbation expansion is the average of the attractive
part of the potential over configurations of the repulsive
reference system. The van der Waals attractive
contribution (equation 40) is recovered when g 0 (r12 ) = 1
confirming the perturbative roots of his approach.
The second-order term
1
2
2
(
− β U1 − U1 0
0
2
) (60)
represents a fluctuation of the attractive energy and is
more difficult to evaluate; it depends on high-order
distribution functions.
41 G. Jackson – Imperial College London
It follows from the ongoing discussion that a knowledge of
the thermodynamic properties and structure of the
reference system allows one to obtain the properties of the
full system (to first-order at the very least).

In this regard, the hard-sphere fluid plays a pivotal role as


a tractable reference system for which accurate analytical
description of the properties is available.

42 G. Jackson – Imperial College London


3.3 Carnahan and Starling (1969) – Hard Sphere Fluid
As is apparent from the previous section the properties of
the hard-sphere fluid are central to the development of
accurate perturbation approaches for fluids of particles
interacting via more realistic intermolecular interactions.

The equation of state proposed by Carnahan and Starling


(1969) is a key development. The (second-virial)
approximation to the hard-sphere repulsive contribution
employed by van der Waals (cf. equation 43) fails rapidly
at moderate densities because the pair excluded volumes
are treated as additive which is clearly only true at very low
density. This means that the van der Waals equation of
state can not be used to represent the pressure (and other
thermodynamic properties) of the hard-sphere fluid at
moderate to high densities.
43 G. Jackson – Imperial College London
The coefficients of the virial expansion (as opposed to the
virial or pressure equation) of a hard-sphere system can
be obtained analytically and numerically to reasonably
high order.

The compressibility factor of the hard-sphere fluid up to


fifth order in density (packing fraction) is accurately
represented by
PV
Z= = 1 + 4η + 10η 2 + 18.36η 3 + 28.2η 4 + 39.5η 5 + ... (61)
NkT

When the van der Waals expression (equation 45) is


expanded as a binomial series only the leading-order term
(second virial coefficient) is reproduced exactly.

44 G. Jackson – Imperial College London


If the virial coefficients are approximated by their nearest
integer values
Z ≈ 1 + 4η + 10η 2 + 18η 3 + 28η 4 + 40η 5 + ... (62)
The series can be reproduced as a Padé approximant of
the following form:
1 +η +η2 −η3
Z≈ (63)
(1 − η )3

This relation (originally proposed by Carnahan and


Starling) together with an accurate integral equation
solution of the radial pair distribution function (see
Appendix D, Hansen and McDonald, 2006) provide a very
accurate representation of the properties of the hard-
sphere fluid for use as a reference in perturbation theories
of molecules with more realistic interactions.
45 G. Jackson – Imperial College London
3.4 Wertheim/SAFT – Associating and Non-spherical
Molecules

Most molecules of industrial relevance are non-spherical


and interact through complex interactions (polar, hydrogen
bonding, association etc.). These types of interactions can
not be captured adequately by simple intermolecular
potentials such as the square-well or Lennard-Jones.
Modern theories of the fluid state have been developed
within the perturbative formalism of the previous sections
to represent the thermodynamic properties and fluid phase
equilibria of more complex molecules.

The focus in this section is on the theory of associating


molecules of Wertheim (1984-86), as implemented in the
SAFT equation of state (Jackson et al. 1988, Chapman et
al. 1988-1990).
46 G. Jackson – Imperial College London
Model of molecules
• Monomeric segments interacting through simple pair-
wise interactions, e.g., square-well, Lennard-Jones
(short-range repulsion and long-range attraction)
• Nonspherical molecules made up as chains of the fused
segments
• Directional interactions and association (hydrogen
bonding, chem. equil.) mediated through additional
bonding sites

e.g., butan-1-ol

47 G. Jackson – Imperial College London


Wertheim’s theory of association

Wertheim, J. Stat. Phys. 35, 19 (1984); 35, 35 (1984)


Wertheim, J. Stat. Phys. 42, 459 (1986); 42, 477 (1986)

Jackson, Chapman, Gubbins, Mol. Phys. 65, 1 (1988)


Chapman, Jackson, Gubbins, Mol. Phys. 65, 1057 (1988)

+
48 G. Jackson – Imperial College London
Wertheim’s theory of association – Alternative view
Galindo, Burton, Jackson, Visco, Kofke, Mol. Phys. 100, 2241 (2002)

Hansen, McDonald Theory of Simple Liquids, 3rd Edition (2006)


11.10 Associating Liquids

+
ρ - total density of segments
ρ0 - density of segments as monomers
ρD - density of segments in dimers

The corresponding mass action equation (cf. chemical


equilibria) is
ρ = ρ0 + 2ρ D (64)
49 G. Jackson – Imperial College London
In the case of an ideal dimerising gas (Hill 1956) we can write

ρ D (1) = ρ 0 (1) ∫ ρ 0 ( 2 ) [exp (− u (12 ) / kT ) − 1]d ( 2 ) (65)


1
2

For an interacting system Wertheim (1984) showed that an the


expression becomes

ρ D (1) = ρ 0 (1) ∫ ρ 0 ( 2 ) g 00 (12 ) [exp (− u (12 ) / kT ) − 1]d ( 2 )


1
2
1
= ρ 0 (1) ∫ ρ 0 ( 2 ) g 00 (12 ) f (12 ) d ( 2 ) (66)
2
using an exact cluster expansion.

A thermodynamic perturbation theory (TPT1) can be developed


with the approximation
g 00 (12 ) ≈ g ref (12 ) (67)
ρ D (1) ≈ ρ 0 (1) ∫ ρ 0 ( 2 ) g ref (12 ) f (12 ) d ( 2 ) (68)
50 G. Jackson – Imperial College London
The mass action equation can then be approximated as
ρ (1) = ρ 0 (1) + 2 ρ D (1)
≈ ρ 0 (1) + ρ 0 (1) ∫ ρ 0 ( 2 ) g ref (12 ) f (12 ) d ( 2 ) (69)

A form of the Helmholtz free energy at the TPT1 level that is


consistent with this mass action equation is
⎛ δ F [ρ 0 (1) ] ⎞
⎜⎜ ⎟⎟ = 0 ⎯
⎯→ mass action equation
⎝ δρ 0 (1) ⎠
F assoc . ⎛ ρ 0 (1) ⎞
kT
=∫ ⎜⎜⎝ ρ (1) ln ρ (1) − ρ 0 (1) + ρ (1) ⎟⎟⎠ d (1) (70)
1
− ∫ ρ 0 (1) ∫ ρ 0 ( 2 ) g ref (12 ) f (12 ) d ( 2 ) d (1)
2
which follows from the variation of the free energy with the
density of monomers.
51 G. Jackson – Imperial College London
In the case of a homogeneous system
ρ (1) = ρ ρ 0 (1) = ρ 0 ( 2 ) = ρ 0 X = ρ0 / ρ
and the mass action equation (Jackson et al. 1988) can be
expressed as a simple quadratic equation:
ρ = ρ 0 + ρ 02 ∫ g ref (12 ) f (12 ) d ( 2 ) (71)

≈ ρ 0 + ρ 02 Δ 12 where Δ 12 = g ref (σ 12 ) K 12 F12


1
1 = X + X ρ Δ 12 2
or X = (72)
1 + X ρ Δ 12
The perturbation to the free energy due to association is then
F assoc . ⎛ ρ 0 (1) ⎞
= ∫ ⎜⎜ ρ (1) ln − ρ 0 (1) + ρ (1) ⎟⎟ d (1)
kT ⎝ ρ (1) ⎠
+ ∫ρ 0 (1) ∫ ρ 0 ( 2 ) g ref (12 ) f (12 ) d ( 2 ) d (1)
ρ0 ρ0 1 ρ 02 X 1
= ln − +1− ρ Δ = ln X − + (73)
ρ ρ 2 ρ 2 12
2 2
52 G. Jackson – Imperial College London
For systems with multiple bonding sites one can assume that
bonding at a given site is independent of that at another and
approximate the free energy with independent terms as
F assoc . ⎛ Xa 1⎞
NkT
=∑A ⎜⎝ ln X a − 2 + 2 ⎟⎠ (74)

1
XA =
1 + ∑ ρ X b Δ ab
(75)
a

One can obtain an expression for the bonding together of


monomeric segments into chains by considering the limit of
complete association (Chapman et al. 1988). In the following,
however, we develop the chain contribution using an alternative
approach to provide further insights into the physical nature of
the approximations that are employed.

53 G. Jackson – Imperial College London


The contribution to the free energy due to chain formation
can be obtained as outlined by Paricaud et al. (2003).

We start by considering the formation of a single chain from a


monomer fluid
N segments N-m segments 1 chain

54 G. Jackson – Imperial College London


The residual (in excess of the ideal contribution) free energy
and chemical potential for these two systems are

Z Z N − m ,1
F N
res
= − kT ln NN F = − kT ln N − m + 1
res
N − m ,1 (76)
V V
⎛ m − 1 Z N − m ,1 ⎞
μ res = F Nres− m ,1 − F Nres = − kT ln ⎜⎜ V ⎟⎟ (77)
⎝ ZN ⎠

where the configuration integrals are defined in the usual way


as

= ∫ exp (− U ( r ) / kT )d r1 ... d rN
rN r r
ZN (78)

Z N − m ,1 = ∫ exp (− U ( r ) / kT )d r1 d rm + 1 ... d rN
r N − m ,1 r r r

= V ∫ exp (− U ( r N − m ) / kT )d rm +1 ... d rN
r r r
(79)

55 G. Jackson – Imperial College London


The monomer m-body distribution function is of the form (cf.
equation 11)

∫ exp (− U ( r ) / kT )d rm +1 ... d rN
Vm rN r r
g (1, 2 ... m ) = (80)
ZN
An appropriate m-body cavity function can be defined as
y (1, 2 ... m ) = g (1, 2 ... m ) exp (U ( r m ) / kT )
r

∫ exp (− U ( r ) / kT )d r
Vm r N −m r r
= m + 1 ... d rN
ZN
V m Z N − m ,1 m −1 Z N − m ,1
= =V (81)
ZN V ZN
The chemical potential of forming a single chain can
therefore be expressed in terms of the cavity function as
⎛ Z N − m ,1 ⎞
μ res
= − kT ln ⎜⎜ V m −1
⎟⎟ = − kT ln y (1, 2 ... m ) (82)
⎝ ZN ⎠
56 G. Jackson – Imperial College London
The residual chemical potential of a hard-sphere chain at
rolling contact is
μ res = − kT ln y (1, 2 ... m ) = − kT ln g (1, 2 ... m ) (83)

The free energy of forming Nm such chains can


therefore approximated as
F chain ≈ N m μ res = − N m kT ln g (1, 2 ... m ) (84)
By employing the linear approximation for m-body function
g (1, 2 ... m ) ≈ g (σ 12 ) g (σ 23 )... g (σ m −1, m ) (85)
g (σ 12 ) ≈ g (σ 23 )... ≈ g (σ m −1, m ) ≈ g ref (σ ) (86)

an expression for chain free energy (TPT1) can be obtained:


F chain
= − ln g (1, 2 ... m ) ≈ − ln g ref (σ ) m −1 = − ( m − 1) ln g ref (σ )
N m kT
(87)
57 G. Jackson – Imperial College London
Statistical Associating Fluid Theory (SAFT)

The SAFT equation of state can be developed within a


standard perturbation approach with the association and
chain contributions that have just been developed.

Original Theory
Chapman, Gubbins, Jackson, Radosz,
Ind. Eng. Chem. Res., 29, 1709 (1990)

SAFT-VR (potentials with variable range)


Gil-Villegas, Galindo, Whitehead, Mills, Jackson, Burgess,
J. Chem. Phys., 106, 4168 (1997)

Numerous other versions (soft-SAFT, PC-SAFT)


Blas and Vega (1997), Gross and Sadowski (2001)
58 G. Jackson – Imperial College London
The SAFT free energy is generally expressed as

F F ideal F mono . F chain F assoc .


= + + + (88)
NkT NkT NkT NkT NkT
which incorporates the ideal, monomer segment, chain and
association contributions. The expressions for the
SAFT-VR version of the theory are summarised below.

The ideal (kinetic) contribution to the free energy is a linear


function of the logarithm of the number density

F ideal
= ln( ρ v / Ω ) − 1 (89)
NkT
where v / Ω is the corresponding de Broglie volume
(constant).
59 G. Jackson – Imperial College London
The monomer segment term is obtained from a standard
second-order perturbation theory (cf. Section 3.2):

Perturbation theory
Barker and Henderson (1975)
2
⎛ 1 ⎞ ⎛ 1 ⎞
= F +⎜ ⎟ F1 + ⎜ ⎟ F2
mono . hs
F (90)
⎝ kT ⎠ ⎝ kT ⎠
Hard sphere reference F hs 4η − 3η 2
= (91)
Carnahan and Starling (1969) N s kT (1 − η )2
Mean value theorem a = F1 ∞
1
N s kT
= − 2π ∫σ u ( r ) g hs ( r ) r 2 dr

≈ − ρ s a vdw g hs (σ ;η eff . ) (92)


Local compressibility F2 1 hs ∂ a 1*
approximation (LCA) a2 = ≈ K ρs (93)
N s kT 2 ∂ρ s
60 G. Jackson – Imperial College London
The contribution to the free energy due to association is
expressed as described earlier:

F chain
= − ( m − 1) ln g mono . (σ ) (94)
NkT

Finally, the chain contribution is given by

F assoc . s
⎛ Xa 1⎞
NkT
= ∑ ⎜ ln X a −
a =1 ⎝ 2
− ⎟
2⎠
(95)
1
Xa =
∑ ρ X b Δ ab (96)
s
1+ b =1

Δ ab ≈ K ab Fab g mono . (σ ) (97)

61 G. Jackson – Imperial College London


The other thermodynamic properties are obtained from
standard thermodynamic relations. For example, the pressure
and chemical potential can be expressed as the following
partial derivatives of the Helmholtz free energy:
⎛ ∂F ⎞ ⎛ ∂F ⎞
P = −⎜ ⎟ μ =⎜ ⎟ (98)
⎝ ∂ V ⎠ NT ⎝ ∂ N ⎠ VT
At equilibria the temperature, pressure and chemical potential
of the two phases must be the same (solved numerically).

Though we have presented the expressions for pure


components these can be generalised to mixtures in a
straightforward manner (e.g., see Gil-Villegas et al. 1997).

The SAFT equation of state provides an accurate description


of the thermodynamic properties and fluid phase equilibria of
complex fluid mixtures. Some examples of the versatility of
the approach are provided in the final section.
62 G. Jackson – Imperial College London
3.5 Examples of the Description of Real Complex
Fluids and Mixtures

63 G. Jackson – Imperial College London


Alkanes
Vapour-liquid equilibria

Gil-Villegas et al.
J. Chem. Phys. (1997)

64 G. Jackson – Imperial College London


Alkanes
Vapour-liquid equilibria

McCabe and Jackson


PCCP (1999)

65 G. Jackson – Imperial College London


C5H12+PE
vapour-liquid equilibria
liquid-liquid equilibria

Paricaud et al.
PE 100 kg/mol Ind. Eng. Chem. Res. (2004)
66 G. Jackson – Imperial College London
Water H2O
Vapour-liquid equilibria

Clark et al.
Mol. Phys. (2006)

67 G. Jackson – Imperial College London


Water H2O
Degree of association

Clark et al.
Mol. Phys. (2006)

68 G. Jackson – Imperial College London


Water H2O
Vapour-liquid tension

Blas et al. Mol. Phys. (2001);


Gloor et al. J. Chem. Phys. (2004)
69 G. Jackson – Imperial College London
HFC-32
Difluoromethane
Vapour-liquid equilibria

Galindo et al.
J. Phys. Chem. B (1998)

70 G. Jackson – Imperial College London


H2O+HF
Vapour-liquid equilibria

Galindo et al.
J. Phys. Chem. B (1997)

71 G. Jackson – Imperial College London


H2O+HF
Vapour-liquid equilibria

Galindo et al.
J. Phys. Chem. B (1997)

72 G. Jackson – Imperial College London


H2O+HF
Vapour-liquid equilibria

Galindo et al.
J. Phys. Chem. B (1997)

73 G. Jackson – Imperial College London


H2O+HF
Vapour-liquid equilibria

Galindo et al.
J. Phys. Chem. B (1997)

74 G. Jackson – Imperial College London


H2O+C4E1
Vapour-liquid equilibria
Liquid-liquid equilibria

Garcia Lisbona et al.


Molec. Phys. (1997)

75 G. Jackson – Imperial College London


H2O+C4E1
Liquid-liquid equilibria

Garcia Lisbona et al.


Molec. Phys. (1997)

76 G. Jackson – Imperial College London


H2O+C10E5
Vapour-liquid equilibria
Liquid-liquid equilibria

Garcia Lisbona et al.


Molec. Phys. (1997)

77 G. Jackson – Imperial College London


H2O+CiEj
Vapour-liquid equilibria
Liquid-liquid equilibria

Garcia Lisbona et al.


JACS (1998)

78 G. Jackson – Imperial College London


H2O+PEG
Liquid-liquid equilibria

Clark et al.
Macromolecules (2008)

PEG 2.18 to 1,020 kg/mol


79 G. Jackson – Imperial College London

You might also like