You are on page 1of 88

Elective Module:

CE4-33 Molecular Modelling of Fluids

Session 2: Classical Statistical Mechanics:


Theory of the Liquid State

Molecular Modelling of Fluids Session 2 – Slide 1 © Imperial College London


Key messages from last time
•  The system sizes we are dealing with are
overwhelmingly large
–  in terms of numbers of particles / molecules
•  This has important consequences in terms of
other key results
–  direction of spontaneous change: the probability of
change in the opposite direction is vanishingly small
–  allows us to simplify maths
•  use crude version of Stirling’s approximation
•  work with Taylor expansions including only a few terms
–  The extremely peaked nature of the Boltzmann
distribution
•  most important configuration is overwhelmingly so
Molecular Modelling of Fluids Session 2 – Slide 2 © Imperial College London
Key messages from last time

•  Boltzmann’s equation: S = kB ln ⌦
–  where S is the entropy, kB is Boltzmann’s constant and
Ω represents “the number of ways”
•  in the molecular context, the number of microstates
•  in the ensemble context, the number of system states
•  Microstate:
–  the set of all the quantum states of all the molecules
•  Macrostate:
–  what we generally just call the state
–  many microstates correspond to the same macrostate

Molecular Modelling of Fluids Session 2 – Slide 3 © Imperial College London


Key messages from last time

•  Principle of equal a priori probabilities


–  each microstate exists with equal probability
•  Stirling’s approximation: ln N ! ⇡ N ln N N
•  Boltzmann distribution law:
exp( "i ) exp( "i )
P ("i ) = P ⌘
i exp( "i ) q
–  where β = 1 / kBT, εi is the energy (level) associated with
molecule i, and
–  q is the molecular partition function

Molecular Modelling of Fluids Session 2 – Slide 4 © Imperial College London


Key messages from last time

•  The concept of ensembles:


–  assembly of replicas of the system
–  each replica shares three macroscopic variables in
common
•  The canonical (NVT) ensemble
–  constant number of particles, volume and temperature
•  Measured properties are time averages
•  The ergodic hypothesis:
–  the ensemble average of a quantity is equal to the time
average

Molecular Modelling of Fluids Session 2 – Slide 5 © Imperial College London


Key messages from last time

•  State of subsystem / member of ensemble


–  set of position and momentum coordinates of all the
molecules
•  more to follow in today’s lecture
•  Principle of equal a priori probabilities
–  each state of subsystem exists with equal probability
•  Canonical / Boltzmann distribution
exp( Ei ) exp( Ei )
P (Ei ) = P ⌘
i exp( Ei ) Q
–  where β = 1 / kBT, Ei is the energy associated with
subsystem i, and
–  Q is the canonical partition function
Molecular Modelling of Fluids Session 2 – Slide 6 © Imperial College London
Key messages from last time

•  Link to macroscopic thermodynamics:


–  Partition function leads to macroscopic thermodynamic
properties
✓ ◆
@ ln Q
U= A = kB T ln Q
@ N,V
✓ ◆ ✓ ◆
@A @A
p= ; µ=
@V N,T @N V,T

Molecular Modelling of Fluids Session 2 – Slide 7 © Imperial College London


Key messages from last time

–  Other thermodynamic quantities then obtained from


standard thermodynamic relationships:

•  enthalpy: H = U + pV
✓ ◆
@U
•  isochoric heat capacity: CV =
@T V
✓ ◆
@H
•  isobaric heat capacity: Cp =
@T p
•  Gibbs free energy: G = H T S = U + pV T S
@A
= A + pV = A V @V N,T

Molecular Modelling of Fluids Session 2 – Slide 8 © Imperial College London


Key messages from last time

–  Other thermodynamic quantities then obtained from


standard thermodynamic relationships:

•  entropy:
✓ ◆ 2
✓ ◆
@Q 1 kB T @Q
A = U TS = TS = TS
Q
@ N,V Q @T N,V
2
✓ ◆
kB T @Q
i.e., kB T ln Q = TS
Q @T N,V
✓ ◆ ✓ ◆
kB T @Q @ ln Q
so S = kB ln Q + = kB ln Q + kB T
Q @T N,V @T N,V
Molecular Modelling of Fluids Session 2 – Slide 9 © Imperial College London
Today’s Lecture

Session 2: Classical Statistical Mechanics

Molecular Modelling of Fluids Session 2 – Slide 10 © Imperial College London


Thermodynamics from Q

•  Recall: Canonical distribution


exp( Ei ) X
P (Ei ) = ; Q= exp( Ei )
Q i
•  We can use this to evaluate thermodynamic
quantities, if we have expressions for quantised
energies, Ei
–  actually very rare
or, for systems of independent particles, if we
have expressions for quantised energy levels, εi
–  via the relationship between q and Q
Molecular Modelling of Fluids Session 2 – Slide 11 © Imperial College London
Example: Ideal Gas Equation

•  Ideal-gas particles have no


interaction with each other
–  so are independent of each other

•  Each ideal-gas molecule


behaves like a “particle in a box”
–  only motion is translation
•  assuming ideal gas is monatomic
–  which we can decompose into components in each of
the Cartesian (x, y, z) directions

Molecular Modelling of Fluids Session 2 – Slide 12 © Imperial College London


Ideal Gas Equation
•  From quantum theory, energy levels
corresponding to translation of a particle in a
box are given by:
n2x h2
"x = 2
; nx = 1, 2, . . .
8mlx
nx, ny and nz are quantum ♯s;
n2y h2 h is Planck’s constant;
"y = 2
; ny = 1, 2, . . . m is the particle mass;
8mly lx, ly and lz are box dimensions
n2z h2
"z = 2
; nz = 1, 2, . . .
8mlz
–  see e.g., Atkins Physical Chemistry
Molecular Modelling of Fluids Session 2 – Slide 13 © Imperial College London
Ideal Gas Equation

•  These expressions allow us to evaluate the


molecular partition function, q
•  We then use the relationship between q and Q
to give us the canonical partition function.
•  However, there is a caveat here:
–  Equation 1.39 (Q = qN ) applies to distinguishable
particles
–  Ideal gas particles are indistinguishable
–  For indistinguishable particles, we over-counted right at
the start:

Molecular Modelling of Fluids Session 2 – Slide 14 © Imperial College London


Indistinguishable Particles

•  Recall our example: how many ways can we


distribute three distinguishable particles among
three energy levels?
2! 3! 2! 3! 2! 1! 1!
1! 2! 3! 1! 1! 3! 2! 6 ways!
0! 1! 1! 2! 3! 2! 3!
Indistinguishable Particles?
2! x!
1! x! 1 way!
0! x!
•  Over-counting is by a factor of 6 = 3 × 2 = N!
Molecular Modelling of Fluids Session 2 – Slide 15 © Imperial College London
Indistinguishable Particles

•  Analogously, Equation 1.39 must be modified


to correct for this over-counting:

N
q
Q= (Equation 1.52)
N!
Indistinguishable Particles

Molecular Modelling of Fluids Session 2 – Slide 16 © Imperial College London


Ideal Gas Equation

•  Recall the expressions for the εi and definition


✓ of◆q :
2 2
nx h X "i
"x = 2
; nx = 1, 2, . . . q = exp
8mlx kB T
i
•  whence ( !)
XXX h2 n2x n2y n2z
qideal = exp 2
+ 2 + 2
n n n
8m l x l y l z
x y z

(Equation 2.3)

•  from which Q can be obtained, and then A, and


finally the ideal gas equation via p = (@A/@V )N,T
–  Left as an Exercise; see Chapter 2 for the solution.
Molecular Modelling of Fluids Session 2 – Slide 17 © Imperial College London
Statistical Mechanics in Practice

•  Recall again: Canonical distribution


exp( Ei ) X
P (Ei ) = ; Q= exp( Ei )
Q i
•  Unfortunately, except in special cases
–  such as the “particle in a box” that we just looked at
•  which leads to ideal gas EoS
we do not have expressions for quantised
energy levels, εi , or for the energies Ei
•  Hence, we cannot write down Q or P(Ei)

Molecular Modelling of Fluids Session 2 – Slide 18 © Imperial College London


Statistical Mechanics in Practice

•  However, we do often have an expression for


the Hamiltonian, H
–  sum of potential and kinetic energies of particles
•  H is a function of positions and momenta of
the particles / molecules
•  Leads to a problem …

Molecular Modelling of Fluids Session 2 – Slide 19 © Imperial College London


Statistical Mechanics in Practice

•  Problem: we cannot know both position and


momentum exactly
–  Heisenberg’s Uncertainty Principle

•  Instead, we know r to within dr and p to within


dp (say)
•  Leads to identification of system state:

Molecular Modelling of Fluids Session 2 – Slide 20 © Imperial College London


System state
Particle i

Position ri

Volume element,
(exaggerated!) size
dr

Molecular Modelling of Fluids Session 2 – Slide 21 © Imperial College London


System state
Particle i

Position ri

Volume element,
(exaggerated!) size
dr

Molecular Modelling of Fluids Session 2 – Slide 22 © Imperial College London


System state
Particle i

Position ri

Volume element,
(exaggerated!) size
dr

Molecular Modelling of Fluids Session 2 – Slide 23 © Imperial College London


System state
Particle i

Position ri

Volume element,
(exaggerated!) size
dr

Molecular Modelling of Fluids Session 2 – Slide 24 © Imperial College London


System state
Particle i

Position ri

Volume element,
(exaggerated!) size
dr

Molecular Modelling of Fluids Session 2 – Slide 25 © Imperial College London


System state

Momentum pi

Momentum element,
Position ri
size dp

Volume element,
(exaggerated!) size
dr
Momentum of
Particle i

Molecular Modelling of Fluids Session 2 – Slide 26 © Imperial College London


System state

Momentum pi

Momentum element,
Position ri
size dp

Volume element,
(exaggerated!) size
dr
Momentum of
Particle i

Molecular Modelling of Fluids Session 2 – Slide 27 © Imperial College London


System state

Momentum pi

Momentum element,
Position ri
size dp

Volume element,
(exaggerated!) size
dr
Momentum of
Particle i

Molecular Modelling of Fluids Session 2 – Slide 28 © Imperial College London


System state

Momentum pi

Momentum element,
Position ri
size dp

Volume element,
(exaggerated!) size
dr
Momentum of
Particle i

Molecular Modelling of Fluids Session 2 – Slide 29 © Imperial College London


System state

Momentum pi

Momentum element,
Position ri
size dp

Volume element,
(exaggerated!) size
dr
Momentum of
Particle i

Molecular Modelling of Fluids Session 2 – Slide 30 © Imperial College London


System state

Positions
r1, … , rN

Molecular Modelling of Fluids Session 2 – Slide 31 © Imperial College London


System state

Positions
r1, … , rN
and momenta
p1, … , pN

Molecular Modelling of Fluids Session 2 – Slide 32 © Imperial College London


System state

Can loosely identify system state j

(
(j)
In state j, lies in volume element dri centred on and ri
•  particle i (j)
has momentum in element dpi centred on pi
–  where i = 1,…, N includes all N particles / molecules in system,
and dri , dpi are infinitesimals
(j)
•  energy of particle i in element dHi centred on Hi
–  dH is infinitesimal
Molecular Modelling of Fluids Session 2 – Slide 33 © Imperial College London
System state j

Can loosely identify system state j

8 ⇣ ⌘
< “position in the interval r(j) ± dr/2 ”
i
•  particle i ⇣ ⌘
: “momentum in the interval p(j) ± dp/2 ”
i

–  where i = 1,…, N includes all N particles / molecules in system,


and dri , dpi are infinitesimals
⇣ ⌘
(j)
•  “energy of particle i in the interval Hi ± dH/2 ”
–  dH is infinitesimal
Molecular Modelling of Fluids Session 2 – Slide 34 © Imperial College London
System state j

•  In a sense: system has an energy (Hamiltonian)


corresponding to positions and momenta
“inside” some region of phase space
–  phase space: 6N-dimensional space defined by
positions and momenta
•  3N for positions (x, y, z); 3N for momenta (px, py, pz)

•  P (system has this energy) must be


proportional to the size of the region
–  the bigger the region, the more likely the system has
energy corresponding to that region

Molecular Modelling of Fluids Session 2 – Slide 35 © Imperial College London


Probability densities

•  This in turn leads us to think about probability


densities
–  rather than direct probabilities, i.e., :
N N (N ) N N N N
P (Ei ) “P H r , p ” f r ,p dr dp
rN ⌘ r1 , r2 , . . . , rN ; drN ⌘ dr1 dr2 . . . drN
pN ⌘ p1 , p2 , . . . , pN ; dpN ⌘ dp1 dp2 . . . dpN

•  f (N)(rN, pN ): phase-space probability density


function
–  or, simply, probability density
Molecular Modelling of Fluids Session 2 – Slide 36 © Imperial College London
Probability densities

•  f (N)(rN, pN ) is a probability density since,


multiplied by a volume (drNdpN) of a phase-
space element, it yields a probability.

f (N ) (rN, pN )drN dpN ⌘ f (N ) (r1 , r2 , . . . , rN , p1 , p2 , . . . , pN )


⇥ dr1 , dr2 , . . . drN , dp1 , dp2 , . . . dpN
⌘ “P (H)”
is the “probability that the physical system is in a
microscopic state represented by a phase point lying in the
6N-dimensional phase-space element drN dpN ”

Molecular Modelling of Fluids Session 2 – Slide 37 © Imperial College London


What does f (N)(rN, pN ) look like?

•  By rules of probability, f (N)(rN, pN ) drNdpN must


normalise,
Z i.e.,:
Z
... f (N ) (rN, pN )drN dpN = 1

–  so dimensions f (N)(rN, pN ) are “1 / (rN pN )”

•  Heisenberg uncertainty principle:


–  each product of r and p ! h
•  Planck’s constant
–  dimensions of h: length × momentum
•  Expect f (N)(rN, pN ) = 1 / h3N × “something”
Molecular Modelling of Fluids Session 2 – Slide 38 © Imperial College London
What does f (N)(rN, pN ) look like?

•  We expect the form to be reminiscent of


exp( Ei )
P (Ei ) =
Q
so we expect, e.g., !
1 exp H(rN, pN )
f (N ) (rN, pN ) =
h3N Q
but the form turns out to be
!
(N ) N N 1 1 exp H(rN, pN )
f (r , p ) =
N ! h3N Q

Molecular Modelling of Fluids Session 2 – Slide 39 © Imperial College London


What does f (N)(rN, pN ) look like?
!
(N ) N N 1 1 exp H(rN, pN )
f (r , p ) =
N ! h3N Q
–  the 1 / N ! accounts for indistinguishability of particles (or
molecules)
–  This is exactly analogous to the over-counting we
encountered previously, in adjusting Equation 1.39,
Q = qN
for distinguishable particles, to Equation 1.52,
Q = q N/N !
for indistinguishable particles
Molecular Modelling of Fluids Session 2 – Slide 40 © Imperial College London
What does f (N)(rN, pN ) look like?

•  Similarly
X Z Z
Q= exp( Ei ) “ . . . exp( Ei )”
i
becomes
Z Z
1
Q= ... exp( H(rN, pN ))drN dpN
N !h3N

Molecular Modelling of Fluids Session 2 – Slide 41 © Imperial College London


Separating the contributions

•  To progress further, we need the Hamiltonian,


H , for the system of interest.
•  Generally, H (rN,pN ) separates:
–  sum of kinetic energy
•  K (pN )
–  and potential (“configurational”) energy
•  V (rN )
•  Consequently,
Z Z partition function
Z Z factorises:
1
Q= ... exp( K(pN ))dpN ... exp( V(rN ))drN
N !h3N
(Equation 3.5)
Molecular Modelling of Fluids Session 2 – Slide 42 © Imperial College London
The kinetic (ideal) contribution
Z Z Z Z
1
Q= . . . exp( K(pN ))dpN . . . exp( V(rN ))drN
N !h3N
•  First integral: independent of the potential
–  To examine this, doesn’t matter what intermolecular
potential we choose
–  pick simplest: V (rZN ) = 0 Z
–  Second integral: exp(0)drN = drN = V N
✓ Z ◆
1
–  so QV=0 = exp( K(p N
))dp N
V N
N !h3N
–  Partition function of fluid with only kinetic energy
•  no potential energy; KE only
•  ideal gas
Molecular Modelling of Fluids Session 2 – Slide 43 © Imperial College London
The kinetic (ideal) contribution
(for simple, spherical molecules)

XN
N p2i Particle mass
•  Kinetic energy: K p =
(spherically symmetric molecules) i=1
2m
Z –  whereby
Z N ZZZ
Y ✓ ◆
p2i
... exp K pN dpN = exp dpi
i=1
2m
•  Since no interactions between particles, all momenta
–  and all components of momenta
are independent. Allows factorisation:
ZZZ ✓ ◆ N Z !
p2i Y p2x,i
exp dpi = exp dpx,i
2m i=1
2m
Z ! Z ! !
p2y,i p2z,i
⇥ exp dpy,i exp dpz,i
2m 2m
Molecular Modelling of Fluids Session 2 – Slide 44 © Imperial College London
The kinetic (ideal) contribution
(for simple, spherical molecules)
Z 1
2 2
p
•  Standard integral: exp( ↵ u )du = ⇡/↵
1
–  whereby
Z ! Z ! Z !
p2x,i p2y,i p2z,i
exp dpx,i = exp dpy,i = exp dpz,i
2m 2m 2m
r
2⇡m
=
Z Z –  so that N ZZZ
Y ✓ ◆ ✓ ◆3N/2
p2i 2⇡m
. . . exp K pN dpN = exp dpi =
i=1
2m
•  Finally:
Z Z ✓ ◆3N/2
1 2⇡m
3N
. . . exp K pN dpN = 2
⌘⇤ 3N
h h
(Equation 3.9)
–  Modelling
Molecular Λ is the thermal
of Fluids de Broglie wavelength Session 2 – Slide 45 © Imperial College London
The ideal free energy
•  Equation
Z Z 3.5 Z Z
1
Q= ... exp K(pN ) dpN . . . exp V(rN ) drN
N !h3N
•  thus becomes
1 VN
Qideal =
N ! ⇤3N
•  whereby
Aideal = kB T ln Qideal
= kB T ln V N ln N ! ln ⇤3N
= kB T N ln V (N ln N N) N ln ⇤3
= N kB T ln(V /N ) + 1 ln ⇤3
Molecular Modelling of Fluids Session 2 – Slide 46 © Imperial College London
The ideal gas equation
•  In other words, we have the ideal free energy
✓ ◆
Aideal N
= ln 1 + ln ⇤3 = ln ⇢ 1 + ln ⇤3
N kB T V
(Equation 2.8)
•  Once we have the ideal free energy, we can use
standard thermodynamic relationships to
obtain other properties
•  For example, recall the Fundamental Equation
X
dU = T dS pdV + µi dNi
i
•  and the definition, A = U – TS
Molecular Modelling of Fluids Session 2 – Slide 47 © Imperial College London
The ideal gas equation
•  This gives us X
dA = SdT pdV + µi dNi
i
•  From this, one can obtain expressions for both the
pressure and✓ the◆chemical potential:
✓ ◆
@A @A
p= ; µi =
@V Ni ,T @Ni V,T,Nj6=i
•  or, for a pure component:
✓ ◆ ✓ ◆
@A @A
p= ; µ=
@V N,T @N V,T

Molecular Modelling of Fluids Session 2 – Slide 48 © Imperial College London


The ideal gas equation
•  Recalling that we had
✓ ◆
Aideal N
= ln 1 + ln ⇤3 = ln ⇢ 1 + ln ⇤3
N kB T V
(Equation 2.8)

it is now straightforward to obtain the ideal gas


equation via p = (@A/@V )N,T :
pideal 1
= , i .e., pV = N kB T
N kB T V
•  Finally, recalling kB = R/NA and N = n NA
pV = nRT
Molecular Modelling of Fluids Session 2 – Slide 49 © Imperial College London
The configurational integral

•  Recall:
Z Z Z Z
1
Q= ... exp K(pN ) dpN . . . exp V(rN ) drN
N !h3N
•  So now:
3N Z Z
⇤ ⇤ 3N
Q= . . . exp V(rN ) drN ⌘ Z (Equation
3.11)
N! N!
−  Non-spherical molecules: Q has same general form but Λ–3 ! “Λ*–3 ”
◦  i.e., “Λ* ” incorporates all rotational and vibrational contributions
Z Z is the
–  Z = . . . exp N N
V(r ) dr configurational
(Equation 3.12) integral
Molecular Modelling of Fluids Session 2 – Slide 50 © Imperial College London
The configurational integral

•  Take care:
–  There is some variation among authors concerning the
exact terminology
•  some people say “configurational integral”, some say
“configuration integral”
•  Z is often called the configuration(al) partition function
–  and, in particular, the notation:
•  Many physicists use Z to represent the (full) partition
function, and Q to represent the configurational integral(!).
•  Many authors use subscript(s) N, or NVT
–  e.g., ZN
–  e.g., QNVT
Molecular Modelling of Fluids Session 2 – Slide 51 © Imperial College London
The configurational integral

•  Is very important!!

•  McQuarrie (in Statistical Mechanics):


–  “ZN is called the Classical Configuration Integral. Since
the intermolecular forces depend on the relative
distances between molecules this integral is, in
general, extremely difficult and is essentially
responsible for the research in equilibrium statistical
mechanics.”

Molecular Modelling of Fluids Session 2 – Slide 52 © Imperial College London


“Separating the probabilities”

•  Recall
P (Hi ) ⌘ P (H(rN , pN )) = f (N ) (rN , pN )drN dpN
–  is the probability of the particles being in the 6N-
dimensional element of phase space drNdpN"
•  f (N)(rN,pN )drNdpN is the probability (not simply f (N)(rN,pN ))
because the probability must depend on the size of the
element of phase space."
N N
–  so could express this simply as" P (r , p )
–  Analogous to the partition function, we factorise:
P rN , p N = P rN P p N

Molecular Modelling of Fluids Session 2 – Slide 53 © Imperial College London


Separating the probabilities

•  Recall !
(N ) N N 1 1 exp H(rN, pN ) ⇤ 3N
f (r , p ) = ; Q= Z
N ! h3N Q N!
so
( )( )
1 1 exp V(rN ) exp K(pN )
f (N ) = 1
N ! h3N N! Z ⇤ 3N
( )( )
exp V(rN ) exp K(pN )
= 3N 3N
;
Z h ⇤
Notice that the 1/N! cancels 3N 2 3N/2
⇤ = 2⇡m/ h
Molecular Modelling of Fluids Session 2 – Slide 54 so Planck’s const cancels too
© Imperial College London
Separating the probabilities

•  Collecting:

P (H) ⌘ P rN , pN = f (N ) drN dpN


( ) ( )
exp V(rN ) N exp K(p N
) N
= dr dp
Z h3N ⇤ 3N
= P rN ⇥ P pN
•  P (pN ):
–  of no further interest Equation 4.3

Molecular Modelling of Fluids Session 2 – Slide 55 © Imperial College London


Separating the probabilities

•  P (rN ):
–  the probability that a given (“labelled”) molecule is in
each of the 3-dimensional volume elements dr1 … drN,
i.e., 0 1
molecule 1 (say) in dr1
B and molecule 2 in dr2 C
B C
B and molecule 3 in dr3 C
P rN = Prob B C
B .. C
@ . A
and molecule N in drN
–  probability of any molecule in each element would be
more useful
Molecular Modelling of Fluids Session 2 – Slide 56 © Imperial College London
Building something useful

•  P (rN ) is not particularly useful as it stands


–  usually, molecules are indistinguishable
•  probability of any molecule in each element would be preferable
•  Build up a step at a time:
–  System can “choose” any molecule to occupy dr1
•  molecule 1 or molecule 2 or … or molecule N
–  Probability theory:
“P (1 or 2 or . . . or Nor ) = P (1) + P (2) + . . . + P (Nor )”
exp V(rN )
so ⇢(1)⇤ (rN ) = N P (rN ) = N drN
Z
represents
P (any molecule in dr1 while given molecules occupy dr2 … drN)
Molecular Modelling of Fluids Session 2 – Slide 57 © Imperial College London
Building something useful

•  ρ(1)* (rN ) is not particularly useful either


–  relates to one molecule
•  indicated by superscript “(1)”
but is a function of positions of all the molecules
–  We integrate out r2 … rN
•  yields a function depending only on position r1:
Z Z
(1) N
⇢ (r1 ) = . . . exp V(rN ) dr2 . . . drN
Z
Z Z
N
⌘ . . . exp V(rN ) drN 1
Z
is called a (singlet) density function and relates to
P (any molecule in dr1, irrespective of the positions of the rest)
Molecular Modelling of Fluids Session 2 – Slide 58 © Imperial College London
More useful …

•  Similarly
–  system has (N – 1) choices for second molecule, so
exp V(rN )
⇢(2)⇤ (rN ) = N (N 1) drN
Z
represents P (any molecule in dr1 and any other in dr2
while given molecules occupy dr3 … drN)
–  relates to two molecules
•  indicated by superscript “(2)”
but is a function of positions of all the molecules
–  We integrate out r3 … rN
•  yields a function depending only on positions r1 and r2:

Molecular Modelling of Fluids Session 2 – Slide 59 © Imperial College London


More useful …
Z Z
(2) N (N 1)
⇢ (r1 , r2 ) = . . . exp V(rN ) dr3 . . . drN
Z
Z Z
N (N 1)
⌘ . . . exp V(rN ) drN 2 (Eqn. 4.9)
Z
is called a pair density and relates to
P (any molecule in dr1, any other in dr2, irrespective of
the positions of the rest)
•  By analogy: n and N-particle densities:
Z Z
(n) (n) n N!
⇢ (r1 , . . . , rn ) ⌘ ⇢ (r ) = ... exp V(rN ) drN n
(N n)!Z

exp V(rN )
⇢(N ) (rN ) = N !
Z
Molecular Modelling of Fluids Session 2 – Slide 60 © Imperial College London
Density functions:
relation to probabilities
Z Z
•  Recalling Z = ... exp V(rN ) drN
and integrating out ! normalisation properties:
exp V(rN )
⇢(N ) (rN ) = N !
Z
Z Z Z Z
(N ) N N exp V(rN )
. . . ⇢ (r )dr = N ! . . . drN = N !
Z
1 (N ) N
–  so ⇢ (r )drN
N!
is P (any molecule in dr1, any in dr2, …, any in drN)

Molecular Modelling of Fluids Session 2 – Slide 61 © Imperial College London


Density functions:
relation to probabilities
Z Z
•  Recalling Z = ... exp V(rN ) drN
and integrating out ! normalisation properties:
Z Z
(1) N
⇢ (r1 ) = . . . exp V(rN ) drN 1
Z
Z Z Z Z
(1) (1) N
⇢ (r1 )dr1 ⌘ ⇢ (r)dr = . . . exp V(rN ) drN = N
Z
1 (1)
–  so ⇢ (r)dr
N
–  is P (any molecule in dr1 irrespective of the positions of
the remainder)
Molecular Modelling of Fluids Session 2 – Slide 62 © Imperial College London
Density functions:
relation to probabilities

•  Keeping in mind
Z Z Z
(1) N
⇢ (r1 )dr1 = . . . exp V(rN ) drN = N
Z
–  Imagine simple, homogeneous fluid
•  ⇢(1) (r) = ⇢(1) , or simply ⇢
Z Z
(1) N
–  so ⇢ (r)dr = N = ⇢ dr = ⇢V =) ⇢ =
⇣ V ⌘
Equations
–  i.e., ρ(1)(r) is the number density at r 4.15, 4.16 & 4.17

Molecular Modelling of Fluids Session 2 – Slide 63 © Imperial College London


Density functions:
relation to probabilities

•  Analogously, for the Zpair Zdensity


(2) N (N 1)
⇢ (r1 , r2 ) = . . . exp V(rN ) drN 2
Z
ZZ Z Z
(2) N (N 1)
⇢ (r1 , r2 ) dr1 dr2 = . . . exp V(rN ) drN
Z
= N (N 1)
1
–  So ⇢(2) (r1 , r2 ) dr1 dr2
N (N 1)
is P (any molecule in dr1, any molecule in dr2,
irrespective of the positions of the remainder)
•  “P (molecule at r1 and another at r2)”
Molecular Modelling of Fluids Session 2 – Slide 64 © Imperial College London
Density functions ! g (r)

•  Define n-particle distribution function:


(n) n
(n) n ⇢ (r )
g (r ) = Qn (1) (r )
i=1 ⇢ i

–  e.g., pair distribution function


(2)
(2) ⇢ (r1 , r2 )
g (r1 , r2 ) = (1) (Equation. 4.21)
⇢ (r1 )⇢(1) (r2 )
•  If we also define P (r2 |r1 ) = (1/N 1)⇢† (r2 |r1 )dr2
⇢(2) (r1 , r2 )dr1 dr2 = ⇢(1) (r1 )dr1 ⇢† (r2 |r1 )dr2 ,
so ⇢(2) (r1 , r2 ) = ⇢(1) (r1 ) ⇢† (r2 |r1 )
Molecular Modelling of Fluids Session 2 – Slide 65 © Imperial College London
Density functions ! g (r)

•  Pair distribution function becomes:


(1) †
(2) ⇢ (r 1 ) ⇢ (r2 |r1 )
g (r1 , r2 ) = (1)
⇢ (r1 )⇢(1) (r2 )
•  If system is isotropic: ⇢† (r2 |r1 ) ⇢† (r) ⌘ ⇢(r)
•  If homogeneous: ⇢(1) (r) = N/V ⌘ ⇢
•  If both isotropic and homogeneous:
(2) ⇢(r)
g (r1 , r2 ) g(r) = (Equation. 4.24)
iso,homog ⇢
(Pair) radial distribution function
Molecular Modelling of Fluids Session 2 – Slide 66 © Imperial College London
Importance of g (r)

•  Tells us about structure:

Molecular Modelling of Fluids Session 2 – Slide 67 © Imperial College London


Importance of g (r)

•  Link to experiment
–  closely related to structure factor (scattering function)
from scattering experiments
and computer simulations
–  accumulated as a histogram by
counting molecules in concentric
shells around origin molecule,
averaged over all molecules

Molecular Modelling of Fluids Session 2 – Slide 68 © Imperial College London


Importance of g (r)

•  Crucial role in theories of fluids


–  Arises out of approximation that intermolecular
potential is expressible as a sum of effective pair
interactions:
N X
X N
V(rN ) = (ri,j ) Equation 4.25
i j>i

–  Demonstrate by example …

Molecular Modelling of Fluids Session 2 – Slide 69 © Imperial College London


g(r) leads to the Energy equation

N N
•  hH(r , p )i = hV(rN )i + hK(pN )i
l l l
U = U residual + U ideal
X
•  By definition hxi = xi P (xi )
•  whence Z Z
i

U residual = . . . V(rN )P (V(rN ))drN


Z Z
1
= . . . V(rN ) exp( V(rN ))drN
Z
(Equation 4.26)
Statistical-mechanical average of V
Molecular Modelling of Fluids Session 2 – Slide 70 © Imperial College London
g(r) leads to the Energy equation

•  Incorporating equation
0 4.25 gives:
1
Z Z N X
X N
1 @
U residual = ... (ri,j )A exp V(rN ) drN
Z i j>i

•  Examine double sum in detail:


1,2 + 1,3 + 1,4 + ... + 1,N (N 1 terms)
N X
X N + 2,3 + 2,4 + ... + 2,N (N 2 terms)
(ri,j ) = + 3,4 + ... + 3,N (N 3 terms)
i j>i .. .. .. ..
. . . .
( shorthand: i,j ⌘ (r)i,j ) + N 1,N (1 term)
1 2 1
](terms) = (N 1) (N 1) + (N 1) = N (N 1)
2 2
Molecular Modelling of Fluids Session 2 – Slide 71 © Imperial College London
g(r) leads to the Energy equation

•  Integral of sum becomes a sum of N(N – 1)/2


integrals: ⇢Z Z
residual 1
U = ... (r1,2 ) exp( V(rN ))drN
Z
Z Z
+ ... (r1,3 ) exp( V(rN ))drN
Z Z
+... + ... (r1,N ) exp( V(rN ))drN
Z Z
+ ... (r2,3 ) exp( V(rN ))drN + . . .
Z Z
+ ... + ... (rN 1,N ) exp( V(rN ))drN

Molecular Modelling of Fluids Session 2 – Slide 72 © Imperial College London


Energy equation

•  All ϕ(ri,j) equivalent, whereby all integrals


identical, thus:
Z Z
residual N (N 1)
U = ... (r1,2 ) exp( V(rN ))dr1 dr2 . . .drN
2Z
•  Now we factorize the integral
–  ϕ(r1,2) depends on r1 and r2 only
ZZ Z Z
residual N (N 1)
U = (r1,2 ) ... exp( V(rN ))dr3 . . .drN dr1 dr2
2Z
ZZ Z Z
N (N 1)
⌘ (r1,2 ) ... exp( V(rN ))drN 2
dr1 dr2
2Z

Molecular Modelling of Fluids Session 2 – Slide 73 © Imperial College London


Energy equation

•  Carrying forward:
ZZ Z Z
residual N (N 1)
U = (r1,2 ) . . . exp V(rN ) drN 2
dr1 dr2
2Z
•  Next recall Equation 4.9
Z Z
(2) N (N 1)
⇢ (r1 , r2 ) = ... exp V(rN ) drN 2
Z
and compare:
Z Z
residual 1
U = ... (r1,2 )⇢(2) (r1 , r2 )dr1 dr2
2

Molecular Modelling of Fluids Session 2 – Slide 74 © Imperial College London


g(r) leads to the Energy equation

•  Carrying forward: Z Z
residual 1
U = ... (r1,2 )⇢(2) (r1 , r2 )dr1 dr2
2
•  Now recall:
–  Equation 4.21 ⇢(2) (r1 , r2 ) = g (2) (r1 , r2 ) ⇢(1) (r1 ) ⇢(1) (r2 )
(2)
–  Isotropic system: g (r1 , r2 ) g (2) (|r2 r1 |) ⌘ g(r1,2 )
–  Homogenous system ⇢(r) ⇢
•  equations 4.15–4.17

•  whereby ZZ
residual 1 2
U = ⇢ (r1,2 )g(r1,2 )dr1 dr2
2
Molecular Modelling of Fluids Session 2 – Slide 75 © Imperial College London
Energy equation
•  Carrying forward: ZZ
residual 1 2
U = ⇢ (r1,2 )g(r1,2 )dr1 dr2
2
•  Integral over vector r1, r2 but integrand depends
only on (scalar) r1,2
@r1,2
–  transform: r1,2 = r2 r1 =1 dr1,2 = dr2
@r2
–  in spherical polar coordinates
Z Z ⇡ Z 2⇡ Z 1
2
dr1,2 = r1,2 sin ✓ d✓ d' dr1,2
✓=0 '=0 r=0
•  whereby
Z Z ⇡ Z 2⇡ Z 1
1 2
U res = ⇢ 2
(r1,2 ) r1,2 g(r1,2 ) sin ✓ d✓ d' dr1,2 dr1
2 r1 ✓=0 '=0 r1,2 =0
Molecular Modelling of Fluids Session 2 – Slide 76 © Imperial College London
Energy equation
•  Carrying
Z Zforward:
⇡ Z 2⇡ Z 1
res 1 2 2
U = ⇢ (r1,2 ) r1,2 g(r1,2 ) sin ✓ d✓ d' dr1,2 dr1
2 r1 ✓=0 '=0 r1,2 =0
Z ZZZ
•  Integrate out r1: dr1 ⌘ drx,1 dry,1 drz,1 = V
•  Integrate out angular dependences ! 4π
•  Yields: (Equation 4.37)
Z 1
U residual = 2⇡⇢2 V 2
(r1,2 ) r1,2 g(r1,2 ) dr1,2
0
Z 1
or U residual = 2⇡⇢N 2
(r1,2 ) r1,2 g(r1,2 ) dr1,2
0

Molecular Modelling of Fluids Session 2 – Slide 77 © Imperial College London


Energy equation

•  “Human” interpretation:

Molecular Modelling of Fluids Session 2 – Slide 78 © Imperial College London


Virial

•  A similar argument leads to the pressure


equation or virial
–  will present just an outline
•  for details, see e.g., McQuarrie, Statistical Mechanics; Hansen
and McDonald, Theory of Simple Liquids; …

•  Begin with A = – kBT ln Q, whereby


✓ ◆ ✓ ◆
@ ln Q @ ln Q
p = kB T pV = kB T V
@V N,T @V N,T

•  but, keeping in mind ideal gas equation


pV = N kB T + (pV )residual
Molecular Modelling of Fluids Session 2 – Slide 79 © Imperial College London
Virial

•  Residual part:
–  obtained from force, – dϕ(r1,2) / dr1,2, between a specified
molecule and all other molecules, accounted for in
concentric shells at increasing separation, r :

Molecular Modelling of Fluids Session 2 – Slide 80 © Imperial College London


Virial

•  Residual part:
–  obtained from force, – dϕ(r1,2) / dr1,2, between a specified
molecule and all other molecules, accounted for in
concentric shells at increasing separation, r :
–  for each shell, force is divided by surface area (4πr2dr) !
pressure
d (r1,2 )
p(r1,2 ) = ⇢g(r1,2 )
dr1,2
–  multiplied by volume contained within shell ! pV for each
shell
d (r1,2 ) 4 3
p(r1,2 )V = ⇢g(r1,2 ) ⇥ ⇡r1,2
dr1,2 3
Molecular Modelling of Fluids Session 2 – Slide 81 © Imperial College London
Virial

•  Residual part:
•  (carried forward)
d (r1,2 ) 4 3
p(r1,2 )V = ⇢g(r1,2 ) ⇥ ⇡r1,2
dr1,2 3
–  integrate from r = 0 to ∞ ! total (pV)residual per molecule
Z 1
4⇡⇢ d (r1,2 ) 3
pV(1) = g(r1,2 )r1,2 dr1,2
3 0 dr1,2
–  Multiply by N / 2 ! contribution for all N molecules
•  factor of half avoids counting each pair interaction twice
Z 1
residual 2N ⇡⇢ d (r1,2 ) 3
pV(N ) ⌘ (pV ) = g(r1,2 )r1,2 dr1,2
3 0 dr1,2
Molecular Modelling of Fluids Session 2 – Slide 82 © Imperial College London
Virial

•  Residual part:
•  (carried forward) Z 1
residual 2N ⇡⇢ d (r1,2 ) 3
pV(N ) ⌘ (pV ) = g(r1,2 )r1,2 dr1,2
3 0 dr1,2
•  Now combine with ideal part:
Z 1
2N ⇡⇢ d (r1,2 ) 3
pV = N kB T g(r1,2 )r1,2 dr1,2
3 0 dr1,2
•  Usual to express in terms of compressibility
factor,
Z ⌘ pV /N kB T = p/⇢kB T

Molecular Modelling of Fluids Session 2 – Slide 83 © Imperial College London


Virial

•  whence we obtain the pressure equation or


virial:
Z 1
p 2⇡⇢ d (r1,2 ) 3
Z= =1 g(r1,2 )r1,2 dr1,2
⇢kB T 3kB T 0 dr1,2
–  NB:
•  don’t confuse Z with the configurational integral
•  this equation is not the same as the virial equation of
state (or virial expansion)

Molecular Modelling of Fluids Session 2 – Slide 84 © Imperial College London


In the last few slides …

•  We demonstrated the usefulness of g(r) in


expressing 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

•  Expressions are simpler in form; difficulty is


shifted to determining g(r)

Molecular Modelling of Fluids Session 3 – Slide 85 © Imperial College London


Molecular Modelling of Fluids Session 2 – Slide 86 © Imperial College London
Z 1 Z 0
2 2
exp( ↵ u )du = 2 exp( ↵2 u2 )du
1 1
Z 1 click

= 2 exp( ↵2 u2 )du
0

since the integration variable (u) appears in the integrand only as its square.
s⇢
Z 1 Z 1 2
2 exp( ↵2 u2 )du = 2 exp( ↵2 u2 )du
0 0
s ⇢Z Z
1 1
= 2 exp( ↵2 u2 )du exp( ↵2 v 2 )dv
0 0
s ⇢Z Z
1 1
= 2 exp( ↵2 (u2 + v 2 ))dudv .
0 0

In polar coordinates, u2 + v 2 = r2 ; area element du dv r d✓ dr, hence


v( )
Z 1 u Z ⇡/2 Z 1
u
2 exp( ↵ u )du = 2 t
2 2
exp( ↵2 r2 )rd✓ dr
0 ✓=0 r=0
s⇢  s⇢ p
1
⇡ exp( ↵2 r 2 ) ⇡ 1 ⇡
= 2 =2 = .
Molecular Modelling of Fluids 2 2↵2
Session 2 – Slide 87
r=0 2↵2
2© Imperial College London

Virial

•  A similar argument leads to the pressure


equation or virial:
Z 1
p 2⇡⇢ d (r1,2 ) 3
Z= =1 g(r1,2 )r1,2 dr1,2
⇢kB T 3kB T 0 dr1,2
where Z is the compressibility factor

(Note that this is not the same as the virial equation of state (or
virial expansion))

Molecular Modelling of Fluids Session 2 – Slide 88 © Imperial College London

You might also like