Introduction to (Statistical)

Thermodynamics
 Molecular Simulations
MD
 Molecular dynamics: solve
equations of motion
r1
r2
rn
MC
 Monte Carlo: importance
sampling r1
r2
rn 2
3
4
 Questions
What is the desired distribution?

• How can we prove that this scheme

generates the desired distribution of
configurations?
• Why make a random selection of the
particle to be displaced?
• Why do we need to take the old
configuration again?
• How large should we take: delx?

5
 Summary
Thermodynamics
– First law: conservation of energy
– Second law: in a closed system entropy increase
and takes its maximum value at equilibrium
System at constant temperature and volume
– Helmholtz free energy decrease and takes its
minimum value at equilibrium
Equilibrium:
– Equal temperature, pressure, and chemical
potential

6
 Entropy
Closed system:
E,V •1st law: total energy remains constant
•2nd law: entropy increases

E = E (S,V ,N i)
M
dE = TdS-pdV + ∑μidN i
i=1

dS ≥0
⎛ ∂E ⎞ 1 ⎛ ∂S ⎞
Temperature T =⎜ ⎟ or =⎜ ⎟
⎝ ∂S ⎠V ,N i T ⎝ ∂E ⎠V ,N i
7
 Helmholtz free energy
n
F ≡ E −TS dE = TdS-pdV + ∑μidN i
n i=1

dF = −SdT − pdV + ∑μidN i

i=1
F = F (T ,V ,N i) dF ≤0

⎛ ∂F ⎞
Pressure: p=- ⎜ ⎟
⎝ ∂V ⎠T ,N i
⎛∂F T ⎞
Energy: E =⎜
⎝ ∂1 T ⎟
⎠ V ,N i

⎛ ∂F ⎞
Chemical potential: μi = ⎜
⎝ ∂N i ⎟
⎠T ,V , N j 8
 Statistical Thermodynamics
System of N molecules: r1 K rN , v 1 K v N

But how do we obtain the

macroscopic properties of this system
from this microscopic information?

9
 Outline (2)
Statistical Thermodynamics
– Basic Assumption
• micro-canonical ensemble
• relation to thermodynamics
– Canonical ensemble
• free energy
• thermodynamic properties
– Other ensembles
• constant pressure
• grand-canonical ensemble

10
 Statistical Thermodynamics:
the basics
• Nature is quantum-mechanical
• Consequence:
– Systems have discrete quantum states.
– For finite “closed” systems, the number of states is
finite (but usually very large)
• Hypothesis: In a closed system, every state
is equally likely to be observed.
• Consequence: ALL of equilibrium
Statistical Mechanics and
Thermodynamics
11
 …, but there are not
Each individual many microstates that
microstate is Basic assumption
give these extreme
equally probable results

If the number of
particles is large (>10)
these functions are
sharply peaked12
 Does the basis assumption lead to
something that is consistent with classical
thermodynamics?

E1 E2 = E −E1
Systems 1 and 2 are weakly coupled
such that they can exchange energy.

What will be E1?

Ω (E1 , E − E1 )= Ω1 (E1 )× Ω 2 (E − E1 )

BA: each configuration is equally probable; but the number of

states that give an energy E1 is not know.
13
 Ω (E1 , E − E1 )= Ω1 (E1 )× Ω 2 (E − E1 )
ln Ω (E1 , E − E1 )= ln Ω1 (E1 )+ ln Ω 2 (E − E1 )
⎛ ∂ ln Ω (E1 , E − E1 )⎞ Energy is conserved!
⎜ ⎟ =0 dE1=-dE2
⎝ ∂E1 ⎠N1 ,V1
⎛ ∂lnΩ1 (E1 )⎞ ⎛ ∂lnΩ2 (E − E1)⎞
⎜ ⎟ +⎜ ⎟ =0
⎝ ∂E1 ⎠N ⎝ ∂E1 ⎠N
V1
1, V2
2,

This can be seen as

⎛ ∂ ln Ω1 (E1 )⎞ ⎛ ∂ ln Ω 2 (E − E1 )⎞
⎜ ⎟ =⎜ ⎟ an equilibrium
⎝ ∂E1 ⎠N1 ,V1 ⎝ ∂E2 ⎠N2 ,V2 condition
⎛ ∂ ln Ω (E )⎞
β ≡⎜ ⎟
⎝ ∂E ⎠N ,V
β1 = β 2 14
 Entropy and number of
configurations
Conjecture: S = ln Ω
Almost right.
•Good features:
•Extensivity
•Third law of thermodynamics comes for free
•Bad feature:
•It assumes that entropy is dimensionless but (for
unfortunate, historical reasons, it is not…)

15
We have to live with the past, therefore

S = k B ln Ω (E )
With kB= 1.380662 10-23 J/K
In thermodynamics, the absolute (Kelvin)
temperature scale was defined such that
n
⎛ ∂S ⎞ 1 dE = TdS-pdV + ∑μidN i
⎜ ⎟ =
⎝ ∂E ⎠N ,V T i=1
But we found (defined):
⎛ ∂ ln Ω (E )⎞
β ≡⎜ ⎟
⎝ ∂E ⎠N ,V
16
And this gives the “statistical” definition of temperature:

1 ⎛ ∂ ln Ω (E )⎞
≡ kB ⎜ ⎟
T ⎝ ∂E ⎠N ,V
In short:
Entropy and temperature are both related to
the fact that we can COUNT states.

Basic assumption:
1. leads to an equilibrium condition: equal temperatures
2. leads to a maximum of entropy
3. leads to the third law of thermodynamics
17
 Number of configurations
How large is ?
•For macroscopic systems, super-astronomically large.
•For instance, for a glass of water at room temperature:

25
2×10
Ω ≈10
•Macroscopic deviations from the second law of
thermodynamics are not forbidden, but they are
extremely unlikely.
18
 Canonical ensemble 1/kBT

Consider a small system that can exchange heat with a big reservoir

∂ ln Ω
Ei E −Ei ln Ω (E − Ei )= ln Ω (E )− Ei + L
∂E
Ω (E − Ei ) Ei
ln =−
Ω (E ) k BT

Hence, the probability to find Ei:

Ω (E − Ei ) exp (− Ei k BT )
P (Ei )= =
∑ Ω (E − E ) ∑
j j j
exp (
− E j k BT )

P (Ei )∝ exp (− Ei k BT )
19
Boltzmann distribution
 Thermodynamics
Thermo recall (2)
What is the average energy
First law of the system?
of thermodynamics
∑ dE = Tdi SE− pdV(− β Ei )
i exp
E ≡ ∑ E P (E )=
exp (
−β E )
i i
i
∑ energy:
Helmholtz Free j j

F ≡E −TS
∂ ln ∑ exp (−β E ) i
dF = −Sd=T−− pdV i
∂β
⎛∂ F T ⎞ 1 ∂F ∂F
⎜ ⎟ = F∂+ln QN ,V ,T= F −T
⎝ ∂1 T ⎠= − T ∂1 T ∂T
Compare: ∂β
= F + TS = E
⎛∂ F T ⎞ F
⎜ ⎟= E Hence: = −ln QN ,V ,T
⎝ ∂1 T ⎠ k BT
20
 Remarks (1)
We have assume quantum mechanics (discrete states) but
we are interested in the classical limit
1 ⎧
⎪ ⎡ p 2
N ⎤⎪
⎫
∑ i exp (−β Ei )→ h3 N ! ∫∫ dp dr exp ⎨−β ⎢∑ i 2m + U (r )⎥⎬
N N i

⎪
⎩ ⎣ i ⎦⎪
⎭
1
3
→ Volume of phase space (particle in a box)
h
1
→ Particles are indistinguishable
N!
Integration over the momenta can be carried out for most systems:
3N 3
N
⎧
⎪ ⎡ p 2
⎤⎫
⎪ ⎡ ⎧ p 2
⎫⎤ ⎛ 2π m ⎞ 2

∫dp exp ⎨ ⎥⎬= ⎢∫dp exp ⎨−β =⎜

N i
−β ⎢∑i ⎬⎥ ⎟
⎪
⎩ ⎣ 2 m i ⎦⎪
⎭ ⎣ ⎩ 2 m ⎭⎦ ⎝ β ⎠
21
 Remarks (2)

Define de Broglie wave length:

1
⎛h β ⎞22
Λ ≡⎜ ⎟
⎝ 2π m ⎠
Partition function:

1 ⎡−β U (
Q (N , V , T )=
Λ3 N N ! ∫ dr N
exp ⎣ r N
)
⎤
⎦

22
 Example: ideal gas
1
Q (N , V , T )Thermo
= 3N
Λ N! ∫ exp ⎡
dr (3)
recall
N
⎣− β U (
r )
⎤
⎦
N

Helmholtz Free energy:N

1 V
= dF3 N= −S∫
d dr− p1d=V 3 N
T
N

Λ N! Λ N!
Free energy: Pressure
⎛ ⎛
∂F ⎞V N
⎞
β F = −ln⎜ ⎜ ⎟3 N = −P⎟
⎝ ∂⎝VΛ⎠T N ! ⎠

Energy: ⎛N ⎞
≈ N ln Λ + N ln ⎜ ⎟ = N ln Λ3 + N ln ρ
3

Pressure: ⎛∂ F T
⎝ V⎞ ⎠
⎞ ⎛ ∂β F
⎜ ⎟= ⎜ ⎟= E
⎝ ∂1 T ⎠ ⎝ Energy:
∂β ⎠
⎛ ∂F ⎞ N
P = −⎜ ⎟ = ⎛ ∂β F ⎞ 3N ∂Λ 3
⎝ ∂V ⎠T β V E =⎜ ⎟= = Nk BT
⎝ ∂β ⎠ Λ ∂β 2 23
 Ideal gas (2)
⎛ ∂F ⎞
Chemical potential: μi = ⎜
⎝ ∂N i ⎟
⎠T ,V , N j
⎛N ⎞
β F = N lnΛ + N ln⎜ ⎟
3

⎝V ⎠

βμ = lnΛ3 + ln ρ +1

βμ IG = βμ 0 + ln ρ

24
 Summary:
Canonical ensemble (N,V,T)
Partition function:

1
Q (N , V , T )= 3 N ∫ dr exp ⎣− β U (
⎡ N
r )
N
⎤
⎦
Λ N!
Probability to find a particular configuration

P (Γ )∝ exp ⎡⎣− β U (Γ )⎤⎦

Free energy

β F = −ln QN ,V ,T
25
 Summary:
micro-canonical ensemble (N,V,E)
Partition function:
1
(( ) )
Q (N ,V , E )= 3 N ∫∫ dp N dr N δ H p N ,r N − E
h N!

Probability to find a particular configuration

P (Γ )∝ 1
Free energy

β S = ln QN ,V , E
26
 Thermodynamic properties

Probability to find a configuration: Γ = {r1 , r2 , K , rN }

P (Γ )∝ exp ⎡
⎣− β U (Γ )⎤⎦
Ensemble average:

1
A = ∫ d ΓA (Γ )exp ⎡⎣− β U (Γ )⎤⎦
Q

For many properties the

momenta do not matter: only
the configurational part27
 Other ensembles?
COURSE:
In the thermodynamic limit the thermodynamic
MD andproperties
MC are
independent of the ensemble: so buy different
a bigger computer
ensembles…

However, it is most of the times much better to think and to carefully

select an appropriate ensemble.

For this it is important to know how to simulate in the various

ensembles.

But for doing this wee need to know the Statistical Thermodynamics
of the various ensembles.

28
 Example (1):
vapour-liquid equilibrium mixture

Measure the composition of

T the coexisting vapour and
V
L+V
liquid phases if we start with
L a homogeneous liquid of two
different compositions:
– How to mimic this with the
N,V,T ensemble?
– What is a better ensemble?
composition

29
 Example (2):
swelling of clays
Deep in the earth clay layers
can swell upon adsorption of
water:
– How to mimic this in the N,V,T
ensemble?
– What is a better ensemble to
use?

30
 Ensembles
• Micro-canonical ensemble: E,V,N
• Canonical ensemble: T,V,N
• Constant pressure ensemble: T,P,N
• Grand-canonical ensemble: T,V,μ

31
 Constant pressure simulations:
N,P,T ensemble
Vi , Ei E −Thermo
Ei , recall (4)
V −Vi
1/kBT p/kBT
First law ofConsider
thermodynamics
a small system that can exchange
dE = TdS volume andμenergy
− pdV +∑ i i
d N i
with a big reservoir

Hence ⎛ ∂ ln Ω ⎞ ⎛ ∂ ln Ω ⎞
ln Ω ( V − Vi , E − Ei )
1 ⎛ ∂S (
= ln Ω ⎞V , E )− ⎜ ⎟ Ei − ⎜ ⎟ Vi + L
=⎜ ⎟ ⎝ ∂E ⎠V ⎝ ∂V ⎠E
Ω (E − Ei , V − Vi )T ⎝ ∂E E ⎠V , N
i pVi
ln =− −
Ω (E , V )and k BT k B T
∂S ⎞ p
Hence, the probability⎛⎜to find⎟ =i i:
E ,V
⎝ ∂V ⎠T , N T exp ⎡⎣− β (Ei + pVi )⎤⎦
Ω (E − Ei , V − Vi )
P (Ei , Vi )= =
( ) (
∑ j ,k Ω E − E j ,V − Vk ∑ j ,k exp ⎡⎣−β E j + pVk ⎤⎦ )
∝ exp ⎡⎣−β (Ei + pVi )⎤⎦ 32
 N,P,T ensemble (2)
In the classical limit, the partition function becomes
1
()
Q (N , P, T )= 3 N ∫ dV exp (−β PV )∫ dr N exp ⎡⎣− β U r N ⎤⎦
Λ N!
The probability to find a particular configuration: r N ,V

( ) ( ( ))
P r N , V ∝ exp ⎡−β PV + U r N ⎤
⎣ ⎦

33
 Grand-canonical simulations:
μ,V,T ensemble
N i , Ei E −Thermo
Ei , recall (5)
N −N i
1/kBT -μ/kBT
First law ofConsider
thermodynamics
a small system that can exchange
dE = TdSparticles
− pdV +∑ andμenergy
i i
d N i
with a big reservoir

Hence ⎛ ∂ ln Ω ⎞ ⎛ ∂ ln Ω ⎞
ln Ω (N − N i , E − Ei )=1 ln⎛Ω∂S(⎞
N , E )− ⎜ ⎟ Ei − ⎜ ⎟ Ni + L
=⎜ ⎟ ⎝ ∂E ⎠N ⎝ ∂N ⎠E
Ω (E − Ei , N − N i )T ⎝ ∂E E ⎠V , N
i μ i Ni
ln =− +
Ω (E , N )and k BT k B T
⎛ ∂S ⎞ μi
Hence, the probability ⎜ to ⎟ =i − i:
find E ,N
∂N i ⎠T ,V T exp ⎡⎣−β (Ei − μ i N i )⎤⎦
Ω (E − Ei , N − N i )
⎝
P (Ei , N i )= =
( ) (
∑ j ,k Ω E − E j , N − N k ∑ j ,k exp ⎡⎣−β E j − μ k N k ⎤⎦ )
∝ exp ⎡⎣−β (Ei − μ i N i )⎤⎦ 34
 μ,V,T ensemble (2)
In the classical limit, the partition function becomes
(
∞ exp βμ N
)
Q (μ , V , T )= ∑ 3N
Λ N! ∫ d r N
exp ⎡
⎣− β ()
U r N
⎤
⎦
N =1

The probability to find a particular configuration: N , r N

( ) ()
P N , r N ∝ exp ⎡⎣βμ N − β U r N ⎤⎦

35