Professional Documents
Culture Documents
by
Wing Kam Liu, Eduard G. Karpov, Harold S. Park
4. Methods of Thermodynamics and Statistical Mechanics
Definition and Features the Thermodynamic Method
The connections between thermodynamic parameters are found from the general laws of
thermodynamics.
Thermodynamics is in fact a theory of equilibrium states, i.e. the states with time-
independent (relaxed) V, P, T and S. Term dynamics is understood only in the sense how
one thermodynamic parameters varies with a change of another parameter in two successive
equilibrium states of the system.
Classification of Thermodynamic Parameters
Internal parameters are determined by the state of the system itself for given values of the
external parameters.
Note: the same parameter may appear as external in one system, and as internal in another
system.
Intensive parameters are independent of the number of particles in the system, and they serve
as general characteristics of the thermal atomic motion (temperature, chemical potential).
Extensive parameters are proportional to the total mass or the number of particles in the
system (internal energy, entropy).
A same parameter may appear both as external and internal in various systems:
System A System B
T Const
M
P P = Const
V = Const V
Application of the thermodynamic method implies that the system if found in the state of
thermodynamic equilibrium, denoted X, which is defined by time-invariant state parameters, such
as volume, temperature and pressure:
X (V , T , P)
f (V , T , P, ) 0
The knowledge of an equation of state allows evaluation of a group of the microscopic system
parameters, such as the compressibility, expansion and pressure coefficients:
f (V , T , P, ) 0
1 V
1) Isothermal compressibility coefficient T
V P T
1 V
2) Isobaric thermal expansion coefficient P
V T P
1 P
3) Isochoric pressure coefficient V
P T V
Examples of the State Equation
Standard forms of the state equations are readily available for ideal gases, real (Van der Waals)
gases, homogeneous liquid and homogeneous isotropic solids.
Ideal gas
PV NkBT
2 a N
2
P n V nb Nk BT n
V NA d effective diameter of
the molecules
2
a 2 N A2 W2 (r )r 2 dr , b N A d 3 W pairwise potential
d
3
Homogeneous isotropic liquid or solid
V V0 1 (T T0 ) ( P P0 )
V0 V (T0 , P0 )
arbitrary initial state
1 V 1 V
,
V0 T P P0 V0 P T T0
The Postulate on Existence of Temperature
Consequently,
All internal parameters of an equilibrium system are functions of the
external parameters and temperature.
The First Law of Thermodynamics
The first law of thermodynamics is essentially a form of the energy conservation law,
written in relation to thermodynamic systems:
Q dU W
An amount of heat absorbed by the system is equal to the summary change of its
internal energy and the work done by the system over external bodies.
Note: The internal energy U is defined solely by the state of system, while the external
thermal energy Q and the mechanical work W may depend both on the internal state of
the system and other factors.
The Second Law of Thermodynamics
The most important consequence of the second law of thermodynamics is that it asserts
the existence of a new function of state, namely the entropy, S.
The concept of entropy plays a crucial role in statistical mechanics.
For quasistatic processes, entropy is an extensive state function, which is defined by the
relation
Q TdS
Based on the first law of thermodynamics we obtain the fundamental differential
equation
dU TdS W TdS PdV
Alternative form of the second law of thermodynamics: All spontaneous processes in
adiabatically isolated systems, Q = 0, occur at a constant or growing entropy:
S 0
The Third Law of Thermodynamics
The second law and the original definition of entropy does not specify the absolute
value of entropy, whose value is provided up to a constant value. The third law
claims that
This allows establishing an absolute scale for measuring the entropy; so that
T 0: S 0
Also,
T 0 : C P V 0
Thermodynamic Potentials
The simplest thermodynamic potential is the internal energy, given by the first law,
U ( S ,V ) : dU Q W TdS PdV
U U
T P
S V V S
G G
S V
T P P T
I I
T V
S P P S
F F
S P
T V V T
Also, system parameters can be evaluated using their relationship with the above state
parameters (see reading assignment [1], Section 4.1.5)
Hamiltonian Mechanics
Equations of motion:
H H
q , p
p q
L
H ( p, q, t ) p q L, p
q
Cartesian coordinates:
p2
H (p, r, t ) T U U (r, t ), p 2 m2 r 2 m2 ( x 2 y 2 z 2 )
2m
Micromodel vs. Macromodel
deterministic
conformity
probabilistic
conformity
Distribution Function
w p1 , p2 ,..., ps ; q1 , q2 ,..., qs ; t
or abbreviated:
w( p, q, t )
( p ,q )
w( p, q, t )dpdq 1
Statistical Ensemble
Within the statistical description, the motion of one single system with given
initial conditions is not considered; thus, p(t), q(t) are not sought.
Instead, the motion of a whole set of phase points, representing the collection of
possible states of the given system.
q
Statistical Averaging
Statistical average (expectation) of an arbitrary physical quantity F(p,q), is given
most generally by the ensemble average,
F (t )
( p ,q )
F ( p, q) w( p, q, t )dpdq
t
For some standard equilibrium systems, thermodynamic parameters can be
obtained, using a single phase space integral. This approach is discussed below.
Ergodic Hypothesis and the Time Average
Evaluation of the ensemble average (previous slide) requires the knowledge of the
distribution function w for a system of interest.
Alternatively, the statistical average can be obtained by utilizing the ergodic
hypothesis in the form,
F F
Here, the right-hand side is the time average
(in practice, time t is chosen finite, though as large as possible)
1 t
F F p( ), q( ) d , t
t 0
This approach requires F as a function of the generalized coordinates.
Some examples
Internal energy: U H ( p, q )
2 Ek Here,
Temperature: T Ek mean kinetic energy per degree of freedom
kB
v mean velocity of molecules
U2 dU T2 C
Change of entropy: S V dT l mean free path of gas particles
U1 T T1 T
vl
Gas diffusion constant: D More examples are given in Ref. [1].
3
Law of Motion of a Statistical Ensemble
Notes: the Hamiltonian gives the total energy required; the Hamiltonian may depend
on the values of external parameters a = (a1, a2,), besides the phase vector X.
This distribution function satisfies the equilibrium equation of motion, because
[ H , ( H )] 0
Exercise: Check the above equality.
Canonical Ensembles
After adoption of the ergodic hypothesis, it then remains to determine the actual
form of the function (H). This function depends on the type of the
thermodynamic system under consideration, i.e. on the character of the interaction
between the system and the external bodies.
1) Adiabatically isolated systems that have no contact with the surroundings and
have a specified energy E.
The corresponding statistical ensemble is referred to as the microcanonical
ensemble, and the distribution function microcanonical distribution.
2) Closed isothermal systems that are in contact and thermal equilibrium with an
external thermostat of a given temperature T.
The corresponding statistical ensemble is referred to as the canonical
ensemble, and the distribution function Gibbs canonical distribution.
H ( p, q, a) E constant
w( p, q) E H ( p, q, a) E, a
( E , a ) E H ( p, q, a) dpdq
( p ,q )
Within the microcanonical ensemble, all the energetically allowed microstates have
an equal probability to occur.
Microcanonical Distribution: Integral Over States
The normalization factor is given by
( E, a)
( E, a)
E a Phase volume
where is the integral over states,
or phase integral:
1
(2 ) f N N ! H ( p ,q,a ) E
( E , a ) dpdq
1
E H (p, q) dp1...dp N dq1...dq N
(2 ) f N N ! (p,q )
0, x 0
( x) , - Planck's constant, j - number of DOF per particle
1, x 0
(E,a) represents the normalized phase volume, enclosed within the
hypersurface of given energy determined by the equation H(X,a) = E.
Phase integral is a dimensionless quantity.
Thus the normalization factor shows the rate at which the phase volume
varies due to a change of total energy at fixed external parameters.
Microcanonical Distribution: Integral Over States
1
S 1
S k ln , T , P
E ( E,V ) V E
(These are the major results in terms of practical calculations over microcanonical ensembles.)
Microcanonical Ensemble: Illustrative Examples
We will use the Hamiltonian equations of motion to get the phase space trajectory,
and then evaluate the phase integral.
Harmonic Oscillator: Hamiltonian
Hamiltonian: general form H T ( p) U ( x)
p2 k m
Kinetic energy T
2m
x
k x2
Potential energy
U
2
Potential energy is a quadratic function of the coordinate (displacement form the
equilibrium position)
p2 k x2 H H p
H ( x, p ) x , p x , p k x
2m 2 p x m
x(0) 0.2
E1 x(0) 0
E2 x(0) 0.4
x(0) 0
E3
x(0) 0.6
x(0) 0
A1
A2
A3
1 0.95 1012
2 3.80 1012
3 8.54 1012
S1 3.811022 J/K
S2 4.00 1022 J/K
S3 4.111022 J/K
1
S
Temperature: T
E
We perturb the initial conditions (on 0.1% or less) and compute new values E and S.
The temperature is computed then, as
1
S S 2 kin
T (benchmark: T E )
E E k
T1 36.3 K
T2 145.1 K
T3 326.5 K
Pendulum: Total Energy
p2
Total energy: E H ( , p) 2
mgl cos Const (at any (t ), p(t ))
2ml
(0) 0.3
(0) 0
E1
E2 (0) 1.8
(0) 0
E3
(0) 3.12
(0) 0
p l - angular momentum
A1
A2
A3
1 0.4 1012
2 11.2 1012
3 21.4 1012
1
S
Temperature: T
E
We perturb the initial conditions (on 0.1% or less) and compute new values E and S.
The temperature is computed then, as
1
S S 2 kin
T (benchmark: T E )
E E k
T1 6.3 K
T2 153.8 K
T3 96.0 K
Summary of the Statistical Method: Microcanonical Distribution
4. Set up initial conditions and solve for the deterministic trajectories (MD).
5. Compute two values of the total energy and the phase integral for the
original and perturbed initial conditions.
One important preliminary issue related to the use of Gibbs canonical distribution is
the additivity of the Hamiltonian of a mechanical system.
Here, kinetic energy and the one-body potential are additive, i.e. they can be
expanded into the components, each corresponding to one particle in the system:
E kin (p1 , p 2 ,..., p N ) E kin (p1 ) E kin (p 2 ) ...
W (r ) W (r ) W (r ) ...
i
1 i 1 1 1 2
Two-body and higher order potentials are non-additive (function Q2 does not exist),
W2 W3 ... E kin W1
the system is described by an additive Hamiltonian,
pi2
H hi , hi W1 (ri )
i 2mi
Here, H is the total Hamiltonian, and hi is the one-particle Hamiltonian.
For the statistical description, it is sufficient that this requirement holds for the averaged
quantities only.
The multi-body components, W>1, cannot be completely excluded from the physical
consideration, as they are responsible for heat transfer and establishing the thermodynamic
equilibrium between constitutive parts of the total system.
A micromodel with small averaged contributions to the total energy due to particle-particle
interactions is called the ideal gas.
3 particles: 5 particles:
W2 0.049 E tot W2 0.026 E tot
Canonical Distribution
Suppose that system under investigation 1 is in
thermal contact and thermal equilibrium with a
much larger system 2 that serve as the
thermostat, or heat bath at the temperature T. Thermostat T
From the microscopic point of view, both 1 and
2 are mechanical systems whose states are
described by the phase vectors (sets of canonical
variables X1 and X2). The entire system 1+2 is N1
adiabatically isolated, and therefore the
microcanonical distribution is applicable to 1+2,
1
1
w( p1 , q1; p2 , q2 ) E H ( p1 , q1; p2 , q2 )
( E ) N2
Assume N1 and N2 are number of particles in 1
and 2 respectively. Provided that N1 << N2, the
2
Gibbs canonical distribution applies to 1:
H ( p ,q )
1
w( p, q) e kT
( p, q) ( p1 , q1 )
Z
Canonical Distribution: Partition Function
Thermostat T
The normalization factor Z for the canonical
distribution called the integral over states or N1
partition function is computed as
1
H ( p ,q ,a )
1
Z
(2 ) N ! ( p ,q )
3N
e kT
dpdq
2
N2
H ( p ,r )
1
(2 ) N ! (p,r )
3N
e kT
dp1...dp N dr1...drN
Before the normalization, this integral represents the statistically averaged phase
volume occupied by the canonical ensemble.
The total energy for the canonical ensemble is not fixed, and, in principle, it may
occur arbitrary in the range from to (for the infinitely large thermostat,
N2 ).
Partition Function and Thermodynamic Properties
Free energy: F (T ,V ) kT ln Z
(relates to mechanical work)
F
Entropy S k ln Z T ln Z
(variety of microstates) T V T
F
Pressure P kT ln Z
V T V
Internal energy U F TS kT 2 ln Z
T
These are the major results in terms of practical calculations over canonical ensembles.
Class exercise: check the last three above formulas with the the method of thermodynamic potentials,
using the first formula for the free energy.
Free Energy and Isothermal Processes
Free energy, also Helmholtz potential is of importance for the description of
isothermal processes. It is defined as the difference between internal energy and
the product of temperature and entropy.
F U TS
Physical content of free energy: the change of the free energy dF of a system at
constant temperature, represents the work accomplished by, or over, the system.
Indeed,
dF dU TdS SdT dU TdS W
SdT W
dF W
Isothermal processes tend to a minimum of free energy, i.e. due to the definition,
simultaneously to a minimum of internal energy and maximum of entropy.
Canonical vs. Microcanonical: Factorization of the Partition Function
In terms of practical calculations, there exists one major
Thermostat T
difference between the canonical and microcanonical 6
distributions: N1
1
For additive Hamiltonians, the canonical distribution N2
factorizes,
2
1 i
hi ( pi , qi )
H ( p ,q )
1
w( p, q) e e
kT kT
Z Z
N hi ( pi , qi )
1
w( p, q) e kT zi is the one-particle partition function
i 1 zi
Analytical results for this system are useful, because they provide acceptable first
guess assessments for a wide class of systems.
One-particle partition function:
pi2
(2 )3 0 2
0
V
(2 mkT ) 3/ 2
(2 )3
N ! (2 )3 N N !
Non-Interactive Ideal Gas: Thermodynamic Parameters
VN
Partition function: Z (2 mkT ) 3N / 2
(2 )3 N N !
Free energy N (2 )3
(recall the method of F (T ,V ) kT ln Z NkT ln 3/ 2
NkT
thermodynamic potentials) V (2 mkT )
F V (2 mkT )3/ 2 5
Entropy S Nk ln Nk
T V N (2 )
3
2
F NkT
Pressure P
V T V
3
Total internal energy E F TS NkT
(differs form the earlier 2
MD definition)
f
(generally: E NkT ,
2
f - number of DOF per atom)
Numerical Example: Interactive Gas
Repulsive interaction between the particles and the
wall is described by the wall function, a one-body
potential that depends on ri distance between the
particle i and the chambers center): y
( R ri ) R
Wwl (ri ) e ri
rij x
Interaction between particles is modeled with the
two-body Lennard-Jones potential (rij distance rj
between particles i and j):
12 6
WLJ (rij ) 4 12 6
r rij
ij
The Hamiltonian:
pi2
H (p, r ) Wwl (| ri |) WLJ (| ri r j |),
i 2 m i i j i
One particle is initially at rest. This illustrates the concept of heat exchange between the
smaller subsystem, for which the canonical distribution holds, and the external thermostat.
Interactive Gas: Equations of Motion and Solution
The total potential:
U Wwl (r1 ) Wwl (r2 ) Wwl (r3 ) Wwl (r4 ) Wwl (r5 )
WLJ (r12 ) WLJ (r13 ) WLJ ( r14 ) WLJ ( r15 )
WLJ (r23 ) WLJ (r24 ) WLJ ( r25 )
WLJ (r35 ) WLJ (r35 )
WLJ (r45 )
Equations of motion:
U U
mi xi , mi yi , i 1, 2,3, 4,5
xi yi
Parameters:
m 1023 kg, 1023 J, 2.7m,
1023 J, 4m 1 , R 10m
Initial conditions (nm, nm/s):
x1 (0) 0, x1 (0) 25, y1 (0) 5.8, y1 (0) 0
x2 (0) 0, x2 (0) 30, y2 (0) 3.1, y2 (0) 0
x3 (0) 0, x3 (0) 0, y3 (0) 0.1, y3 (0) 0
x4 (0) 0, x4 (0) 24, y4 (0) 2.9, y4 (0) 0
x5 (0) 0, x5 (0) 22, y5 (0) 5.7, y5 (0) 0
Interactive Gas: Temperature
Hamiltonian H hi ,
(can be viewed as additive in i 5 particles:
the statistical sense, due to pi2 W2 0.026 E tot
smallness of the time averaged hi e ( R |ri |)
2mi
pair-wise interaction)
p2 / 2 m e ( R |r|)
1
One-particle partition
z
(2 ) 2
( p ,r )
e kT dpdr
function
p2 p2 , r r,
(value at given dpdr pdpd p rdrd r polar coordinates
T and V = R2)
p 2 / 2 m e ( Rr )
4 2
(2 ) 2 0
e 2 mkT pdp rdr 1.383 1026
Z Z 21
Entropy: S k ln Z T ln Z k ln Z T 4.16 10 J/K
T T
Z Z
Pressure: P kT ln Z kT 7.211023 Pa
T T
Z Z
Internal energy: U kT 2 ln Z kT 2 1.27 1020 J
T T
f
Ideal gas benchmark: U (i.g.) NkT 1.22 1020 J f 2, N 5
2
Other parameters can be computed using the method of thermodynamic potentials
Phase Integral, Free Energy and Entropy vs. Temperature
Assume that we observe the same isothermal system at various
temperatures of thermostat. The following trends are available:
Partition function, and therefore the phase volume occupied by this canonical ensemble, grows
exponentially vs. temperature.
Free energy decreases linearly; the work done by the system does not depend on temperature.
Entropy decays vs. temperature. Physical implication (according to the second law):
temperature cannot grow spontaneously in an isothermal system, once thermal equilibrium
with the thermostat is established.
Specifics of Calculations for Liquids and Solids
In liquids, the energy due to pair-wise interaction between
H
particles is close to the kinetic energy (per particle).
However, interaction, wij, between separate constitutive
parts (subdomains) i and j is still weak, if compared with i j
the total kinetic energy of the smaller domain. Indeed, the hi
kinetic energy depends on the subdomain volume, while wij
depends on the surface area.
4. Set up initial conditions and solve for the deterministic trajectories (MD).
7. Compute the free energy and other thermodynamic parameters, using the
method of thermodynamic potentials. If possible compare the obtained value
of internal energy with a benchmark value.
Berendsen thermostat
Adelman-Doll thermostatting GLE
Phonon heat bath
Time-history kernel and transform techniques
Random force and lattice normal modes
Finite Temperatures
mi ui (t ) iU u i (t ) R i (t )
R i (t ) R j (t ) 2mi i k BT0 ( ) ij
T / T0
t
Finite Temperatures
t
ui (t ) 02ui (t ) (t ) ui ( )d R i (t )
0
R i (t ) R j (t ) k BT0 ( ) ij
atom next to
the interface
mui (t ) iU (t ) ui ( ) R i ( ) d k R j (t )
t
t
Time History Kernel (THK)
The time history kernel shows the dependence of dynamics in two adjacent cells.
Any time history kernel is related to the response function.
f(t) u1 (t ) u0 (t ) , ?
-2 -1 0 1 2 fn (t ) n,0f (t )
t
u n (t ) g n n ' (t ) f ( )d , U n ( s) G n ( s) F( s), U1 ( s) G1 ( s)G 01 ( s) U 0 ( s)
0
t
u1 (t ) (t )u 0 ( )d , (t ) L 1 G1 ( s)G 01 ( s)
0
0.6
2
(t ) J 2 (2t )
0.4
1 2 2 t
(t ) L 1 s 2 4 s J 2 (2t )
4 t 0.2
-0.2
0 2 4 6 8 10 12 14
Bridging Scale at T = 0: Impedance Boundary Conditions
The MD domain is too large to solve, so that we eliminate the MD degrees of freedom
outside the localized domain of interest.
Collective atomic behavior of in the bulk material is represented by an impedance force
applied at the formal MD/continuum interface:
Md N T f (u)
t
MD degrees of freedom outside the
localized domain are solved implicitly
M Aq f (u) (t ) q( ) u( ) d
0 FE + Reduced MD +
Impedance BC
MD FE
Due to atomistic nature of the model, the structural impedance is evaluated computed at
the molecular scale.
Dynamic Response Function: 1D Illustration
Assume first neighbor interaction only:
n-2 n-1 n n+1 n+2
n 1
2U
Mu n (t ) K
n ' n 1
nn ' un ' (t ) n ,0 (t ), K 0 2k , K 1 k K nn '
un un ' u 0
s M k (e
2n
1 1 ip ip 1 1 1
g n (t ) L F 2
2e L s 4s
2
2 n
( M k 1)
2 s s 4
2
Displacements Velocities
1
0.7
0.8
0.6
0.5
0.6
g n (t ) J 2 n (2t )
0.4 0.4
0.3 0.2
t
0.2
g n (t ) J 2 n (2 )d 0
0.1 -0.2
0
0 -0.4
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Illustration
Transfer of a unit pulse
due to collision (movie):
Discrete Fourier Transform (DFT)
Discrete convolution
K un K n n ' un ' F K un K ( p ) u ( p )
n'
Numerical Laplace Transform Inversion
Most numerical algorithms for the Laplace transform inversion utilize series
decompositions of the sought originals f(t) in terms of functions whose Laplace
transform is tabulated. The expansion coefficients are found numerically from F(s).
Examples:
Weeks algorithm (J
Assoc Comp Machinery 13, 1966, p.419)
S
f t e c T / 2t
a L t / T
0
L (t ) Laguerre polynomials,
a coefficients computed using F(s)