Sunteți pe pagina 1din 18

Statistical Thermodynamics: Molecules to Machines

Venkat Viswanathan
May 5, 2015

Module 4: Electrons, Phonons and Photons


Learning Objectives:
Analyze the statistical thermodynamics of Bose and Fermi particles
Demonstrate consistency with our analysis for the ideal gas of Bose
particles (ideal gas of diatomic molecules)

Analyze the behavior of electrons in a metal

Discuss the thermodynamic behavior of a crystalline solid.

Find the role of lattice vibrations, resulting in the definition of


quasi- particles called phonons.

Compare two models for lattice vibrations

Look at photons in the context of Bose-Einstein statistics

Derive Plancks law of radiation and use density of states to derive


thermodynamic properties of a photon gas.

Key Concepts:
Bose and Fermi statistics, bosons and fermions, electron gas, Fermi
energy, Fermi momentum, Fermi temperature, electron pressure,
elec- tron heat capacity, crystal lattice, lattice vibrations, phonons,
Einstein model, Debye model, Black Body radiation, statistical
mechanics of pho- tons, Bose occupation, Plancks law.

statistical thermodynamics: molecules to machines

Non-interacting particles obeying Bose and Fermi


statistics
Consider a system of non-interacting, indistinguishable particles that
can have energies ( = 1, 2, ...) associated with their quantum mechanical states. The state of the system can be specified by the number
of particles at each energy level, i.e. n is the number of particles at
energy state .
.
Total number of particles is n = N .
.
Total system energy is n = E.
In the canonical ensemble, we write the partition function:
Q(T , V , N ) =

n1,n2,...

.
N

.
ex
p

.
(1)

where we include the delta-function constraint on the summation in


.
order to fix n = N , and we define = 1/(kBT ).
The indistinguishability of the particles is properly accounted for in
this representation since any given set of n contributes a single term
without over-counting indistinguishable states. In the grand canonical
ensemble, we write the grand canonical partition function:

(T , V , ) =

Q(T , V , N ) exp(N )

N =0

=
=

N =0 n1,n2,...

ex
p

.
n +

n1,n2,...

.
n N

enn

The Landau potential is written as:


pV = kBT log = kBT
log

.
enn

(2)

If the particles are Bosons, there are no restrictions on the number


of particles in a given state n ( = 1, 2, ...), thus:
.n
.
1
e +
=
(3)

1 e +
n =0

If the particles are Fermions, any given state can only have either
n = 0 or n = 1 particles, thus:
.n
.
e +
= 1 + e +
n =0

(4)

Therefore, we generally write:


.
.

pV = kB T log 1 e +

(5)

where "-" is for Bosons, and "+" is for Fermions.


From this, we find the average number of particles:

(N) = .

log
.

e +
= (n)

1 e

(6)


where (n) is the average occupation number in the state.
In the ideal gas limit, is very large and negative. Noting that
+
e
1, we have:

(N) e + = eq and pV k B Teq = k B T ( N )

where we define the single-particle partition function q =


.
e
.
The grand canonical partition function is:
.
.

1 eN qN =
= epV /(kB T ) = exp e q =
Q(T , V , N )eN
N
!
N =0
N =0

thus Q =

1
N
!

qN for ideal gas (agrees with previous lecture).

Fermi-Dirac statistics for conducting electrons in a


metal
Electrons in a metal can be modeled as a gas of non-interacting
fermions:
Electrons in a metal are at high densities (many atoms per volume
with each contributing to the conducting electrons).
Since no two electrons (fermions) can exist in the same state, a high
density system fills many of the single particle energy levels
The lowest unoccupied state will have a kinetic energy much larger
than kBT , thus thermal excitations result in energetics with large
kinetic energy and comparatively negligible potential energy of
inter- action.
The large kinetic energy associated with these electrons results in
large conductivity of electrons in metals
Using the results for the thermodynamic behavior of noninteracting fermions, find the behavior of an ideal gas of electrons in a
metal. Elec- trons in a metal act as non-interacting particles with
quantized energies given by:
h2
2
2
2
.
.
En =
nx + ny + nz
translational modes
2
8mL

which are the energy levels


for a particle in a box for a particles with
mass m = 9.10938 1031kg.
The average number of electrons in the n state is:
1

(nn) =

= F (En)
(7)
.
.1
where we have defined the Fermi function F () = e() + 1 .
The total number of electrons is given by:

(N) = 2

e(En) + 1

(E)n
e
nx =1 ny =1 nz =1

+ 1= 2n =1 n
x

F (En)

(8)

=1 nz =1

where we include a factor of 2 since the electrons can exist in spin-up


and a spin-down states.
For sufficiently large V , the spectrum of translational wavemodes
is effectively a continuum. Therefore, we can convert this summation
to an integral over n, resulting in:

dnx
(N) = 2

0
dnz F (En)
dny

=2
=2

dkx L

1
L
3

dk
3 23

L
dkz F [(k)]

dky L

dk

dk F [(k)] =
2V
y
z

dkF [(k)]

(2)3

where we have used a coordinate change from n to k = n, and


L
the
2k 2
k
energy is now written as (k) = 2
.
m
Define the chemical potential at T = 0 to be (T = 0) = F (Fermi
Energy). To proceed, consider the form of the Fermi function at T = 0,
F () = 1 for F and F () = 0 for > F . Define the
Fermi
momentum pF according to F =
F
found to be:
2
2V 4

(N ) =

(2)3 3

kF =

p2
m

8 V
(2m)3/2
3

k2k2
F
2
m

. Thus, at T = 0, (N) is
.

3/
2

h3

F =

3
8

.2/3

h2

2m

A typical metal (Cu) has a mass density of 9g/cm3. Assuming each


atom donates a single electron to the conducting electron gas, this
den-

particles

pV = 2kB T
.
log 1 + e
.

sity has a Fermi temperature F = Fk 80, 000K. This verifies


B
that
the Fermi energy F is sufficiently large to make the ideal gas approximation valid at room temperature. At room temperature, only states
with energy very near F will be affected by thermal energy kBT .
The spread in the distribution is approximately 2kBT (Fig. 1).
The pressure is found using the relationship for Fermi

(9)

Figure 1: The Fermi function F () =

.1

e() + 1

Following a similar derivation


as before, we write

2V

dk log{1 + [( k)]}
pV =
3

(2)
e

.
4V (2m)2/3 d1/2

1 + .
log
=
e()
h3
0
k2

(10)

where we have used = 2k .


m ), we have 1 + e() e(F ) for
In the limit T 0(or
< F and 1 + e() 1 for > F .
Therefore, the pressure is written as:

3/2
d1/2(F )
pV = 4V (2m)
F
=
h3
0

16V
(2m)2/3
15h3

(11)

25/
F

This pressure at T = 0 is approximately 106atm. This large


pressure plays an important role in halting the collapse of a star
(white dwarf) because this enormous pressure offsets the gravitational
forces that oth- erwise drive the collapse.
The average energy (E) is found using:

e+

(E) = 2
1 + e

(12)

Following a similar derivation


as before, we write

2V
1
dk(

(E) =
e[( k)] +
(2)3 k)
1

4V (2m)2/3
1

d3/2
=
e[] + 1
h3
0
4V (2m)2/3 d3/2F ()
=

h3
k2

where we have used = 2k .


We apply integrationmby parts to this equation and use ( N) =
to get:
3(N )

where
dF

5F3/2

(2m)
V2/3

h3

3/2

5/2

dF

(E) =

(13)

.
= e
[e()+1]2
In the limit T 0, the function d becomes peaked near F ,
d

dF

and we can effectively expand the integrand near = F5/2


(Fig. 2).
3/2
Expand the integrand about = F to get 5/2
+ 5 (
2
F
2 F
F ) + 15 1/2
8 F ( F + ....
) dF
Since
()
is even about F in this limit, the odd-order terms will
d
integrate to zero, leaving only the even terms. Thus, the final form of

Figure
2:
the Fermi

The
derivative
d

dF ()
function

of
=

the average energy is going to be:


.

( E) = ( N ) F

. .2
A+ B T
+ ..
F
.

e()[e() + 1]

.
(14)

A precise calculation of this low-temperature expansion (outside


scope of this module) gives:
.
.
. .2
3
5 2 T
(15)
+ ..
(E ) = (N)F 1 +
F
5
12
.
Therefore, the heat capacity for a metal in the limit of small temperature is given by:
52

3
CV =

(N)F 12

T
F2

( N) k B

(16)

2
giving a linear temperature dependence in the small-T limit.
The limiting behavior in the small-T limit suggests that a plot of
CV /T approaches a constant as T 0. This proves to be the behavior
of the heat capacity for metals in the limit of small T . As temperature
increases, the fluctuations in the metal nuclei also contribute to the
heat capacity (Fig. 3).

Crystal line Solid


The atoms in a crystal are arranged in a regular array of points in
space with a variety of possible crystalline lattices (Fig. 4). At zero
temper- ature, the atomic coordinates are uniquely locked into spatial
positions that minimize the energy. At finite temperature, the atoms
fluctuate about the energy-minimum positions, leading to lattice
vibrations that govern the thermodynamic behavior (Fig. 5).

Figure 3: The heat capacity CV of


T i3SiC2 exhibits the temperature
dependence CV = 1T + 3T 3 + ... in
the low-temperature limit [Ho et al.,
1999]

Thermodynamic contribution of lattice vibrations


Consider N atoms with positions {r} = r1, r2, ..., rN in a crystalline
lattice. Define the potential energy V ({r}) that describes .the energy
.
for a given
system(0)
configuration
{(0)
r}. A minimum-energy configura(0)
tion {r (0) } = r , r , ..., r
satisfies the
V {r (0) } =
condition
1

ri V |0 = 0 for i = 1, 2, ..., N . The atomic


positions {r
regular crystalline lattice.

ri
(0)

} define the

Expand the potential energy about {r } = {r (0) }, such that:


V ({r}) V
.
1
2

.
. N .
.
. .
( )
{r (0) } +
r V |0 ri r0
+

i
i
N

i=1 j=1

.
.
.
(0)
. .
(0)
rj
ri rj V | : ri
+ ..
0
ri
rj
.
i=1

Figure 4: Crystal lattice structures

The linear term is zero (by definition), thus the energy is:
N

where si

sisj Kij
V V0 +
2 i=1 j=1 ,=x,y,z
.
.
V|.
(0)
= e ri r
and Kij =
1

rj

(17)

ri

The Matrix Kij can be diagonalized into normal modes (eigenvectors) with effective elastic constants (eigenvalues). Since there are
3N 6 3N degrees of freedom, there are 3N normal modes.
The potential energy is written as:
3N
1

V = V0 +

2
Kll

(18)

2 l=1
where l is the magnitude of the lth normal mode.
The total energy of the system is then written as:

3N

3N
E=
p + 1 K
2
+
V
2
l

2m
l

l=1

2 l=1

(19)

l l

where pl and m l are the effective momentum and mass of the lth normal
mode, respectively.
The total energy E is decomposed into normal modes with an
individual- mode energy in the form of a harmonic oscillator. These
normal modes are called phonons.
Phonons act as quasi-particles which means they are
distinguishable and independent, i.e. they dont interact between each
other.
The Hamiltonian (sum of
kinetic and potential energy) of the lth
p2l
phonon is given by: Hl = 2 + 1 K l 2. This harmonic oscillator
energy
m

results in the quantized energy:

.
El =

.
1 k
l
jl +

(20)

.
K l
2
.
where jl = 0, 1, 2..., and the phonon frequency is l
m l
=
The canonical partition function Q is given by:
.
.3N

V0
(j + 1 )k.

l
l=1 l 2
...
e
Q(T , V , N ) =

j1 =0 j2 =0
3N

= eV0

j3N =0

l=1 jl
=0
3N

= eV0

21
e(jl + )kl

1 kl

Figure 5: Schematic showing the


atoms in a crystal in their locked,
energy- minimum positions at T = 0
and the lattice vibrations that occur at
finite temperature

(21)
l=1

1 ekl
.
where we use the mathematical property n= z n

1
.
1z

This gives the Helmholtz free energy:


.
.
. 1
.
kl
kl
F = kBT log Q = V0 + kBT
log 1 e
2
l=1
+
The average energy is given by:
.
.
log Q
3N
kl 1

e
k
l
= V0
+ k
(E) =
kl
l
+
l=1 1 e
2

3N

The next step is to analyze two competing models for the phonon
frequencies l.

Einstein model
In the Einstein model, we assume there is a single characteristic frequency of the crystal, defined as the Einstein frequency E . The
average energy for this model is given by:
3N kE

(E) = V0 +

3N kE e kl

ekE

1
For this model, the heat capacity is given by:
.
.
CV =

(E)
T

.k

3Nk
=

E
T
kB

1e
kE
k

kE

kE

V ,N

where we define E =

.2

3NkB

BT

.2

E
T

1e

.2

k BT

B
In the limit T the
heat capacity CV 3NkB
(Dulong-Petit
Law). As we learned before, the equipartition theorem states that the
energy receives kBT per thermally active degree of freedom for a harmonic oscillator (quantized energy is linear in the quantum index). In

the limit T 0 the heat capacity CV 0. The heat capacity approaches zero exponentially in the small-T limit (Fig. 6).

Debye model
The Debye model treats the solid as an elastic material. Vibrational
modes in an elastic solid correspond to sound waves, thus the
frequencies satisfy = ck, where c is the speed of sound in the solid
and k = m/L, where m = 1, 2, ...

Figure 6: Heat capacity of a crystal


predicted by the Einstein model of lattice vibration

Convert the sum over normal modes to a sum over k using:

(...) =
(...) = dm 1 dm2 dm3 (...)
m1 m2 m3

= kc

.3

= 4 .

dk1

dk2

.3

kc

kc

dk3 (...)

dkk2 (...) =
4

kc

.3

d2 (...)

0
c3 0
where kc is a cutoff wavemode (to be determined), and D = Ckc is
called the Debye frequency.
A complete conversion will include one longitudinal mode with cl
and two transverse modes with ct . This gives:
. L.

3
. D
(...) = 4
d2 (...)
(22)
2
. 1

+
3
l

c
t
.

To find D we must enforce that l 1 = 3N , thus we have:


. L.

3
. D
1 = 4
d2 (...)
2
. 1

c3

cl

2
3

.3 .
1

c3 + c3
l

1/3

3 = 3N D

9N

.
= .

L c13 + 23
ct

For convenience, use D as a parameter, thus:

1 D
2
(...) = 9N
D 0 d (...)

(23)

D
Define the Debye temperature D k=
, which defines the
temper- ature scale for vibrational fluctuations.
To test this model, we
B
find the heat capacity:

CV = k B 2 . ( E )
=
.
9NkB

3
D

D /T

x4ex

dx

(1
ex)

In the limit T the heat capacity approaches:

/T
T 3 D
dxx2 = 3NkB
CV 9Nk B 3
D 0
which is expected (Dulong-Petit Law).
In the limit T 0 the heat capacity scales as:
3
3
Nk B T
CV 9Nk BT 3
dx x4e

x
2
3
x
(1 e )
D
D 0

Figure 7: Heat capacity of a crystal


predicted by the Einstein model and
Debye model of lattice vibration

The heat capacity predicted by the Einstein and the Debye models
are very similar. However, the low temperature of the heat capacity
of non-conducting solids matches the Debye model, i.e. CV NT 3
(Fig. 7).

Black Body Radiation


We are all familiar with the idea that hot objects emit radiation, a light
bulb, for example. In the hot wire filament, an electron, originally in
an excited state drops to a lower energy state and the energy difference
is
given off as a photon, (s = h = s2 s1). We are also familiar with the
absorption of radiation by surfaces. For example, clothes in the
summer absorb photons from the sun and heat up. Black clothes
absorb more radiation than lighter ones. This means, of course that
lighter colored clothes reflect a larger fraction of the light falling on
them.
A black body is defined as one which absorbs all the radiation incident upon it, i.e. a perfect absorber. It also emits the radiation
subsequently. If radiation is falling on a black body, its temperature
rises until it reaches equilibrium with the radiation. At equilibrium, it
re-emits as much radiation as it absorbs so there is no net gain in
energy and the temperature remains constant. In this case, the
surface is in equilibrium with the radiation and the temperature of the
surface must be the same as the temperature of the radiation.
To develop the idea of radiation temperature we construct an enclosure having walls which are perfect absorbers (see Fig. 8). Inside
the enclosure is radiation. Eventually this radiation reaches
equilibrium with the enclosure walls, equal amounts are emitted and
absorbed by the walls. Also, the amount of radiation travelling in each
direction be- comes equal and is uniform. In this case the radiation
may be regarded as a gas of photons in equilibrium having a uniform
temperature. The enclosure is then called an isothermal enclosure.
An enclosure of this type containing a small hole is itself a black
body. Any radiation passing through the hole will be absorbed. The
radiation emitted from the hole is characteristic of a black body at the
temperature of the photon gas. The properties of the emitted radiation is then independent of the materials of the wall provided they are
sufficiently absorbing that essentially all radiation entering the hole is
absorbed. This universal radiation is called Black Body Radiation.
An everyday example of a photon gas is the background radiation
in the universe. This photon gas is at a temperature of about 5 K. Thus
the earths surface, at a temperature of about 300 K, is not in
equilibrium with this gas. The earth is a net emitter of radiation
(excluding the sun) and this is why it is dark at night and why it is
coldest on clear nights when there is no cloud cover to increase the
reflection of the earths radiation back to the earth. A second example
is a Bessemer converter used in steel manufacture containing molten
steel. These vessels actually contain holes like the isothermal
enclosure of Fig. 8. The radiation emitted from the hole is used in
steel making to measure the temperature in the vessel, by means of an
optical pyrometer.

Figure 8: Isothermal Enclosure.

Statistical Thermodynamics
To derive Plancks radiation law directly from our statistical
mechanics, we note that number of photons in the gas is not fixed.
The photons are absorbed and re-emitted by the enclosure walls.
Since the photons are non-interacting it is by this absorption and reemission that equilib- rium is maintained in the gas. Since, also the
free energy F (T , V , N ) is constant in equilibrium (at constant T and
V ) while N varies it follows that F /N = 0, that is
.
F .
=0
= N .
.
V ,N

(24)

The photon gas is then a Bose gas with = 0 so that the canonical
partition function is given directly as
r

Q=

(1 exp(ss ))1

(25)

s=1

And the expected Bose occupation is


ns = (exp(ss) 1)1
(26)
Using s = h and the equation of density of states, we then obtain:
u() = 1
or

s()n ()g()

(27)

2
(28)
3
c (exp(h) 1)
Which is Plancks Radiation Law. We may also recover Wien and
Rayleigh-Jeans laws as limits of Plancks law,
(a) Long wavelength, hc
kT<< 1. Here
u() =

hc/

(29)

8
kT
s()g() 4
V

(30)

(exp(hc/kT) 1)
kT

And Eq. 28 becomes


u() = 1

c
Which is Wiens law valid at long wavelengths.
(b) Short wavelength, hc
kT>> 1. Here

s =

hc

exp(

hc/kT)

(31)

And Eq. 30 becomes


u() = 1
V

s()g()

8h exp(hc/kT)
c
5

(32)

Which is the Wiens law valid at short wavelength. Employing our


statistical mechanics we readily obtained Plancks radiation law. We
may also derive the Stefan-Boltzmann law for


h
8
2
(33)
u=
du()
=

c3

(exp(h) 1)

Introducing x = h, this reduces to

x3
8k4
4
= aT
dx

exp(x) 1 4
u=
T
(hc)3 0
Where

(34)

8k4 4
a=

(35)
(hc)3 15
In this way we obtain, using statistical mechanics, a law derived
previously using thermodynamics including all the numerical factors.
This gives Stefans constant in
1
E=
As

caT = T

4
2ck
4

=
10

(hc)3 15

erg

= 5.67

(36)

cm2

Sec K 4

(37)

A measurement of could then be used, for example, to determine


Plancks constant. Planck, in fact, determined h as the constant
needed in his radiation law to fit the observed spectral distribution
law. This
gave him the value h = 6.55 1027 erg.sec which compares with the
present value of h = 6.625 1027 erg.sec

Thermodynamic Properties
We may calculate all the thermodynamic properties of black body
radia- tion using statistical mechanics through the partition function Q,
where
F = kT log(Q)

(38)

And Q was derived as shown in Eq. 25. This is the basic method
of statistical thermodynamics. The aim is to reproduce all the thermodynamic properties with all factors and constants evaluated. This
gives

F = kT log . Y
r
r

F = kT

.
(1 exp(ss ))

s=1

log(1 exp(ss))1

s=1

F=
2kT

d
h3

log(1

(39)

exp(ss))

Where the ss are the single photon states and the factor of 2
arises from the two polarizations available to each photon. This can
be in- tegrated in a variety of ways. Perhaps the most direct is to
integrate over phase space (d = dV 4p2 dp) and write s = pc.
Introducing the dimensionless variable x = s = pc, the
Helmholtz free energy is:
. 8k4
.
1

d(x) log(1
VT
(40)
F=
4
3
3
exp(x))
(hc)3 0
The dimensionless integral here can be transformed into that
appear- ing in the constant of a of Eq. 35, by an integration by parts,
i.e.,

I=

d(x)3 log(1 exp(x))

..
= (x ) log(1 exp(x)). +
x3d[log(1 exp(x))] (41)
0
.
3

The first term vanishes because:


3
3
(a) limx
log(1 exp(x)) c exp(x) = 0
x3
limx 3x
(b) limx0 log(1 exp(x)) c
log(x) = 0
x
limx0 x
And

I=
15

dx

x3
=
exp(x) 1

(42)

Comparing Eq. 40 and Eq. 35, we get:


1

(43)
F = aV T
3
From F we may determine all other thermodynamic properties by
differentiation. For example, the entropy is
.
. .
F .
4
3
. = aV
(44)
S=
T
T
3
.
The internal energy is:
V
U = F + TS = avT 4

(45)

The pressure
is:

. .
F .
1 4
. = aT
p=
3
V .T
Finally, the Gibbs free energy is: s
1
G = F + pV =
T

aV

1
+

(46)

aV T

=0

(47)

This is zero as required G = N and the chemical potential = 0.


We may use these expressions to further
verify
.
. thermodynamic consisU .
tency, for example that CV = T S . = T .
T
.
.
V
In summary, we have obtainedVthe spectral
distribution from the
Bose occupation in much the same way as we obtained the MaxwellBoltzmann distribution for a classical gas. The only other required
ingredient was the density of states. We have also obtained all the
thermodynamic properties using the partition function Q.

References
J. C. Ho, H. H. Hamdeh, M. W. Barsoum, and T. El-Raghy. Low
temperature heat capacity of Ti3 SiC 2 . Journal of Applied Physics, 85
(11):7970, June 1999.

S-ar putea să vă placă și