Anda di halaman 1dari 6

09Mar2009

Chemistry 21b Spectroscopy & Statistical Thermodynamics


Lecture # 26 Vibrational Partition Functions of Diatomic Polyatomic Molecules
Our starting point is again the approximation that we can treat the molecular degrees
of freedom independently, or
H

Htrans + Hrot + Hvib + Helec ,

(26.1)

to yield an overall wavefunction that is the product of the solutions to the individual terms,
or = trans (rot nucl )vib elec . Here we concentrate on the vibrational degrees of
freedom, first for diatomic molecules and then for polyatomic systems (and finally solids,
treated as a very large molecule). For simplicity, we will use a harmonic oscillator model
throughout and ignore the effects of anharmonicity.
The Diatomic Harmonic Oscillator
As we have seen before, the vibrational energy levels of a diatomic molecule are simply
vib = (v + 21 )hvib , v=0,1,2,... for a harmonic potential; where the vibrational frequency
p
is vib = k//(2), k being the force constant (or second derivative of the interatomic
potential) and the reduced mass. Each vibrational state is non-degenerate, and the
zero-point energy is h/2. Again, the reduced temperature for the system can be defined,
and here is T /vib , where vib = hvib /k. For the rotational degrees of freedom, even very
light molecules possessed a B of at most a few 10s of Kelvin. Here, the values of vib run
from a high of 6000 K for H2 out to closer to 300 K for I2 . Thus, even for heavy diatomic
species with weak bonds, T /vib is close to unity while T /vib  1 for typical diatomic
composed of first row atoms.
Furthermore, the zero-point energy, while real, cannot be accessed by the molecules in
question when responding to changes in physical conditions. It must, however, be included
when considering state functions such as the vibrational energy (and would become the
U (0) we have discussed previously). Including the zero-point energy, qvib becomes


X
qvib (T ) =
evib (v+1/2)/T = evib /2T 1 + evib /T + e2vib /T +
(26.2)
v

For translation and rotation we typically invoke the classical limit approximation and
replace the summation with an integral. Here, the terms in the exponent are often <1,
and so the series converges rather rapidly and it is better to simply to evaluate the sum.
Since we will ultimately be concerned with functions such as ln(Qvib ), the zero-point energy
multiplier in front becomes a negligible additive term for many species since ln(1) = 0.
The geometric series in the parenthesis of eq. (26.2) is a fairly simple one to examine
as v (see Apostol, Vol. I, pp. 388-389), and the closed form expression for q vib of a
diatomic molecule under these conditions and ignoring zero-point energy is
qvib (T ) =

1
.
1 evib /T
252

(26.3)

As T /vib 0, qvib 1; and as T /vib , qvib T /vib . As noted above,


for most molecules containing first row atoms or hydrogen, T /vib  1 near room
temperature, and the vibrational partition function is close to unity. This is some two
orders of magnitude smaller than qrot , and as we shall see helps explain the typically small
vibrational contribution to state functions such as the entropy of a perfect gas.
Polyatomic Systems and Normal Modes
From Lecture #10, we know that for an N atom molecule there are 3N 6 vibrational
degrees of freedom if it is non-linear and 3N 5 vibrational degrees of freedom if it is linear.
In either case, the vibrational modes can be treated as independent harmonic oscillators,
and so the full microcanonical vibrational partition function is simply the product of the
partition functions associated with each normal mode, or for a non-linear molecule
qvib (T )

3N
6
Y

qiN M (T ) ,

(26.4)

i=1

where the index i runs over the normal modes. Recall that certain modes can have
degeneracies larger than one, and these must be included in the appropriate qiN M .
Unlike diatomic molecules, the vibrational frequencies in polyatomic systems can vary
over a wide range. In CO2 , for example, the three distinct vibrational frequencies span
the range from 1000-3400 K (for vib ). Vibrations that are in some sense local, or
characteristic of various functional groups, often have frequencies of 3000-700 cm 1 , and
so yield individual terms in eq. (26.4) that are close to unity. As molecules become
larger, however, the product of many terms of, say, 1.1, can lead to a sizeable q vib (think
about a protein, for example). Furthermore, in large systems there are often very soft
vibrational degrees of freedom that are attached to what are called torsional modes. Their
(vibrational frequencies)/k can become quite close to or even less than 300 K, and when
present they tend to dominate the vibrational partition function. If the barriers to these
torsions or oscillations are high, the vibrational approach outlined here can be used. If
they are extremely low, then a rotational approach can be used. If, however, the barriers
are intermediate, as is almost always the case, the effects on the partition function must
be numerically calculated (see Maczek for additional details).
Molecular/Vibrational Thermodynamic State Functions
The localization of vibrations within molecules means that, like rotations, these
degrees of freedom are distinguishable. Thus, the canonical partition function Qvib is
Qvib =

N
qvib

1
1 evib /T

N

(26.5)

for a diatomic perfect gas. For a non-linear polyatomic system it is


N
Qvib = qvib
= (q1 q2 q3N 6 )

253

For a diatomic molecule, the expression for the vibrational energy of a mole of
oscillators, neglecting the zero-point contribution, becomes


Rvib
ln Qvib
2
 .
=
(26.6)
Uvib,m = Evib = kT
T
evib /T 1
V

If zero-point is included, we need to add Avogadros number times the individual molecular
zero-point energy, or for a diatomic molecule
Rvib
N h
 + A vib .
2
evib /T 1

Uvib,m = Evib =

At high temperatures, and ignoring the zero-point energy, the diatomic vibrational
energy for one mole is Uvib,m = RT , as expected from the equipartition principle. For
a polyatomic molecule, we would predict RT per each normal mode. This, of course,
can become quite large in (bio)polymers. Near room temperature, however, many of the
vibrational modes will have characteristic vibrational frequencies that are difficult to excite,
and so the energy content will fall far short of that predicted by equipartition.
The heat capacity at constant volume is given by the temperature derivative of U vib ,
and so the zero-point energy (or reference energy, if you will) is unimportant, so that
CV,

vib,m

= R

vib
T

2

evib /T
evib /T 1

2 .

(26.7)

This very famous result is known as the Einstein equation, and well come back to it in
greater detail in the next section. For now, note that at as T this expression goes
to CV, vib,m R as predicted by equipartition, and that as T 0, CV, vib,m 0. This
latter result is perfectly understandable from the quantum chemical perspective of a two
level system (as the temperature becomes very low it is impossible to access the excited
vibrational state, and so the system cannot respond to a change in temperature), but was
quite perplexing to physical chemists at the turn of the last century.
For the entropy we use, as before,
Svib =
or

Svib,m
R

Uvib
+ k ln Qvib
T


vib /T
 ln 1 evib /T
evib /T 1

(26.8)
(26.9)

for diatomic molecules. Again, if vib /T  1, as is typical near room temperature,


Svib,m /R is close to zero, and the entropy of a perfect gas is dominated by the translational
and rotational degrees of freedom (as we saw numerically for F2 in Lecture #25).
Vibrations and Crystalline Solids
As we saw in Lecture #13, in a macroscopic solid the translational and rotational
degrees of freedom are unimportant, and vibrational effects come to the fore. To remind
254

ourselves of how to proceed in this case, consider the N atom variant of the 1D coupled
oscillator we examined in Lecture #9 (where N NA ) and generalized in Lecture #13
that is depicted in Figure 26.1. Clearly, the interatomic interactions cannot be neglected,
indeed they are central to the problem at hand. It is, nevertheless, possible to use the
weakly interacting machinery we have set up to consider the problem.

#1

#2

#N

Figure 26.1 A schematic of a one-dimensional array of coupled oscillators as a simple


model for solids, where N NA .
As with our treatment of normal modes in a polyatomic molecule, we begin by
expanding the potential energy surface about its equilibrium configuration. The first
derivatives are zero since we are at a potential minimum, and we truncate the expansion
at the quadratic term. Defining the potential energy at the minimum as U (0) (where the
bold face stresses that this is a multi-dimensional system), the second derivatives of the
potential with the force constants kij , and the displacement coordinates as i , we have
1X
U (1 , 2 , ...., N ) U (0) +
kij i j .
(26.10)
2 i,j
From such an analysis we know that there would be 3N - 6 3N vibrational frequencies
(since N NA , the approximation here is an extremely good one!), and that the analog
of eq. (26.2) for the crystalline vibrational partition function becomes
Qvib (T ) = e

U (0)/kT

3N
Y

qvib,j ,

j=1

where qvib,j is the harmonic oscillator partition function for each normal mode j.
In a reasonable sized molecule it is possible to enumerate the vibrational frequencies
and assign them to distinct modes. Here, since N is so large, the vibrational frequencies
blend into a continuum. For single atom crystals, these become the acoustic branch derived
in eq. (13.7). Here, we introduce a spectral function g()d which denotes the number of
normal mode frequencies between and + d. The function g() is normalized such that
Z
g()d = 3N .
(26.11)
0

Inserting this into the partition function, including the zero point energy, and the
expressions for the vibrational energy Evib and CV for N = NA yields
#
Z "
heh/kT
h
 +
Evib = U (0) +
g()d
(26.12)
2
1 eh/kT
0
255

and
CV

(h/kT )2 eh/kT g()d


.
2
1 eh/kT

(26.13)

These expressions contain the relevant physics, the difficult part is to evaluate g().
We have encountered one of the most famous approximations that leads to the
relationship in eq. (26.7), the Einstein equation. This approximation says that all normal
frequencies are the same, that is
g() = 3N ( E ) ,
where E is called the Einstein frequency for a given crystal. Defining the Einstein
temperature as E = hE /k and inserting the approximation for g() into eq. (26.12)
yields, for a mole of oscillators,
CV,m (crystal) = 3R

E
T

2

eE /T
eE /T 1

2 ,

(26.14)

where the multiplicative factor of three compared to the diatomic arises because the atoms
in a realistic crystal are free to vibrate in three dimensions.
The Einstein relationship successfully predicts that the heat capacity should go to
zero at low temperature, and this was a major step forward in the early days of quantum
mechanics since the low temperature heat capacity behavior of metals was impossible
to understand from a classical perspective. However, the low temperature behavior is
predicted to behave as
CV,m (Einstein) 3R

E
T

2

eE /T ,

while the experimental observation is CV,m (exp.) T 3 as T 0. The exponential term


clearly dominates the low temperature character of the Einstein equation, and so the
discrepancy between theory and experiment is, in fact, quite large at low temperature.
The culprit is the extremely crude assumption made for the frequency dependence of
g() to derive the Einstein equation, namely that the vibrational behavior of crystalline
normal modes can be approximated by a single frequency. The Debye model briefly
outlined in eq. (13.9) is much more realistic model in that it examines the low frequency
behavior of crystals when they are treated as a continuous elastic body. This approach
weds the quantum nature of the vibrational partition function embodied in eq. (26.10)
with a mechanically realistic implementation of the crystalline vibrational spectrum as
modeled by the continuous function g().
Specifically, if the crystal is treated as a cube and the number of standing waves/unit
frequency interval is determined (this is the same calculation that is used to examine
the Rayleigh-Jeans limit of the blackbody equation), we can recast the density of states
expression derived in Lecture #13 to yield:
256

9N 2
3 d
D
= 0

g()d =

0 D
> D ,

(26.15)

where D is called the Debye frequency for the crystal, and leads in turn to the Debye
temperature, or D = hD /k. While the actual form of the acoustic branch frequencies
does not exactly follow this behavior, the lowest frequencies do and it is precisely these
modes that matter most at low temperature.
By inserting the Debye formula for g(), it is found that the molar heat capacity at
constant volume is given by (where x = h/kT )

3 Z D /T
x4 e x
T
CV,m (Debye) = 9R
(26.16)
2 dx .
D
(ex 1)
0

Figure 26.2 (Left) Data for the molar heat capacity of Al, Cu, and Pb as a function
of reduced temperature are plotted as symbols. The solid curve presents a Debye heat
capacity fit to the data. (Right) Experimental acoustic mode spectrum of Fe. Note that
below 3 THz the expected g 2 for the Debye model is met (3 THz=100 cm1 140 K).
Figure 26.2 presents a fit of the Debye equation to heat capacity data for metals. At
high temperatures, both the Einstein and Debye molar heat capacities approaches 3R, the
Dulong and Petit value motivated by the equipartition principle. At low temperatures,
however,
and the integral in eq. (26.16) becomes definite. Its value is
R 4 x xx ,
2
x e /(e 1) dx = 4 2 /15. Thus, the limiting form of the Debye formula becomes
0
3

12 4
T
CV,m (Debye, T 0)
R
,
(26.17)
5
D
which does correspond to the low temperature behavior of the heat capacity of monatomic
metals. Clearly, by plotting CV /3R versus T /D , CV s for all such systems can be shown
on one plot! Experimentally, the Debye temperatures of metals are found to be a few
hundred Kelvin (D,Pb = 88 K, D,Na = 150 K, D,Cu = 315 K, D,Fe = 420 K, and
D,Be = 1000 K, for example). The actual form of the acoustic spectrum for iron is shown
at right in Fig. 26.2. At sufficiently low temperature, only the longest wavelength acoustic
modes are excited, and the structure in the actual density of states becomes unimportant.
257

Anda mungkin juga menyukai