Professional Documents
Culture Documents
http://ocw.mit.edu
Spring 2008
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
#
5.62 Spring 2008 Lecture 22 Page 1
The Einstein (quantum) model (all vibrational modes have the same frequency) gave
much better agreement with experiment than the Dulong-Petit (classical) model
(equipartition). But as T decreases, the Einstein model CV decreases too fast relative to
the experimentally observed (approximately T+3 dependence) behavior of CV. Perhaps it
would be more realistic to allow the vibrational frequencies to follow a plausible,
computationally convenient, but non-constant probability distribution, ρ(ν)
Debye Treatment
$ h# '
2
e!h# kT
=k*
#max
CDebye
d#+(#) & )
V 0 % kT ( (1 ! e!h# kT )2
For sound waves traveling in a three-dimensional solid ! ( ") # "2 [see Nonlecture
c
(
c 2
kx + ky2 + kz2 ) , where
1/2
derivation, below]. One way of seeing this is that ! = k~ =
2" 2"
i k~ •r~ 2"
k~ is the wave vector such that the wave is described by e and ! = . The density
|k|
!
of states with a given ν is the number of ways of choosing k~ with the corresponding
magnitude, ⎟ k~ ⎟.
Just as the degeneracy for a given speed state is proportional to c2 in the kinetic
theory of gases (as you'll see later in 5.62), the number of ways of picking k~ with
magnitude k = | k~ | is proportional to k2.
So ρ(ν) ∝ k2 ∝ ν2
This figure is supposed to show a spherical shell of radius |k| and thickness dk. The
!
number of k states with |k| between |k| and |k| + dk is proportional to the volume of this
shell, 4π|k|2.
Solution: Find the harmonic frequencies which satisfy the boundary condition that the
displacements are zero at the surface of a crystal of volume V. The wave equation for
this problem is very similar to the Schrödinger Equation for a particle in a 3D infinite
cubical well.
At t ≠ 0, the initial wave has moved in the +x direction by vst, where vs is the speed of
sound in this medium
Find the values of ν for which Φ(x,t) = 0, where x is at the surface of the crystal of
volume, V.
#2m$
Schrödinger Equation: ! 2 " ( x, y, z ) = " ( x, y, z ) .
!2
2m!nx ny nz "2 2
2 ( x
= n + ny2 + nz2 )
!2 a
where {!nx ny nz } satisfy the boundary condition that ψ(x,y,z) = 0 on all six surfaces of the
cube.
Now use the required values of {ε} to solve for allowed ν's:
d 2!(x,t) d 2 + % 2"#x (.
= - A cos ' $ 2"#t *0
dx 2 dx 2 , & vs )/
% 2"# (
2
= $' * !(x,t).
& vs )
Generalize to 3 dimensions
& 2$% )
2
# 2!" & ! 2 2
2
( = 2 ( n x + n y + nz ) .
2 2
%
$ vs ' a
So now we know how ν depends on the number of standing waves in each of the three
crystal directions. We want to know the density of vibrational modes as a function of
frequency, ρ(ν), but it is easier to derive the density of modes as a function of n, ρ(n),
where
This is the equation for a sphere. So the number of modes between n and n + dn is given
by the volume of one octant (nx, ny, nz, and n are all positive) of a spherical shell of radius
n and thickness dn
!(n)dn = 18 4"n 2 dn
We want ρ(ν)
dn
!(") = !(n) .
d"
dn
To find the value of the Jacobian, , for the transformation between n and ν as the
d!
independent variable,
2!" ! !
= n = ( nx2 + ny2 + nz2 )
1/2
vs a a
2a"
n=
vs
dn 2a
=
d" vs
dn 1 ( 2a 4!a 3 "2
#(") = #(n) = 4!n 2 ) =
d" 8 vs v3s
a 3 = V.
4#V 2
!(") = 3
"
v3s
This is the frequency distribution function that goes into the Debye model.
But Debye had one more trick up his sleeve before inputting ρ(ν) to a statistical
mechanical calculation of macroscopic thermodynamic properties.
There cannot be an infinite number of modes; only 3N–6 ≈ 3N. So Debye cut off the
mode distribution arbitrarily at νmax to give the correct number of modes.
ρ(ν) = Aν2 where A is determined by Debye's cutoff at νmax
!max !max 3
A!max 9N 9N
3N = " #(!)d! = A " !2 d! =
3
$ A = 3 $ # ( !) = 3 ! 2
!max !max
0 0
NOTE: We still don't know what νmax is, only that the mode distribution is normalized to
this parameter.
Cv = k " %(#)d#
(1 $ e$h#/ kT )
0 2
#max
9N 2 & h# )
2
e$h#/ kT
C =k "
Debye
# ( + d#
V
0 #max
3
' kT * (1 $ e$h#/ kT )2
h# h# kT
Debye Temp. , -D , max x = d# = dx
k kT h
!D T
9N 2 $ kT ' 2 e* x $ kT '
2
CDebye
V =k " #max
3
x & ) x
% h (
& )dx
(1 * e* x ) % h (
2
0
3 !D T
$ kT ' x 4 e* x
= 9Nk & )
% h#max (
" (1 * e* x )
2
dx
0
3 !D T
$T ' x 4 ex 9Nk
y
x 4 ex
= 9Nk & )
% !D (
" ( e *1)
x 2
dx + ++, 3
y
y=! D T
" ( e x *1)
2
dx
0 0
not a generally
tabulated function
Integrate by parts
u = x4
!1
v=
( e !1)
x
ex
du = 4x 3dx dv = dx
( e x !1)
2
# &
3# y4 &
y
9Nk % !x 4
y y
4x 3dx ( x 3dx
C debye
V = 3
y %$ e x !1 0
+ " ex !1 (' = 3Nk • y 3 %$4 " ex !1 ! ey !1('
0 0
# 3
y
x 3
3y &
= 3Nk • %4 • 3
$ y
" e x !1
dx ! (
e y !1'
0
" 3y %
CVDebye = 3Nk • $4D(y) ! y '
# e !1&
Debye Einstein
Function Function
high T limit
$ D h#max
y
CvDebye 12 x3 3y h#
3Nk
= 3
y
! e "1
x
dx " y
e "1
x=
kT
y=
T
=
kT
0
y
12 x3 3y
% 3
y
! (1+ x +… "1)
dx "
(1+ y +… "1)
T ! $ D & x, y ' 0
0
y
12 12 y 3
=
y3
! x 2 dx " 3 = • " 3= 4 " 3=1
y3 3
0
!CVDebye ""
# 3Nk for T ! $D
(agrees with classical and Einstein treatments)
low T limit
y
CvDebye 12 x3 3y
3Nk
= 3
y
! e "1
x
dx " y
e "1
T # 0 $ x, y # %
0
12 x3
!
%
& dx " 0
y30 e x "1
12 ( ' 4 +
= 3* -
y ) 15 ,
4 $T '
3
!C Debye
= 3Nk• " 4 & ) for T * 0 low T limit
5 % #D (
V
3N !6
h# / kT
(U ! E0 )vib = kT "
i=1 eh#/ kT !1
3N !6
(h# / kT ) e!h#/ kT
= kT " 1 ! e!h#/ kT
i=1
$
(h# / kT ) e!h#/ kT
= kT % 1 ! e!h#/ kT
&(#)d#
0
9N 2
!(") = "
"max
3
"max
9N 2 (h" / kT ) e!h"/ kT h"max
(U ! E0 )vib = kT # "max
3
"
1 ! e!h"/ kT
d" x = h" / kT ,$ D =
k
0
% kT (
3
$D T x ' * e! x
3
3 & h )
= 3NkT •
"max
3 # 1 ! e! x
dx
0
% T ( $ D T x 3 e! x
= 3NkT • 3' * # !x
dx
& $D ) 0 1 ! e
!D
y 3
y=
3 x
= 3NkT
y3
! e "1
x
dx = 3NkT • D(y)
T
y
x3
0
D(y) = 3y"3 # dx
0 e x "1
3N !6 #max
3 - x3 x 3 dx 0
y
= 3NkT • 3 / ln (1 ! e! x ) 0 ! $
y
2
y .3 0 3 ( e x !1) 1
Integrate by parts:
x3
u = ln (1 ! e! x ) v=
3
e! x dx
du = !x
dx = x dv = x 2 dx
1! e e !1
# 1 3
y
x 3dx & # 1 &
= 3NkT %ln (1 ! e! y ) !
$ 3 y3
" e !1'
x ( = 3NkT %ln (1 ! e ) ! D(y)(
$
!y
3 '
0
"4 %
= 3Nk $ D(y) ! ln(1 ! e! y )'
#3 &
NOTE: The Debye model does not fit the phonon mode distribution of actual solids
terribly well, but Cv is not too sensitive to these differences. It works well for insulating
crystals but fails badly for metals. What is special about metals? Also, fails near
melting point of solid because the harmonic approximation fails. Why? Large
displacements are necessarily anharmonic.