Density of states

There’s been some confusion about the density of states. First one finds the density of states
in k-space. Assume we have a system with periodic structure with lattice constant is L. Then
the quantum mechanical boundary conditions says that particle wav functions have periodic
boundary conditions

ψ(x + L) = ψ(x); , (1)

and since plane waves ψ(x) = eikx give

ψ(x + L) = eik(x+L) = eikx eikL (2)

ψ(x) = eikx (3)

we must have
2π
eikL = 1 ⇔ kL = n2π ⇔ k = n , n = any integer . (4)
L
From this we see that one k-state takes space 2π L
in k-space. Doing the same in 3 dimensions,
there is one k-state in a box of size ( 2π
L
) 3
.
If we are dealing with electrons, we can put one spin up and one spin down electron in each
k-state ⇒ factor g = 2. This g is a degeneracy factor that tells how many internal states (like
spin states) there are in one k-space state.
Usually we want to sum over states. In the thermodynamical limit L → ∞ the k space is
densily packed with k-space states, so it becomes very cumbersome to sum them one by one.
Instead we convert the sum to an integral and say that
1 state
In 3D, the density of states in k space is constant,
, (5)
2π 3
L

d k 3
meaning that in an infinitesimal volume d3 k there are 3 states. The sum over states of
( 2π
L )
some function F (k) in the thermodynamical limit is (volume V = L3 )

d3 k
X X Z
F (k) = g F (k) ≈ g
3 F (k) (6)
2π
states k L
π 2π
dkk 2
Z ∞ Z Z Z
gV
= g sin(θ) dφF (k) = 4π dkk 2 F (k) . (7)
2π 3 (2π)3


0 0 0
L

Warning: The k = 0 state of bosons is an exception at T < Tc ; that single state has macro-
scopic occupation, which the integral completely ignores (in the integral k 2 F (k) = 0 for k = 0
no matter how big F (k = 0) is). That single state must be dealt with separately.

1
 With the “density of states” one usually means the energy density of states, so we convert
k-space to E-space. This is simple for free particles (such as ideal Bose or Fermi gas), which
have
1 1
~2 k 2
 
2m 2 1 2m 2 1 − 1
E= ⇒k= E 2 ⇒ dk = E 2 dE . (8)
2m ~2 ~2 2

We need a simple change of variables, k → E,

Z "  12 # "  12 #2
2m 1 2m
Z
X gV gV 1 1
F (E) = dkk 2 F (E) = 2 E − 2 dE E 2 F (E) (9)
states
2π 2 2π ~2 2 ~2
 3/2 Z
gV 2m 1
= dEE 2 f (E) . (10)
4π 2 ~2

This must be the same as

X Z
F (E) = dEf (E)F (E) , (11)
states

where f (E) is the (energy) density of states. Comparing the integrals we see that
 3/2
gV 2m 1
f (E) = 2 E 2 (in 3D) . (12)
4π ~2