Chapter 2

The Fermi-Gas Model

The simplest quantum-mechanical systems are free particles. There is no spatial

variation in the potential and the relevant Hamiltonian is given merely by the kinetic
energy operator. Eigenstates are plane waves labeled by their eigenmomentum k
with eigenenergies εk = 2 k2 /(2m). An interaction-free, continuous many-fermion
system is called a Fermi gas. All plane-wave states are occupied up to a certain
Fermi energy ε F , or corresponding Fermi momentum kF . The density of each plane
wave is constant and so is the total density. This, in turn, means that interacting bulk
matter when treated at the level of a mean-field approximation (Hartree–Fock or
density functional approach, see Chaps. 5 and 6) still has plane waves as eigenstates
for the single particle wavefunctions. This allows to use the Fermi gas as an instruc-
tive and powerful lowest-order approximation to a great variety of many-fermion
systems. The most prominent example for a Fermi gas comes from bulk metals. An
example from a quite different regime is given by compact stars (neutron star, white
dwarf, see Sect. 1.1.8). A dilute gas is realized for large clouds of fermions in traps
(see Sect. 1.1.8).
Bound finite fermion systems differ in that the occupied states reside in the
regime of discrete spectra. Nonetheless, the Fermi gas is also a useful starting
point for examining volume properties and orders of magnitude. For sufficiently
large particle numbers, the spectrum becomes dense and can be approximated by
a continuum (as will be discussed in Sect. 2.1). The complementing view comes
from the spatial density distribution ρ(r). A homogeneous gas has, by definition,
a homogeneous density, ρ = constant. As an example, Fig. 1.3 shows nuclear
charge densities for a broad range of nuclei. All four cases indicate a region of
nearly constant density in the interior growing with increasing system size. This
suggests that from the density point of view a Fermi-gas limit can provide a rea-
sonable lowest-order approximation also for finite fermion systems (see also the
examples in Sects. 2.1.2.2, 2.1.2.3, and 2.4).
This reasoning is applicable at least to all systems which have a nearly constant
density in the interior. It holds for saturating systems, i.e., systems where the radius
grows as R ∝ N 1/3 (see e.g. Fig. 1.2) which applies to metallic nano-particles (metal
clusters), atomic nuclei, or droplets of liquid 3 He. There are other systems whose
radius does not behave ∝ N 1/3 but which, nonetheless, have constant density like
compact stars (their radius results from an interplay with gravitation, see Sect. 2.4),

J.A. Maruhn et al., Simple Models of Many-Fermion Systems, 45

DOI 10.1007/978-3-642-03839-6 2,  C Springer-Verlag Berlin Heidelberg 2010
46 2 The Fermi-Gas Model

electrons in quantum dots (the radius is a design property), and fermions in a trap
(the radius is tunable). All these systems can be well approximated by a Fermi gas.
For a quick quantitative overview, let us recall Table 1.2 which summarizes key
parameters for a variety of fermion systems, Wigner–Seitz radius rs , Fermi momen-
tum kF , and Fermi energy εF . The table demonstrates the wide variety of physical
dimensions for the systems of interest.
The Fermi gas also plays a crucial role in density functional theory. Many prac-
tical density functionals are derived by carefully computing the properties of the
electron gas and transferring that, so to say piecewise, to an energy-density func-
tional for inhomogeneous systems. That is the well-known local-density approxi-
mation (LDA), see Sect. 6.1.2 for the exchange functional and Sect. 6.5 about the
Thomas–Fermi approximation for the kinetic energy functional.
One often encounters the notion of a “Fermi liquid” [85]. The difference to a
Fermi gas consists in the role played by correlations. A Fermi gas applies if the
interaction can be accounted for at a mean-field level. One speaks of a Fermi liquid
if the two-body short-range interaction becomes dominant. Simple models are rare
in that regime. A profound theory of finite droplets of Fermi liquids goes far beyond
the scope of this book.

2.1 From Finite Box to Continuum States

2.1.1 Single-Fermion Wave Functions in General

The Fermi gas is a generic model for nearly homogeneous systems of independent
fermions. Let us address first the features of a finite system. Each fermion is associ-
ated with a wave function |ϕα
. It is a state in the Hilbert space of one-particle wave
functions. It is often used in coordinate-space representation |ϕα
−→ ϕα (r). This
still hides the fact that fermions necessarily have spin and that the wave function is
to carry a spinor part. To make it quite explicit we write it in detailed components

1 1
|ϕα
≡ ϕα (r, )χ 1 + ϕα (r, − )χ− 1 .
2 2 2 2

We assume here and in the following that the fermions we consider are spin 1/2
which correspond to the vast majority of physically relevant cases (electrons, nucle-
ons in particular . . .). The components ϕα (r, ν) compose a complex field in R3 .
The two ν components together define a spinor. It is often advantageous to com-
bine the coordinates into a super-vector x = (r, ν). The integration
and summa-
tion over the whole space is then abbreviated as ν d3r −→ dx. This can be
summarized as
2.1 From Finite Box to Continuum States 47

x = (r, ν), r ∈ R3 , ν ∈ {− 12 , 12 }, (2.1a)


|ϕα
= dx|x
ϕα (x) = d3r |r, ν
ϕα (r, ν), (2.1b)
ν
ϕα (x) = x|ϕα
, (2.1c)

ϕβ |ϕα
= dx ϕβ∗ (x)ϕα (x) = d3r ϕβ∗ (r, ν)ϕα (r, ν). (2.1d)
ν

A formally sound connection is established by the Hilbert space basis of eigenstates

|r, ν
= |x
of space coordinate r and spin ν. It depends on the particular application
whether the compact notation with x or the detailed form r, ν is better suited.
The N -fermion state built from those occupied single-particle states is a Slater
determinant as given in (1.3), for details see Appendix A.3. The notion of a determi-
nant becomes impracticable for continuum states. One needs to find a more robust
definition. That is achieved by employing the operators for creation or annihilation
of a fermion in the state ϕα , i.e., the âα† or âα , see Sect. 1.3.2 and Appendix A.4.

2.1.2 From Bound States to Plane Waves

2.1.2.1 1D Periodic Boundary Conditions – Plane Wave Limit
Here we discuss the handling of infinite systems by deriving their properties from
a continuum limit of the corresponding finite system. We do that for the simpler
case of one dimension (1D). We also ignore spin because it remains unchanged
throughout all considerations. It allows to demonstrate all essential steps with simple
sums and integrals. The generalization to realistic three dimensions (3D) is obvious
but tedious.
The homogeneous 1D system is first limited to a finite box 0 ≤ z ≤ L with
periodic boundary conditions. The normalized eigenstates thus become

1 2π
ϕn (z) = √ exp(ikn z), kn = n, n = 0, ±1, ±2, . . . . (2.2)
L L

 2 2
The energies are εn = 2m n . The ground state is obtained by filling the lowest
energies up to the given particle number N , which means |n| ≤ (N − 1)/2 (note that
N must be odd to produce a unique ground state). The states are orthonormalized

L L
dz  2π z 
ϕn |ϕn
= dz ϕn∗ (z)ϕn (z) = exp i(n − n ) = δnn . (2.3)
0 0 L L

The density becomes

nF
exp(ikn z) exp(−ikn z) N N −1
ρ0 = √ √ = , nF = , (2.4)
n=−n F L L L 2
48 2 The Fermi-Gas Model

as it should be (remember that in 1D L plays the role of the volume V in 3D). We

now
view the summation
 as a discrete approximation to a continuous integral, i.e.,
dk . . . ←→ n Δk . . . and identify Δk = kn+1 − kn = 2π/L. This suggests to
rewrite (2.4) as


nF
exp(ikn z) exp(−ikn z) 2π
ρ0 = Δk √ √ , Δk = kn+1 −kn =
n=−n F 2π 2π L
  

Δk . . . −→ dk . . .
n
  
kF
exp(ikz) exp(−ikz)
= dk √ √ , kF = Δk n F = 2πn F
L
−kF 2π 2π
kF
1 N
= dk = .
2π −kF L

The integral trivially becomes kF /π and allows to recover the appropriate

√ N /L in
the limit N  1, while N /L = constant. This shows that the factor 1/ 2π provides
the correct normalization of the continuum states, which are then

exp(ikz)
ϕk (z) = √ . (2.5)
2π

The continuum limit has an important consequence for the orthonormality relation.
Form (2.3) applies only to a discrete spectrum because the Kronecker-δ, i.e., the
δnm , is defined only for integer numbers, while the continuum limit produces the
continuous label k. The wave functions now have to be “normalized” to a Dirac
δ-distribution as
∞ ∞
† exp(i(k − k)z)
ϕk |ϕk
= dz ϕk (z)ϕk (z) = dz = δ(k − k). (2.6)
−∞ −∞ 2π

2.1.2.2 N Particles in a 1D Box – Limit for the Density ρ(z)

The periodic boundary conditions were particularly useful to explain the contin-
uum limit in terms of plane waves. Realistic situations start from bound states and
consider increasing system size. For simplicity, we stay in 1D and consider the
Schrödinger equation in a square-well potential

p̂ 2
Ĥ = + V (z),
2m
0 for 0≤z≤L N
V (z) = , L= . (2.7)
∞ for z < 0 or &z > L ρ0
2.1 From Finite Box to Continuum States 49

This potential produces bound states throughout, it is very well suited for the con-
tinuum limit because it has a constant value inside the infinite walls, so that it is a
useful model for saturating many-fermion systems which develop constant potential
bottoms in the interior for large system sizes (see the discussion at the beginning
of Chap. 3). If the system were composed of N classical particles, the average
density would be constant in space with ρ = N /L, which is also the density
(2.4) in case of periodic boundary conditions. We take that as a guideline for the
quantum-mechanical result. This is why we scale the length L = N /ρ0 to sim-
ulate a saturating system which has a strong tendency to a constant equilibrium
density ρ0 .
The eigenstates for the Hamiltonian (2.7) are [24]

2 π
ϕn (z) = sin(kn z), kn = nΔk, Δk = , (2.8)
L L

as can be seen from the boundary conditions ϕ(0) = ϕ(L) = 0. The total density of
such a many-fermion system is the incoherent sum of the densities of the occupied
single-particle states (see Appendix A.5.1). We use


N
N cos ((N +1)a) sin (N a)
sin2 (na) = −
n=1
2 2 sin(a)

to obtain


N
2 2
N
N cos ((N +1)Δkz) sin (N Δkz)
ρ(z) = |ϕα |2 = sin (nΔkz) = − ,
n=1
L n=1 L
 L sin(Δkz)
ρ0

which we finally write as

 
1 cos ((N +1)Δkz) sin (N Δkz)
ρ(z) = ρ0 1 − . (2.9)
N sin(Δkz)

The resulting densities for three system sizes N are illustrated in Fig. 2.1. The con-
vergence toward a homogeneous system with increasing N is clearly apparent. The
deviation are largest at the boundaries because the step down to ρ(0) = ρ(L) = 0
has to be enforced. In the interior, however, there is soon a smooth pattern.
One can read Fig. 2.1 the reverse way and point out the irrepressible spatial fluc-
tuations in the densities. This is a quantum-mechanical effect, often referred to as
“shell fluctuations” [37]. The probability distributions of bound-state wave functions
are necessarily inhomogeneous and excited states do have large fluctuations with
zeroes, the more the higher the excitation. These fluctuations persist in the total
densities, but in fact tend to compensate each other, the better the larger the particle
number. This is why we see the nice approach to the continuum limit with constant
50 2 The Fermi-Gas Model

1
density ρ [ρ0]

N=5 N=10 N=50

0
0 5 5 10 0 5 10 15 20 25 30 35 40 45 50
x x x

Fig. 2.1 Spatial density distributions for N -fermion states in the 1D box potential. The length of
the box scales as L ∝ N to maintain the same average density. The faint horizontal line indicates
the (classical and) continuum limit of equi-distribution

density. Consider, e.g., the nuclear charge densities in Fig. 1.3. Unlike the present
rough model of particles in a box, they show an extended surface zone, but in the
interior one can also spot small oscillations about a constant value, which are mainly
due to these shell effects. The patterns also do show up in the density of the cluster
Na+339 , see the right-hand part of Fig. 2.4.

2.1.2.3 2D Periodic Box – Limit For Eigenvalues

In the last step of illustrating the continuum limit, we are now going to discuss
the density of states and the associated level density (see Sect. 1.2.4.3). The start-
ing point is the Fermi sphere of occupied single-particle states in a homogeneous
fermion system. The Fermi “sphere” is literally true in a 3D system. For simplicity
of graphical representation, we here choose a 2D system which then will yield a
Fermi circle to describe the set of occupied states.
The homogeneous 2D system limited to a finite box 0 ≤ x, y ≤ L with periodic
boundary conditions in both directions has solutions quite similar to the 1D case
discussed in Sect. 2.1.2.1. They read

1
ϕn (x, y) = exp(ikn x x) exp(ikn y y), (2.10a)
L
2π 2π
kn = n= (n x , n y ), n x , n y = 0, ±1, ±2, . . . , (2.10b)
L L
2 k2 2 (k x2 + k 2y )
εn = = . (2.10c)
2m 2m

We consider again systematically increasing√the system size N while keeping the

average density constant, which means L = N /ρ0 for that 2D system. The result-
ing spectrum of quantum numbers (k x , k y ) is√illustrated in Fig. 2.2. The single-
particle energy (2.10c) is a function of k = |k| only. The occupied states then
reside within a circle whose radius is called the Fermi momentum kF , as indicated
in the figure. The selection of states does not so perfectly match the circle for small
system sizes (leftmost panel), but becomes a smoother distribution and approaches
2.2 Basics: Density, Fermi Momentum, Fermi Energy 51

kF
ky

kx kx kx

Fig. 2.2 The momenta (k x , k y ) of the eigenvalues for the 2D box with periodic boundary conditions
√
for different values of L (∝ N ), increasing from left to right

the circle quickly with increasing size (see Sect. 2.2.5 for the detailed computation
of the density of states in a 3D Fermi gas.)

2.2 Basics: Density, Fermi Momentum, Fermi Energy

2.2.1 Fermi Momentum, Fermi Energy, Density

Consider the ground state of independent fermions with the symmetries of free
space. “Independent” means that each particle can be associated with one single-
particle wave function ϕkσ (r) and “fermion” implies that each single-particle state
can be occupied only once. The “symmetries of free space” are invariance under
translation and rotation, called together homogeneity of space. Homogeneous sys-
tems are necessarily infinite. This causes no technical problems but conceptual sub-
tleties like dealing with continuum states. Translational and rotational symmetry
apply to the Hamiltonian ĥ which defines the single-particle wave functions, e.g.,
the self-consistent mean-field Hamiltonian as will be discussed in Chap. 5. For the
moment we need only the symmetry property
 of ĥ. The most general translationally
and rotationally invariant form is ĥ = n an (p2 )n with constant coefficients an . The
Fermi-gas model assumes the simplest form

p2
ĥ = . (2.11)
2m

In any case, plane waves, the eigenstates of the momentum operator p̂ = −i∇, will
also be eigenstates of ĥ. We have obtained them for the 1D case in Sect. 2.1.2.1. The
plane waves in 3D similarly read

exp (ik·r) 2 2
ĥϕkσ = εk ϕkσ ←→ ϕkσ (r, ν) = δσ ν , εk = k , (2.12)
(2π )3/2 2m
52 2 The Fermi-Gas Model

where σ = ±1/2 labels the spin and ν labels the components of the Pauli–Spinor.
This form applies to a gas from a single-fermion species like the electron gas or
neutron matter. Their degeneracy factor is two due to the two-spin orientations. That
feature carries through to most formulae in the sequel. Symmetric nuclear matter
would also have a dependence on an isospinor (which treats proton and neutron as
two states of the nucleon distinguished by isospin ± 12 ) and consequently a degen-
eracy factor four, for details see Sect. 2.4. Note that the energies depend only on
k = |k| which is a consequence of rotational invariance. The wave-vectors k ∈ R3
constitute a continuum of quantum numbers. The orthonormality relation is the 3D
generalization of (2.6), i.e.,

 †
ϕkσ |ϕk σ
= d3r ϕkσ ϕk σ = δ 3 (k − k )δσ σ . (2.13)
ν

A many-fermion system occupies a large number of these states. The many-body

ground state |Φ0
fills the lowest kinetic energies up to a Fermi momentum kF , i.e.,
 
|Φ0
≡ ϕkσ , k = |k| ≤ kF , σ = ± 12 . (2.14)

The filling of occupied states may equally well be expressed in terms of the Fermi
energy εF as

2 2
εk ≤ εF = k . (2.15)
2m F

Fermion creation and destruction operators, see Sect. 1.3.2 and Appendix A.4, allow
a clear and compact definition of the ground state |Φ0
as

†
âkσ |Φ0
= 0 for k ≤ kF
(2.16)
âkσ |Φ0
= 0 for k > kF

expressing formally that all states k ≤ kF are occupied.

The spatial density
 for independent-Fermion
 systems is generally (see Appendix
A.5.1) ρ(r) = α∈occup. ν |ϕα (r, ν)| . For the present continuum of states, we
2

have to replace the summation over discrete states by an integral over the continuum
states. Thus

 
∗
ρ(r) = d3 k ϕkσ (r, ν)ϕkσ (r, ν)
σ k≤kF ν

1  2 4π 3
= d3 k δσ ν exp (−ik·r) exp (ik·r) δσ ν = k ,
(2π )3 σ ν k≤kF    (2π )3 3 F
=1
2.2 Basics: Density, Fermi Momentum, Fermi Energy 53

which finally yields

kF3
ρ0 = ρ(r) = . (2.17)
3π 2

The density is constant, i.e., it also obeys the symmetries of free space. From another
aspect, the density is the number of particles per volume, ρ0 = N /V . Both particle
number and volume are infinite, but the density is finite. This provides a way to regu-
late the particle number in homogeneous systems, and (2.17) is often read in reverse.
The density ρ0 is given and determines the corresponding Fermi momentum as

 1/3 1/3
kF = 3π 2 ρ0  3.18 ρ0 . (2.18)

2/3
Consequently the Fermi energy behaves as εF ∝ ρ0 .
A further way to characterize the density of the system is the Wigner–Seitz radius
rs . It is the radius of the sphere whose volume fits just one particle, i.e.,

 1/3  1/3
1 4π 3 9π 1 1.92 3
= r =⇒ rs =  = . (2.19)
ρ0 3 s 4 kF kF 4πρ0

2.2.2 The One-Body Density Matrix

For finite N the one-body density matrix is defined as (see Appendix A.5)

(x, x ) = ϕα (x)ϕα† (x ).
α∈occup.

It contains the same amount of information as the independent-fermion state |Φ0

.
It is instructive to evaluate the one-body density matrix for the Fermi gas as a model
system. Identifying α ↔ (kσ ) this yields

 d3 k  
(rν, r ν ) = 3
δνσ exp ik·(r − r ) δν σ .
σ k≤kF (2π )

The spin overlaps are trivially evaluated. The integration of the wave-vectors is per-
formed in spherical coordinates k ≡ (k, θ, φ). We align the coordinate system along
the direction of r − r . Thus
54 2 The Fermi-Gas Model

 
δνν kF 1 2π
(rν, r ν ) = dk k 2
d(cos θ ) dφ exp ik |r − r | cos(θ )
(2π )3 0 −1 0   
y

δνν kF
exp (iky) − exp (−iky)
= dk k 2
(2π )2 0 iky

δνν kF
δνν k 3F sin(k F y) − k F y cos(k F y)
= dk k sin (ky) = ,
2π 2 2π 2 (k F y)3
0
  
j1 (k F y)/(k F y)

where j1 is the spherical Bessel function of first order [96]. The final result is

δνν 3 j1 (x)
(rν, r ν ) = ρ0 J (k F |r − r |), J (x) = . (2.20)
2 x

It is a rather simple universal function, depending only on |r − r |, which expresses

translational and rotational invariance. The shape of the function is shown in
Fig. 2.3. Note that it scales with kF−1 and so applies to all densities ρ0 . The larger ρ0 ,
the faster the decay in relative distance.
The spatial dependence can also be expressed in terms of the Wigner–Seitz radius
(2.19) as J (1.92|r − r |/rs ), which shows that exchange properties scale with rs .
The typical decay length of J is given by the first zero. That appears at a distance
of |r − r | ≈ 1.6 rs .

0.8

0.6
J(x)=3j1/x

0.4

0.2

0 5 10 15 20
kF|r-r’|

Fig. 2.3 The function J carrying the spatial dependence of the one-body density matrix (2.20) of
the homogeneous Fermi gas
2.2 Basics: Density, Fermi Momentum, Fermi Energy 55

2.2.3 A Model for the One-Body Density Matrix

in Finite Systems

The one-body density matrix is a rather powerful quantity carrying a large deal of
information about a system. For example, it allows to compute all one-body observ-
ables directly (Sect. 1.3.3). It thus provides a rather detailed view of the state of a
system and, correspondingly, a critical analyzing instrument for the quality of the
Fermi-gas approximation. As a test case, we consider the valence electron cloud of
the metal cluster Na+ 339 . The ionic background is simplified in terms of the jellium
approximation (1.1). A DFT calculation (see Sect. 6.1) is performed and yields the
one-body density matrix (r, r ) for the finite system, which is then the benchmark
result for the fully detailed description of the system.
The Fermi-gas model yields a compact expression of the one-body density matrix
with a very specific and pronounced dependence on the difference coordinate |r−r |,
see (2.20) and Fig. 2.3. The pattern in a finite system close to a Fermi gas should
look similar in that difference coordinate while the local density has to reproduce
the finite spatial distribution. That suggests a simple model for the one-body density
matrix

δνν  
(LDA) (rν, r ν ) = ρ(r) J k F (r)|r − r | , (2.21a)
2
 2 1/3
k F (r) = 3π ρ(r) , (2.21b)
1 
r= r + r , (2.21c)
2

which employs a local Fermi momentum k F (r) deduced from the local density
according to (2.17). It is called a Local Density Approximation (LDA) for the one-
body density matrix. (An application of LDA to the energy is at the heart of DFT as
we will see in Sect. 6.1.)
Figure 2.4 shows a comparison between the exact one-body density and the
LDA (2.21). We look at the pattern in the relative coordinate |r − r | for a series
of average positions r and normalize to the value at |r − r | = 0. There is nice
agreement between the exact pattern and the LDA for most reference radii. The
largest deviation is seen at the center of the cluster. A quick glance at the local
density distribution on the right side of Fig. 2.4 shows a large deviation from the
average local density just at the center. This is an effect of spatial shell fluctuations
as discussed in Sect. 2.1.2.2, which here happen to accumulate particularly at the
center. The mismatch of the LDA model in the lowest left panel of Fig. 2.4 is then
understandable considering that going along growing |r − r | one always runs into
a regime of higher densities than those at the reference point. On the other hand,
the center occupies only a very small fraction of the total volume and thus Fig. 2.4
indicates that LDA is fairly well justified on the average. Note also that the density
matrix falls off properly to zero outside the cluster.
56 2 The Fermi-Gas Model

1 exact
raver = 24 a0 Fermi gas 0.4
0.8

ρ(r,r’)/ρ(raver)

ρ(r) [100–1a0–3 ]
0.6 0.3
0.4
0.2 0.2 Na339+
0

1 exact 0.1
raver = 16 a0 Fermi gas
0.8
ρ(r,r’)/ρ(raver)

0
0.6 0 5 10 15 20 25 30 35
0.4
|r| [a0]
0.2
0

1 exact exact
raver = 8 a0 Fermi gas raver = 0 Fermi gas
0.8
ρ(r,r’)/ρ(raver)

0.6
0.4
0.2
0

0 5 10 15 20 25 0 5 10 15 20 25 30 35
|r-r’| [a0] |r-r’| [a0]

Fig. 2.4 Upper right panel: The local electron-density distribution of the cluster Na+ 339 . The ionic
background is described in the spherical jellium model (i.e. (1.1) with deformation δ = 0). The
electronic wave functions were computed with density functional theory (DFT) in local density
approximation, see Sect. 6.1. The vertical dashed lines indicate the reference radii for the plots
in the left and the lower right panels. Left and lower right panels: The one-body density matrices
(r, r ) are normalized to the local density at ρ(raver ) (where raver = 12 (r + r )) along the difference
coordinate |r − r | for a variety of average radii raver as indicated. The full line shows the result
from the DFT calculation of the finite system and the dashed line the density matrix in Fermi-gas
approximation (2.21)

2.2.4 The Two-Body Density

The two-body density matrix 2 for a Slater state can be expressed completely
through the one-body density matrix  as (see Appendix A.5)

2 (x1 , x2 ; x1 , x2 ) = (x1 ; x1 )(x2 ; x2 ) − (x1 ; x2 )(x2 ; x1 ).

We consider its diagonal element, the local two-body density ρ2 (x1 , x2 ) = 2 (x1 , x2 ;
x1 , x2 ) which in terms of the one-body density matrix becomes

ρ2 (x1 , x2 ) = (x1 ; x1 )(x2 ; x2 ) − (x1 ; x2 )(x2 ; x1 ). (2.22)

It represents the probability to find one particle at x1 and at the same time another
particle at x2 . It thus characterizes the spatial correlations between particle 1 and
particle 2 for the various spin relations between ν1 and ν2 . To evaluate the local
two-body density for the Fermi gas, we use the result (2.20) for the one-body density
matrix and very quickly obtain
2.2 Basics: Density, Fermi Momentum, Fermi Energy 57

Fig. 2.5 The two-body ρ20

density ρ2 (x1 , x2 ) as given in
(2.23) for the case of aligned 4
spin ν1 = ν2 as function of

ρ2(x1,x2) for ν1 = ν2
one-particle
kF |r1 − r2 |. The vertical volume
dashed line indicates the
radius of a sphere covering
one fermion

0
0 1 2 3 4 5 6 7
kF|r−r’|

ρ02   3 j1 (y)
ρ2 (x1 , x2 ) = 1 − δν1 ν2 J 2 (kF |r1 −r2 |) , J (y) = . (2.23)
4 y

The result sensitively depends on the spin of the two particles. The case that they
have the same spin is shown in Fig. 2.5. There is a deep hole around distance
zero which is the coordinate-space appearance of the Pauli principle: two identical
fermions cannot occupy the same position. For larger r , the function approaches
quickly the value (ρ0 /2)2 which means that the two particles become independent
(no more “Pauli correlated”). The excluded volume around r = 0 is often called the
exchange hole [27]. The vertical dashed line in Fig. 2.5 indicates the phase space
volume of a sphere covering one fermion. This one-particle equivalent radius cuts
almost precisely the half-value of the correlation function. This corresponds to the
fact that the “exchange hole” excludes precisely one fermion. The local two-body
density becomes (ρ0 /2)2 constantly at all r if the two particles have different spin.
For then they are distinguishable and thus independent everywhere.

2.2.5 The Kinetic Energy

The kinetic energy E kin can be computed with the same steps as were used for
the density in the previous subsection. The plane waves are eigenstates of p̂ 2 with
p̂ 2 ϕkσ = 2 k 2 ϕkσ . Thus we have just to modify the weight in the k-integration by a
factor 2 k 2 /(2m). This proceeds as
58 2 The Fermi-Gas Model

3  k
2 2
2
E kin = dr 3
d k
(2π )3 k≤kF 2m
2 kF π 2π
 2
= d3r dk k 4
d cos(θ ) dφ
2m (2π )3 0
    0
 0

kF5 /5 = dΩ=4π

 2
kF3 3 2  3 2 2
= d3r k = k ρ0 V,
2m 3π 2 5 F 2m 5 F
  
V

where V = d3r stands for the volume of the system. It is infinite and so is the
kinetic energy. Thus it is more appropriate to discuss the energy density, which is a
finite quantity. In fact, one usually prefers to specify the energy per particle which
is related to the energy per volume through the density ρ0 = N /V. This then yields
the finite result

E kin E kin V E kin 1 2 3 2 3

= = = k = εF . (2.24)
N V N V ρ0 2m 5 F 5

The kinetic energy per particle grows ∝ kF2 . The result as such is also quite interest-
ing as it gives a typical estimate of the kinetic energy of a particle in a Fermi gas
(3
F /5). This again confirms the key importance of the Fermi energy to provide a
relevant scale in simple fermion systems.
It is interesting to reformulate the kinetic energy in terms of system density using
relation (2.18). That yields

E kin 2 3 2/3
= (3π 2 )2/3 ρ0 .
N 2m 5

It explicitly shows the growth of the kinetic energy with system density. Let us
consider now a change the volume for fixed particle number which amounts to a
change of density. This defines a kinetic pressure

∂ E kin ∂ E kin 2 2 5/3

Pkin = = −ρ02 = (3π 2 )2/3 ρ0 . (2.25)
∂V ∂ρ0 N 2m 5

This is the Pauli pressure opposing compression of the system because it is easier to
accommodate the Pauli principle if the particles are farther apart.
2.3 Fermi Gas at Finite Temperature 59

2.2.6 The Level Density

In the computation of the kinetic energy the angular integration was trivial while
the crucial integral ran over the absolute value of the momentum k = |k|. For that
very common situation, it is instructive and advantageous to replace the momentum
integration by one over the energy. This substitution uses ε as
√
2 k 2 2 m dε
ε= =⇒ dε = k dk =⇒ dk = √ (2.26)
2m e me  ε

to yield the kinetic energy as

εF
2 2m 3/2 √ 3
E kin = d3r dΩ dε 3 ε ε = V εF ρ0 . (2.27)
(2π ) 3  5
      0
V 4π
  
εF
0 dεD(ε)

The volume element for energy integration can be expressed through one physical
quantity, the density of states D(ε), which then becomes useful for all expectation
values of functions of energy alone. Assume that we consider an observable f (ε)
which is a function of energy. The expectation value can then be written as
εF
f 2m 3/2 √
= dε D(ε) f (ε), D(ε) = ε. (2.28a)
V 0 3 π 3

The simplest examples are the particle number N for f = 1, and the kinetic energy
E kin for f = ε and given in (2.27). The density of states can be rewritten in several
different forms. Recall (2.17) connecting ρ with kF and (2.15) connecting kF with
εF . Combining these two equations yields
√ √
3N ε 3 ε 3 N √
D(ε) = = ρ0 , VD(ε) = ε. (2.28b)
4 V εF3/2 4 εF3/2 4 εF3/2

It is important to note that this expression for the density of states holds for a Fermi
gas in three dimensions. For a general dimension D, one has D ∝ ε D/2−1 .

2.3 Fermi Gas at Finite Temperature

One of the interests of a description in terms of level density is that temperature

effects can easily be included in the picture. To that end, one introduces occupa-
tions numbers n(ε) which regulate the probability to find a particle at single-particle
energy ε. Particle number and kinetic energy can then be written as
60 2 The Fermi-Gas Model
∞ ∞
N =V dε D(ε)n(ε), E =V dε D(ε)n(ε)ε. (2.29)
0 0

The occupation number n(ε) for a Fermi gas at zero temperature is just a step
function

n(ε) = ϑ(ε − εF ) = n T =0 (ε). (2.30)

Switching to a finite temperature is achieved through replacing the step function for
zero-temperature occupation numbers (2.30) by a smooth Fermi distribution

1
ϑ(ε − εF ) −→ n T (ε) = ε−μ(T )
, (2.31)
1+ kB T

where μ = μ(T ) is the chemical potential which has to be adjusted such that the
first integral in (2.29) reproduces the desired particle number. At temperature zero,
the chemical potential merges into the Fermi energy, μ(T = 0) = εF . kB = 0.8617 ×
10−4 eV/K is the Boltzmann factor, so that it also often called the (temperature-
dependent) Fermi energy.
The Fermi gas at finite temperature thus involves Fermi occupation numbers,
which render most calculations non-analytical, except in the case of sufficiently
small temperatures. Let us explore that case. Note first that the term “small temper-
ature” deserves some explanations. In a Fermi system the temperature, as defined in
classical kinetic theory, for example, in the case of perfect gases, has to be under-
stood with respect to the scale determined by the Fermi energy εF or rather the
associated temperature TF = εF /kB . Even at zero temperature, the Pauli exclusion
principle implies a non-vanishing kinetic energy due to the motion of the fermions
in the various occupied single-fermion states. This scales with Fermi temperature
TF , and the notion of low or high temperature is then to be defined with respect to
TF . We will confine the discussions to the case of low temperatures T  TF .
The aim of the calculation in this subsection is to compute the thermal excitation
energy E ∗ stored in the system and relate it to the temperature T . This amounts
to evaluating the specific heat of the system. The excitation energy is defined as
E ∗ (T ) = E(T ) − E(T = 0) where the energy in the Fermi gas is purely kinetic and
reads
∞ ∞
1
E(T ) = V dε D(ε)n T (ε)ε = dε VD(ε) ε−μ(T )
εg(ε), (2.32)
0 0 1+ kB T

where the Fermi distribution (2.31) defines the thermal occupation weights. It has
to be kept in mind that the Fermi energy μ(T ) depends on temperature. It is deter-
mined from the condition that the distribution reproduces the given total number of
particles N as
2.3 Fermi Gas at Finite Temperature 61

1 nT
nT=0
nT–nT=0

0.5
nT

–0.5
|ε−μ|<2T

μ ε
Fig. 2.6 The Fermi distribution n T at finite temperature (dashed), the distribution at T = 0 (dotted)
and their difference (solid)

∞
N= dε n T (ε)g(ε). (2.33)
0

The condition (2.33) and the integral (2.32) for the total energy cannot be
expressed in closed form, but the case of small temperatures (T  TF ) still can
be worked out in detail. Consider the change in occupation relative to the ground
state at T = 0. The situation is sketched in Fig. 2.6. The difference

1
n T − n T =0 = ε−μ(T )
− ϑ(εF − ε)
1+ kB T

is concentrated in a region around μ which is as narrow as T is small (remember

that for small T we have εF ≈ μ). That difference is thus a rapidly changing func-
tion, whereas all other ingredients in the integral vary smoothly. The idea is then to
approximate the smooth parts by a Taylor expansion. We do that in one stroke for the
normalization integral (2.33) and for the energy (2.32) by considering a functional

∞ 
f = VD(ε) for N
I (T )[ f ] = dε n T (ε) f (ε), f (ε) = . (2.34)
0 f = εVD(ε) for E

The Taylor expansion will be applied to the smooth function f (ε) ≈ f (μ) + (ε −
μ) f (μ). The steps read
62 2 The Fermi-Gas Model
∞ ∞
I (T )[ f ] = dεn T =0 (ε) f (ε) + dε [n T − n T =0 ] f (ε)
 0
  0
∞
dε f (ε)=I0

0
∞ ∞

≈ I0 + f (μ) dε [n T − n T =0 ] + f (μ) dε [n T − n T =0 ] (ε − μ).
 0
  0
=0

Substituting (ε − μ)/T = x the remaining integral becomes

∞   ∞
1 1
I − I0 ≈ f (μ)T 2 dx − ϑ(−x) = 2 f
(μ)T 2
dx ,
1+e x 1 + ex
0
0  
π 2 /12

from which finally

∞ μ
T →0 π 2T 2
dε n T f (ε) −→ dε f (ε) + f (μ). (2.35)
0 0 6

This relation is evaluated first to determine μ(T ) in lowest order. According to

(2.34) we identify f = D(ε) and, furthermore, expand μ(T ) = εF + δμ(T ). This
yields
εF εF +δμ
π2
N ≈ dε VD(ε) + dε VD(ε) + T 2 VD (εF ),
6
0    εF  
=N δμVD(εF )

and thus

π 2 D (εF )
δμ = −T 2 . (2.36)
6 D(εF )

One can now use the above expression for μ(T ) in the expansion of the energy
(2.32) for the limit of small temperature. Here, we identify f = εD(ε) and insert it
into expansion (2.35). This yields

εF εF +δμ
π 2 2 d(εD(ε))

E(T ) ≈ dε εVD(ε) + dε εVD(ε) + T V

εF 6 dε εF
0  
E(T=0)

π2 2 π2
≈ E(T = 0) + δμεF VD(εF ) + T εF VD (εF ) + T 2 VD(εF ).
 6  6
=0
2.4 Fermi Gas in Stars: The Example of White Dwarfs 63

The first term of this latter equation is nothing but the energy of the system at zero
temperature. The excitation energy is thus directly obtained as

π2
E ∗ (T ) = E(T ) − E(T = 0) ≈ VD(εF ))T 2 . (2.37)
6
It is to be noted that starting from T = 0 the energy initially grows quadratically
with the temperature. The specific heat c = ∂T E ∗ of a Fermi gas thus begins with a
linear growth and becomes zero in the limit T −→ 0.
Formula (2.37) for the excitation energy is still very general as it allows to insert
different expressions for the level density D. (For an example, see the discussion in
Sect. 1.2.4.3.) The standard application is a Fermi gas in three dimensions for which
the level density (2.28b) applies. Inserting that relation for the level density yields
the excitation energy E ∗ , and the specific heat c, as

π2 N 2 dE ∗ π2 N
E ∗ (T ) = T , c= = T. (2.38)
8 εF dT 4 εF

This simple relation can serve as a quick estimate for a great variety of 3D systems
when inserting the Fermi energies as given in Table 1.2. A typical example of appli-
cation is the level density parameter in nuclei (Sect. 1.2.4.3) and its use for studying
statistical nuclear deexcitation. It holds for low temperatures T  TF . There are
systems, however, for which the temperature range of applicability is also limited
from below. A transition to a BCS condensate (see Sect. 9.4) changes the pattern
below a critical temperature TBCS . Small many-fermion systems can also experience
sizeable shell effects. Estimate (2.38) then still holds averaged over system sizes and
becomes valid for any N as soon as T is of the order of the shell gap which, in turn,
shrinks with N −1/3 .

2.4 Fermi Gas in Stars: The Example of White Dwarfs

One of the probably most famous examples of a Fermi gas is provided by the elec-
tron gas in white dwarfs, for a brief introduction see Sect. 1.1.8. Chandrasekhar
showed that in order to reach stability such a self-gravitating object should have a
mass smaller than the Chandrasekhar mass MCh which is about 1.4 solar masses
[77]. We give a short outline of the derivation below. Let us denote the total number
of electrons in the star by Ne and the stellar radius by R.

2.4.1 “Low” Densities

For a particle density ρ ∼ 0.5 × 1030 cm−3 = 0.5 × 105 a−3
0 (Sect. 1.1.8) and Fermi
energy of order εF ∼ 50 keV, the electrons still move at non-relativistic velocities,
64 2 The Fermi-Gas Model

or momenta k  m e c. The total kinetic energy is then given by (2.24). We express

the Fermi momentum therein by the Wigner–Seitz radius using (2.19), and the star
radius as rs N 1/3 . For the kinetic energy this yields
 2/3  2/3 5/3
3 2 kF2 3 9π 2 N e 3 9π 2 N e
E kin = Ne = 2
= . (2.39)
5 2m e 5 4 2m e rs 5 4 2m e R 2

The electron charge is neutralized by the positive proton charges. Thus the Coulomb
energy becomes negligible and the potential energy is purely gravitational. The
mass of the star mostly stems from the nucleon masses. Assuming equal number
of protons and neutrons, the total number of nucleons becomes 2Ne . We, further-
more, ignore the small mass difference between protons and neutrons. The potential
energy is then the gravitational energy of a homogeneous sphere with radius R and
mass 2Ne m p which reads

3 (2Ne m p )2 12 m 2p 2
E pot = − G =− G N , (2.40)
5 R 5 R e

where m p is the proton (or neutron) mass and G = 6.67 × 10−11 m3 /(kg s2 ) is the
gravitational constant. The total energy of the star then becomes
 2/3 5/3
3 9π 2 N e 12 m 2p 2
E = E kin + E pot = − G N .
5 4 2m e R 2 5 R e

It should be minimized with respect to the radius R at constant mass M, namely

constant Ne . This leads to
 2/3
1 9π 2 (2m p )1/3 −1/3
Req = M , (2.41)
2 4 2m e Gm 2p

expressed as a function of the total mass M = 2m p Ne . Taking for M the solar mass
M = 2 1030 kg leads to a radius R ∼ 8 × 106 m in agreement with observations.

2.4.2 Stability in the Relativistic Domain

The stability radius (2.41) shrinks with increasing total mass M. Decreasing R
means increasing density and consequently increasing Fermi momentum. The non-
relativistic procedure, as outlined in the previous section, provides a stable solution
for every M, or R, respectively. With increasing M, however, one quickly reaches
a regime where the assumption of non-relativistic momenta does not hold anymore.
In this section we want to discuss an upper limit for the mass M, i.e., a lower limit
for the radius R. To that end, we consider the relativistic case. The electronic kinetic
energy (properly subtracting the electron rest mass) then becomes
2.4 Fermi Gas in Stars: The Example of White Dwarfs 65
kF  
2V
E kin = d3
k (ck) 2 + (mc2 )2 − mc2
(2π )3 0
 
2V4π kF m 2 c3
≈ dk k ck − mc +
2 2
(2π )3 0 2k
V V V m 2 3
c 2
= ckF4 − mc2 kF3 + k
4π 2 3π 2 4π  F
2

3 3 m 2 c3
= Ne ckF − Ne mc2 + Ne
4 4 kF
 2/3 4/3  
3 9π Ne 3 4 2/3 2/3 m 2 c3
= c − Ne mc +2
Ne R, (2.42)
4 4 R 4 9π 

where we have used that the volume V is directly linked to the electron number
1/3
by the relation VkF3 /(3π 2 ) = Ne and the relation kF = (4/(9π ))1/3 Ne /R. It is to
be noted that the integration starts from a regime of very small values of k where
the high-momentum expansion is grossly wrong. Looking at the whole integral,
this region is extremely small (k 2 weight) and the error made therein is outweighed
by the huge region of relativistic k, thus validating the above approximation. The
potential energy remains as given before in (2.40). The leading term of the kinetic
energy now has the same trend ∝ R −1 as the potential energy. We thus write the
total energy

 
C 3 4 2/3 2/3 m 2 c3
E= − Ne mc +
2
Ne R, (2.43a)
R 4 9π 
 
3 9π 2/3 12
C= cNe4/3 − Gm 2p Ne2 . (2.43b)
4 4 5

It has a stable minimum if C > 0. For C < 0, however, the minimal energy cor-
responds to a singularity R −→ 0 which means that the white dwarf would be
unstable and implode. The critical point lies at C = 0. It corresponds to a critical
particle number Ne,c and critical mass MCh = 2m p Ne,c , which are given by

  2/3 3/2   2/3 3/2

5 9π c 5 9π c
Ne,c = , MCh = 4/3
. (2.44)
16 4 Gm 2p 16 2 Gm p

The critical mass MCh is known as the Chandrasekhar mass [77]. The star becomes
unstable if M > MCh . On the contrary for M < MCh an equilibrium is possible.
The Chandrasekhar mass thus fixes the maximum mass a self-gravitating object may
possess. Our simple-minded calculation leads to MCh  1.7M (Ne,c ∼ 1.2 1057 )
while more elaborate calculations give a slightly smaller value of MCh  1.45M
[77].
66 2 The Fermi-Gas Model

2.5 Coulomb Energy of a Charged Fermi Gas

We now go one step beyond the Fermi-gas model and consider a homogeneous
electron gas including the repulsive Coulomb interaction between the electrons.
This is, e.g., a model for the valence electrons in bulk metal where the positive
ionic background has been smoothed to a homogeneous jellium density (1.1). The
Hamiltonian of that many-body system is the sum of kinetic energy and Coulomb
interaction:

N
p̂i2 1 
N
e2
Ĥ = + . (2.45)
i=1
2m e 2 i= j=1 |ri − r j |

For a finite system of independent particles the energy then reads

E = E kin + E pot,dir + E pot,ex . (2.46a)

The kinetic energy was discussed and evaluated in Sect. 2.2.5. The potential energy
consists of two terms,

1 e2
E pot,dir = dx dx ϕα∗ (x)ϕβ∗ (x ) ϕ (x)ϕβ (x )
2 αβ |r − r | α

1 e2
= d3 r d3r ρ(r) ρ(r ) , (2.46b)
2 |r − r |

1 e2
E pot,ex =− dx dx ϕα∗ (x)ϕβ∗ (x ) ϕ (x)ϕα (x )
2 αβ |r − r | β

1 e2
=− d3r d3 r δνν (rν, r ν ) (r ν , rν). (2.46c)
2 νν |r − r |

The direct part of the Coulomb energy, E pot,dir , diverges dramatically. Even when
considering the energy per volume (or per particle), there remains a quadratic diver-
gence, as the long-range Coulomb force does not allow infinite amounts of charge.
The average electron charge needs to be compensated by an equally dense positively
charged background with ρback = ρ0 = constant, since the slightest amount of
finite charge density would again lead to a divergent Coulomb energy. Actually, an
electron gas in the degenerate regime (T  εF ) is realized by the valence electrons
in a metal. Here, the positive metal ions serve as neutralizing background. It is not
homogeneous. But one often ignores the detailed ionic structure and smooths the
total positive charge to a homogeneous density distribution, often called the jellium
approximation, see (1.1). It is justified by the argument that the electrons have long
wavelengths which cannot resolve the ionic details anyway and thus deliver nearly
2.5 Coulomb Energy of a Charged Fermi Gas 67

constant density. After the approximation ρback = ρ0 , we have the quite trivial
result that the total direct Coulomb energy is zero because the total charge density
(ρback − ρ0 ) vanishes everywhere.
The exchange energy (2.46c) requires quite a lengthy calculation. On the other
hand, it serves as a welcome example to practice a moderately lengthy formal cal-
culation in detail. Before carrying on, we first summarize a few integrals which will
be needed in the following:

exp (ik·r) 4π
d3r = 2 , (2.47a)
r k

log |ax + b|
dx (ax + b)−1 = , (2.47b)
a

k + k

dk k log

= kk + 1 (k 2 − k 2 ) log
k + k
(2.47c)
k−k
2
k − k

kF + k
1 4

dk k 3 log

= (k − k 4 ) log
k F + k
+ 1 k 2 k 2 + 1 k F k 3 .(2.47d)
kF − k
4 F
kF − k
2 F 6

Inserting 
 the plane-wave eigenstates (2.12) and performing the continuum limit
α −→ σ d3 k yields

 d3r d3r d3 kd3 k exp i(k − k )·(r − r )

+ +
E pot,ex =e 2
χ σ χ σ χ σ χσ .
(2π )6    |r − r |
σσ
δσ σ

It is advantageous to transform to center of mass and relative coordinates as

1
R= (r + r ) , r̃ = r − r .
2

The integration volume element changes as d3r d3r −→ d3 Rd3 r̃ . We are now in
shape to go through the steps quickly, using the integrals of (2.47). In a first step,
spatial integration is performed:

e2 exp i(k − k )· r̃
E pot,ex = − d3 R d3 kd3 k d3r̃
(2π )6 r̃
     
V (2.47a)

4π e2 1
=− V d3 kd3 k 2 .
(2π )6 k + k − 2k·k
2

Again, it is advantageous switching from Cartesian coordinates for the wave vec-
tors to spherical ones. The choice of the polar axis is free and a proper choice can
save a huge amount of work. First, we observe that the integrand depends only
on three quantities, the absolute values of the wave vectors k as well as k and
the angle between the two vectors cos(θ ) = k · k /(kk ). After the d3 k integration
68 2 The Fermi-Gas Model

and the change to spherical coordinates d cos(θ ) dφ dk k 2 , the integration over d3 k

may be attacked for the given k. The choice of the axis for the spherical coordi-
nates of k is still free. We choose the z-axis parallel to the vector k . The polar
angle θ in that frame is precisely the angle between k and k as given above. The
azimuthal angle φ thus does not appear in the integrand because the relative distance
|k − k | is invariant under azimuthal rotations. The angles θ and φ also do not
appear anywhere in the integrand. This simplifies the integration greatly. The trivial
1 2π 2π
part is −1 d cos(θ ) 0 dφ 0 dφ = 2(2π )2 . There remain only three non-trivial
integrals
kF kF +1
32π 3 e2 1
E pot,ex = − V dk k 2 dk k 2 d(cos θ )
(2π )6 0 0 −1 k2 + k 2 − 2kk cos θ
use (2.47b)
kF

e2 kF
k + k

=− V dk k dk k log

2π 3 0 0 k − k

use (2.47c)





e2 kF
2 1

k + kF
31

k + kF

=− V

dk k k F + kk F log

− k log

2π 3 0 2 k − kF
2 k − kF

use (2.47d)
 
e2 k 4F 1 1 1 1 2e2 k 4F
=− V + − − =− V .
2π 3 3 2 4 12 (2π )3

This energy also grows to infinity with V like the kinetic energy, so that it is prefer-
able similarly to consider the finite energy per particle

E pot,ex E pot,ex 1 2e2 kF

= =− . (2.48)
N V ρ0 (2π )3

This grows ∝ kF , i.e., more slowly than the kinetic energy.

After all, the total energy becomes

E E kin E pot,ex 2 3 2 3e2 2 2

= + = kF − kF = 0.6 k − 0.239 e2 kF . (2.49)
N N N 2m e 5 4π 2m e F

The electron–electron interaction is repulsive while the exchange term has the oppo-
site sign and so becomes attractive. That is plausible. But a quick glance at the
total energy (2.49) gives the puzzling impression that all binding comes from the
electron–electron interaction which is usually considered to be repulsive. Note that
the direct term is the one responsible for repulsion and we have seen that it is huge,
in fact, insurmountably infinite. It is, however, fully counterweighted by the equally
huge attractive contribution from the external positively charged background. The
2.6 Concluding Remarks 69

Fig. 2.7 The binding energy

6 kinetic
per particle for the Fermi gas
of electrons, (2.49), and its
two constituents, the kinetic 4
energy and the exchange
energy 2

E/N [eV]
0
total
-2

-4
exchange
-6
0 0.2 0.4 0.6 0.8
kF [1/a0]

full compensation of the leading forces leaves the small exchange effect on the final
tip in the balance and that small contribution provides the binding in the Fermi-gas
model, which is counterbalanced by the Fermi pressure from the kinetic energy. The
trends are sketched in Fig. 2.7. The linear growth of (negative) exchange energy
dominates for small Fermi momenta kF . It is overruled by the quadratic trend of the
repulsive kinetic energy for increasing kF . The turnover produces a unique mini-
mum which establishes the equilibrium state of the electron gas, having the energy
E/N = −1.295 eV at the Fermi momentum kF = 0.4/a0 corresponding to rs = 4.8
a0 . It is remarkable how well this number fits typical Wigner–Seitz radii of metals,
e.g., rs = 3 a0 for Ag, rs = 4 a0 for Na, or rs = 5 a0 for K. Of course, the ionic
structure, adding core polarizability and core repulsion, determines the final detailed
value of rs , but the basic balance seems to be determined already by the electron gas,
a simple model with surprisingly realistic aspects.

2.6 Concluding Remarks

The Fermi-gas approximation, simple though it is, was shown to be surprisingly suc-
cessful in understanding many features of quite distinct systems at least in a semi-
quantitative way. The reason was that many systems are characterized by a relatively
constant density and weak correlations, which makes the approximation by a gas of
non-interacting particles trapped in a constant potential quite adequate. Through
the exploration of the thermal behavior, relativistic formulation, and the addition
of gravitational and Coulomb interactions we have shown the tremendous flexibility
that also characterizes the model. For a first basic understanding of fermion systems,
the Fermi-gas model remains an approach of choice.
The next chapters will show how to introduce more realism by discarding the
constant potential using either a phenomenological potential adapted to specific
70 2 The Fermi-Gas Model

systems or a self-consistent one calculated from the single-particle wave functions

themselves. In both cases, the many-body state is still a wave function built out
of single-particle states that are filled up to Fermi energy, so that many qualitative
properties of the Fermi-gas model remain valid.
http://www.springer.com/978-3-642-03838-9