UNIVERSIDADE DE COIMBRA

Departamento de Física
Lectures by Pedro Borlido, Tiago Cerqueira

Estrutura Electrónica e Modelação Computacional

WS 2024/2025
Homework - 02
• Please upload your answers to the Inforstudent platform. If you do not have access to this
platform, send it by email from your institutional address.

• Submitted answers must be uploaded as a compiled file (.tar, .tar.gz, .bzip)

• Any written answers must be uploaded in PDF format.

• The topics covered here have been discussed in the theory classes and in solid state physics
courses.

• There is a lot of text in this homework, but there aren’t really many calculations. Please read
the text carefully to avoid wasting time... Solving this shouldn’t take more than 1-2 hours.

1 Homogeneous Electron Gas

1. 1
Here we will discuss jellium, a classical problem which is explored in almost all condensed matter
physics courses, and is particularly useful to showcase some of the deficiencies of Hartree-Fock
for solids.
The jellium model (or homogeneous electron gas) is model system comprised of a homogeneous
electron chage density, n, which is sometimes represented in terms of the Wigner-Seitz radius
(the average radius for a sphere containing 1 electron),

‹1/3
3

rs = . (1)
4πn

To this system is added a homogeneous positive charge background which maintains electrical
neutrality. The Hartree-Fock equations of the homogeneous (interacting) electron gas can be
solved exactly using plane waves, i.e. the single-particle wave-functions are written as (within
the unrestricted formalism)

1
ϕk,m (x) = p e ik·r ηm (σ) , (2)
V

where V is the size of the box, η are the spinors and k is the wave-vector which defines the
plane-waves. In general, we have to assume that the two spin channels are not equivalent, i.e.
we have N↑ electrons in the spin-up channel and N↓ electrons in the spin-down channel, meaning
that the density can be split as

1
Estrutura Electrónica e Modelação Computacional Homework - 02

n(r) = n↑ (r) + n↓ (r) , (3)

leading to a magnetization density

m(r) = n↑ (r) − n↓ (r) . (4)

Equivalently, we can work with the total density and the relative polarization

n↑ (r) − n↓ (r)
ζ= . (5)
n↑ (r) + n↓ (r)
(a) Write the Fermi momentum for each channel in terms of that’s channel density (or equiva-
lently, the total density and the polarization). (Hint: You know that each spin channel has
a fixed number of electrons! Just have to count the number of wave vectors that fill up to
that number...)
(b) Starting from the HF equations, find the expression for the eigenvalues of the single particle
wavefunctions. (Hint 1: in this case the Hartree and external potentials cancel each other
- why? - simplifying the equations) (Hint 2: This is not a Calculus class! If you get to an
annoying integral feel free to invoke integration tables or to use Wolfram...)
(c) What happens to the eigenvalue (and its derivative) at the Fermi level?
(d) Write the total energy of the system for the paramagnetic case (N↑ = N↓ ⇔ ζ = 0) and for
the ferromagnetic case (N↑ = N , N↓ = 0 ⇔ ζ = 1)
(e) This system has a transition between these two regimes. At which value of the density does
this happen within Hartree-Fock theory? (Hint: the actual transition takes place at rs ≈ 75
Bohr−3 .)

More Hints:

Ω
R
= d3 k for k-vector sums in reciprocal space.
P
• You may want to exploit the relation k (2π)3

• The occurring volume integral over k can be solved analytically in spherical coordinates choos-
ing the z axis of your coordinate system to be parallel to r − r′ .

Page 2
Estrutura Electrónica e Modelação Computacional Homework - 02

2 Configuration Interaction

The following problem is designed to introduce the concepts of basis sets, Configuration Interac-
tion, spin adapted configurations and so on that were mentioned in the theory classes, with the H2
molecule as an underlying example.

The problem may seem rather large, but notice that there is a lot of text and the questions themselves
are rather straighfoward, focusing heavily on identifying and regrouping integral terms. Give your
answers in terms of one- and two-particle integrals using the notation of Szabo & Ostlund for the
spin

〈ϕα |h|ϕβ 〉 , 〈ϕα ϕβ |ϕγ ϕδ 〉

and space integrals

(ϕa |h|ϕ b ) , (ϕa ϕ b |ϕc ϕd )

with α = {a, m}.

No explicit integration of the orbitals is necessary.

1. Minimal Basis for H2

We will follow the traditional approach of building molecular orbitals as a linear combination of
atomic orbitals (LCAO). These atomic orbitals form a basis that is normalized but not orthogonal,
i.e., they obey (
〈χA|χA〉 = 〈χB |χB 〉 = 1
. (6)
〈χA|χB 〉 = 〈χB |χA〉 = SAB

For simplicity, we will use a minimal basis consisting of a single orbital per atom (e.g. 1s hydro-
gen orbitals), each centered at the corresponding atom of the molecule:
(
χA(r) = χ1s (r − RA)
, (7)
χB (r) = χ1s (r − RB )

where RA (RB ) is the position of the first (second) nucleus.

Using this basis we can then build two spatial molecular orbitals of the form

ϕ(r) ∝ cAχA(r) + cB χB (r) (8)

where the coefficients by solving the secular equation |H−ES| = 0 and corresponding eigenvalue
problem. However, in this simple case we can avoid all of these troubles as (i) we can assume

Page 3
Estrutura Electrónica e Modelação Computacional Homework - 02

the coeficients and the wavefunctions to be real, (ii) the molecular orbitals are agnostic to the
A ↔ B change (as we have a homonuclear molecule), (iii) the spatial wavefunctions must
be either symmetric/anti-symmetric with respect to the center of the molecule and (iv) the
molecular orbitals are orthonormalized.

(a) Write the values for the cA and cB coefficients for each of the two possible molecular orbtials,
in terms of SAB .

(b) Identify each of the two orbitals you found with the bonding (ϕ b ) and anti-bonding (ϕa )
orbitals commonly seen in molecular physics.

2. Restricted Hartree-Fock for H2

Having identified the bonding (b) and anti-bonding (a) molecular orbitals we can build the
Hartree-Fock ground state determinant for the H2 molecule. In this case, it is a determinant
where both electrons are in the bonding orbital with opposite spins, i.e.,

1 ϕ b↑ (x1 ) ϕ b↑ (x2 )
Φ b↑,b↓ (x1 , x2 ) = p
2 ϕ b↓ (x1 ) ϕ b↓ (x2 )

where ϕ b↑ (x1 ) = ϕ b (r1 )η↑ (σ1 ), ϕ b↓ (x1 ) = ϕ b (r1 )η↓ (σ1 ) and ηm (σ) are the up/down spinors.

(a) Expanding the molecular orbitals in the determinant, show that this wavefunction can be
split into Slater determinants of the atomic orbitals.

(b) Identify each of the determinants with the corresponding electronic configuration. Con-
sidering that the dissociation of H2 leads to two H, argue why this wavefunction cannot
correctly describe the dissociation of the H2 molecule.

(c) Compute the energy of the H2 molecule, and show that in the limit of infinite separation it
becomes

1
lim EH2 = 2(χA|hA|χA) + (χAχA|χAχA)
RAB →+∞ 2
1
= 2EH + (χAχA|χAχA)
2

where hA = −1/2∇2 − 1/|r − RA|. Due to the spurious Coulomb interaction term, RHS is
not size consistent. (Hint: You do not need to explicitly compute all integrals over atomic
wavefunctions, just argue what happens to them in the limit of vanishing overlap of the
atomic wavefunctions).

3. Excited determinants for H2

To bypass the problems with RHS we can turn to Configuration Interaction (CI). The CI wave-
function is built as a sum of excited Slater determinants, for which in this minimal basis
set we have 5 (4 singly excited and 1 doubly excited), corresponding to the configurations

Page 4
Estrutura Electrónica e Modelação Computacional Homework - 02

ϕa ↑↓ ↑ ↓ ↑ ↓
ϕb ↑↓ ↑ ↓ ↓ ↑

Φ0 Φ1 Φ2 Φ3 Φ4 Φ5

where Φ0 = Φ b↑,b↓ is the ground state configuration. Although these functions are eigenfunctions
of the Sz operator,
N
X
Sz = sz,i ,
i

some of them are not eigenvalues of the total spin operator,

N X
X N
S2 = si · s j
i j

which complicates their use in CI...

(a) Write the Slater determinants for each of these orbitals, separating them (if possible) into
a product of molecular space and spin functions (i.e. in the form f (x) = g(r)h(σ)).

(b) We can solve our problems if instead of using the previous determinants we use spin adapted
configurations. By using linear combinations of determinants we can generate new wave-
functions that are eigenvalues of S 2 and can be separated into spin and spatial parts. In
this case we just have to replace two of them:
 Φ4 + Φ5
Φ4′ = p

2
Φ4 − Φ5
Φ5′ = p

2

Repeat the previous part for the new wavefunctions.

4. CI with spin adapted configurations for H2

The CI wavefunction can then be written as

|ΨCI 〉 = c0 |Φ0 〉 + c1 |Φ1 〉 + c2 |Φ2 〉 + c3 |Φ3 〉 + c4′ |Φ4′ 〉 + c5′ |Φ5′ 〉

where the Ψi correspond to the configurations indicated in the previous part. In principle solving
the CI problem would require solving a 6x6 matrix eigenvalue problem,

Page 5
Estrutura Electrónica e Modelação Computacional Homework - 02

Φ0 Ĥ Φ0 Φ0 Ĥ Φ1 Φ0 Ĥ Φ2 Φ0 Φ3 Φ0 Φ 4′ Φ0 Φ 5′
 
Ĥ Ĥ Ĥ

 · Φ1 Ĥ Φ1 Φ1 Ĥ Φ2 Φ1 Ĥ Φ3 Φ1 Ĥ Φ 4′ Φ1 Ĥ Φ 5′ 

· · Φ2 Ĥ Φ2 Φ2 Ĥ Φ3 Φ2 Ĥ Φ 4′ Φ2 Ĥ Φ 5′
 
H= ,
 
 · · · Φ3 Ĥ Φ3 Φ3 Ĥ Φ 4′ Φ3 Ĥ Φ 5′ 
 
 · · · · Φ4′ Ĥ Φ4′ Φ4′ Ĥ Φ5′ 
· · · · · Φ5′ Ĥ Φ5′
which is possible but rather tedious to do by hand.
Fortunately, the problem can be greatly simplified thanks to the use of spin adapted configura-
tions, which transform the 6x6 matrix into a block diagonal matrix. From this, our problem for
the ground state simply becomes a 2x2 matrix which is easily solvable by hand.
(a) Prove the previous statement by showing that the matrix elements in blue in the previous
matrix are zero. (Hint: argue that you only need to look at the spin parts of the wavefunc-
tions; use the spinor orthogonality; remember Brillouin’s theorem).

5. Full CI for H2
Given everything we saw in the previous question, our CI wavefunction for the ground states
becomes simply
|ΨCI 〉 = c0 |Φ0 〉 + c1 |Φ1 〉 = c0 |Φ b↑,b↓ 〉 + c1 |Φa↑,a↓ 〉

i.e., only the singlet ground state and doubly exicted states couple. The corresponding CI eigen-
value problem is therefore
Hc = ECI c

where H is now a 2x2 matrix and ECI = EHF + Ecorr .

(a) Show that the previous CI matrix can be recast into the form
‚ Œ
0 K
c = Ecorr c ,
K 2∆

and find the values for K and ∆.

(b) Find the expression for the CI coefficients and the und state CI energy (Hint: use interme-
diate normalization, c0 = 1).
(c) Using the information from the previous parts, write the CI wavefunction and the CI ground
state energy in the limit of infinite separation. What are the major differences with respect
to HF? Do you recover the correct limit? (Hint: Use the same reasoning as for the infinite
separation limit in Hartree-Fock.)

Page 6