Tight-Binding Model for Graphene

Franz Utermohlen

September 12, 2018

Contents
1 Introduction 2

2 Tight-binding Hamiltonian 2
2.1 Energy bands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Hamiltonian in terms of Pauli matrices . . . . . . . . . . . . . . . . . . . . . 4
2.3 σz term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3 Behavior near the Dirac points 6

3.1 Near K . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Near K0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3 Linear dispersion relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.4 σz term gapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1
1 Introduction
The unit cell of graphene’s lattice consists of two different types of sites, which we will call
A sites and B sites (see Fig. 1).

Figure 1: Honeycomb lattice and its Brillouin zone. Left: lattice structure of graphene, made out
of two interpenetrating triangular lattices a1 and a2 are the lattice unit vectors, and δi , i = 1, 2, 3
are the nearest-neighbor vectors. Right: corresponding Brillouin zone. The Dirac cones are located
at the K and K’ points.
Note: This figure and caption are from Castro et al. [2].

These vectors are given by

a √

a1 = 3, 3
2
a √

a2 = 3, − 3
2
a √

δ1 = 1, 3
2
a √

δ2 = 1, − 3
2

δ3 = −a 1, 0
2π √

K= √ 3, 1
3 3a
2π √
K0 = √


3, −1 , (1)
3 3a
where K, K0 are the corners of graphene’s first Brillouin zone, or Dirac points.

2 Tight-binding Hamiltonian
Considering only nearest-neighbor hopping, the tight-binding Hamiltonian for graphene
is X †
Ĥ = −t (âi b̂j + b̂†j âi ) , (2)
hiji

2
where i (j) labels sites in sublattice A (B), the fermionic operator â†i (âi ) creates (annihilates)
an electron at the A site whose position is ri , and similarly for b̂†j , b̂j . We can rewrite the
sum over nearest neighbors as
X † XX †
(âi b̂j + b̂†j âi ) = (âi b̂i+δ + b̂†i+δ âi ) , (3)
hiji i∈A δ

where the sum over δ is carried out over the nearest-neighbor vectors δ1 , δ2 , and δ3 , and the
operator b̂i+δ annihilates a fermion at the B site whose position is ri + δ. Using
1 X ik·ri †
â†i = p e âk , (4)
N/2 k

where N/2 is the number of A sites, and similarly for b̂†i+δ , we can write the tight-binding
Hamiltonian for graphene (Eq. 2) as
t X X i(k−k0 )·ri −ik0 ·δ †
Ĥ = − [e e âk b̂k0 + H.c.]
N/2 i∈A δ,k,k0
X
= −t (e−ik·δ â†k b̂k + H.c.)
δ,k
X
= −t (e−ik·δ â†k b̂k + eik·δ b̂†k âk ) , (5)
δ,k

where in the second line we have used

X 0 N
ei(k−k )·ri = δkk0 . (6)
i∈A
2

We can therefore express the Hamiltonian as

X
Ĥ = Ψ† h(k)Ψ , (7)
k

where  
âk
Ψ† = â†k b̂†k ,


Ψ≡ , (8)
b̂k
and  
0 ∆k
h(k) ≡ −t (9)
∆∗k 0
is the matrix representation of the Hamiltonian and
X
∆k ≡ eik·δ . (10)
δ

3
2.1 Energy bands
p
The eigenvalues of this matrix are E± = ±t ∆k ∆∗k . We can compute this by writing ∆k
out more explicitly:
∆k = eik·δ1 + eik·δ2 + eik·δ3
= eik·δ3 [1 + eik·(δ1 −δ3 ) + eik·(δ2 −δ3 ) ]
h √ √ i
= e−ikx a 1 + ei3kx a/2 ei 3ky a/2 + ei3kx a/2 e−i 3ky a/2
h √ √ i
−ikx a i3kx a/2 i 3ky a/2 −i 3ky a/2
=e 1+e (e +e )
"  √  #
−ikx a i3kx a/2 3
=e 1 + 2e cos ky a . (11)
2
The energy bands are therefore given by
s   √  √ 
3 3 2
3
E± (k) = ±t 1 + 4 cos kx a cos ky a + 4 cos ky a , (12)
2 2 2
or, as it is sometimes written,
p
E± (k) = ±t 3 + f (k) , (13)
where  √
√
 
3 3
f (k) = 2 cos( 3ky a) + 4 cos kx a cos ky a . (14)
2 2
These are two gapless bands that touch at the Dirac points K and K0 (see Fig. 2). In other
words, the Dirac points are the points in k-space for which E± (k) = 0.

Figure 2: Energy bands for graphene from nearest-neighbor interactions. The bands meet at the
Dirac points, at which the energy is zero.

2.2 Hamiltonian in terms of Pauli matrices

We can express the Hamiltonian

 
0 ∆k
h(k) = −t (15)
∆∗k 0

4
in terms of Pauli matrices by expressing ∆k as
X X
∆k = eik·δ = [cos(k · δ) + i sin(k · δ)] , (16)
δ δ

so the Hamiltonian reads

X 0 cos(k · δ) + i sin(k · δ)

h(k) = −t . (17)
cos(k · δ) − i sin(k · δ) 0
δ

In terms of Pauli matrices, the Hamiltonian is then

X
h(k) = −t [cos(k · δ)σx − sin(k · δ)σy ] . (18)
δ

2.3 σz term

Writing the Hamiltonian in the form

X
h(k) = −t [cos(k · δ)σx − sin(k · δ)σy ] (19)
δ

it is natural to ask whether we can have also have a σz term. Let’s see what happens when
we do:
X
hM (k, M ) = −t [cos(k · δ)σx − sin(k · δ)σy + M σz ]
δ
X M cos(k · δ) + i sin(k · δ)

= −t . (20)
cos(k · δ) − i sin(k · δ) −M
δ

From the matrix form of the Hamiltonian we can see that this additional σz term raises the
energy of A sites and lowers the energy of B sites, thereby gapping the bands. We can think
of this as being caused by on-site potential terms of the form
!
X X
Ĥpotential = M n̂a,α − n̂b,β , (21)
α β
{A sites} {B sites}

where n̂a,α = â†α âα and n̂b,β = b̂†β b̂β . This could arise when there are two different types of
atoms on the A and B sites, such as in boron nitride, which has the same lattice structure
as graphene, but whose A and B sites correspond to boron atoms and nitrogen atoms.

5
3 Behavior near the Dirac points

3.1 Near K

Let’s look at the behavior of ∆k about the Dirac point K. Defining the relative momentum
q ≡ k − K, we can write ∆k in terms of q as
" √ #
3(K y + q y )a
∆K+q = e−iKx a e−iqx a 1 + 2ei3(Kx +qx )a/2 cos
2
"  √ #
π 3a
= e−iKx a e−iqx a 1 − 2e3iaqx /2 cos + qy . (22)
3 2

Now, expanding this about q = 0 to first order, we have

3a
∆K+q = −ie−iKx a (qx + iqy ) . (23)
2
The phase of ∆K+q carriesp
no physical significance (since, for example, the energy bands are
given by E± (K + q) = ±t ∆K+q ∆∗K+q ), so it is convenient to ignore the phase of ie−iKx a .
We thus have
3a
∆K+q = − (qx + iqy ) . (24)
2
About the Dirac point K, the Hamiltonian is thus
 
0 qx + iqy
h(K + q) = vF , (25)
qx − iqy 0
where
3at
vF = (26)
2
is the Fermi velocity. We can express this in terms of Pauli matrices as
h(K + q) = vF (qx σx − qy σy ) , (27)
or, defining the vector  
qx
q̄ ≡ , (28)
−qy
we can express h(K + q) more naturally as
h(K + q) = vF q̄ · σ . (29)

3.2 Near K0

Similarly, defining the relative momentum q ≡ k − K0 and expanding ∆k about the Dirac
point K0 , we find
3a
∆K0 +q = − (qx − iqy ) , (30)
2
6
so the Hamiltonian about this Dirac point is
 
0 0 qx − iqy
h(K + q) = vF = vF (qx σx + qy σy ) = vF q · σ . (31)
qx + iqy 0

3.3 Linear dispersion relation

From the matrix form of the Hamiltonian near the Dirac points (Eqs. 25, 31) we find that
the energy bands near the Dirac points are given by

E± (q) = vF |q| . (32)

Near a Dirac point, the dispersion relation is therefore linear in momentum, so the energy
bands form cone (called a Dirac cone; see Fig. 3) with the vertex lying on the Dirac point.

Figure 3: Conical behavior of the bands near a Dirac point, known as a Dirac cone, where the
energy is linear in momentum.

3.4 σz term gapping

If we add the σz term that we introduced in Section 2.3, the Hamiltonian near the Dirac
points will be of the form
 
M qx − iqy
hM = vF (qx σx + qy σy + M σz ) = vF . (33)
qx + iqy −M

As we saw, this additional term gaps the energies, so that near the Dirac points, the energies
become q
EM,± = ± qx2 + qy2 + M 2 , (34)
See Fig. 4 for a plot of the band gap near a Dirac point.

7
 Figure 4: Energy gap near a Dirac point produced by a σz term. Compare to Fig. 3.

References
[1] Anthony J. Leggett. Lecture 5: Graphene: Electronic band structure and Dirac fermions.
http://web.physics.ucsb.edu/~phys123B/w2015/leggett-lecture.pdf, 2010.

[2] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim.

The electronic properties of graphene. https://journals.aps.org/rmp/abstract/10.
1103/RevModPhys.81.109, 2009.