Tight-Binding Model of Electronic Structures

Consider a collection of N atoms. The electronic structure of this system refers to its electronic wave
function and the description of how it is related to the binding energy that keeps the atoms together. In
an independent electron approximation, a single electron time-independent Schrdinger equation,
! !
H (r ) = (r ) , (1)
is solved to find the eigenstates, ( = 1, 2, ...), and the corresponding eigenenergies, , of the

Hamiltonian operator,
1
H = 2 +V (r ) , (2)
2

where V (r ) is the potential energy operator and we have used the atomic unit. (For example, the
density functional theory provides a framework to derive an effective single-electron potential energy
operator, which incorporates the interaction among the many electrons [1-3].)
1. Tight-Binding Model
In the tight-banding model of electronic structures, single-electron wave functions are expanded in
terms of atomic orbitals [4,5],
nlm (r, , ) = Rnl (r)Ylm ( , ) , (3)
centered around each atom, where Rlm and Ylm are radial and spherical-harmonics functions in polar
coordinates. (This scheme is also called the linear combination of atomic orbitals, or LCAO.) In Eq.
(3), n, l and m are the principal, angular-momentum and magnetic quantum numbers, respectively. For
example, 1s atomic orbital (n = 1 and l = m = 0) is spherically symmetric and is taken as positive
definite (see the figure below).

For p atomic orbitals, we work with the Cartesian representation, such that the three orthogonal p
orbitals are along the x, y and z axes [4-6] (see the figure below),

3
1/2 x / r
n1m (x, y, z) = Rn1 (r) y / r , (4)
4 z / r
where r = x 2 + y 2 + z 2 .

1
HOPPING INTEGRALS
As a specific example, we consider a collection of silicon atoms, such as a silicon cluster. The
electron configuration in a free silicon atom is 1s22s22p63s23p2, and thus there are four valance electrons
(shown in bold) in the outer shell, which mainly contribute to the chemical bonding. We will represent
the electronic structures of silicon clusters as a linear combination of four atomic orbitals per atomone
3s orbital and three 3p orbitals, 3px, 3py and 3pz, centered around each silicon atom,
N
(r ) = ci (r ri ) . (5)
i=1 {s, px , py , pz }

(The effects of inner shell electrons can be effectively included using pseudopotential methods [7].)
To solve the eigenvalue problem, Eq. (1), we need the Hamiltonian matrix elements between these
atomic orbitals at different interatomic distances. In tight-binding methods, these so called hopping
integrals are fitted as analytic functions of the interatomic distance, r. For the sp-bonding, there are only
four nonzero hopping integrals as shown in the figure below, in which and bondings are defined
such that the axis of the involved p orbitals are parallel and normal to the interatomic vector,
respectively.

In this lecture, we adopt the tight-binding model of silicon by Kwon, et al. [8], in which the hopping
integrals are fitted as
! s1 H s2 = ss %
# #
## s1 H p2d = sp ##
h (r) = " &
# p1d H p2d = pp #
# # , (6)
#$ p1n H p2n = pp #
'
n ( / ( +n ( +n 2+
( r0 + r r
= h (r0 ) * - exp * n 1 * - + * 0 - 4-
)r, * 1 ) r , ) r , 4-
) 0 3,
whereas the diagonal Hamiltonian elements on each atom are given by
s H s = Es
. (7)
px H px = py H py = pz H pz = E p
In Eq. (6), p1d and p1n denote the p orbitals parallel and normal to the bonding axis, respectively, centered
at the first atom. The parameters in Eqs. (6) and (7) are listed in the tables below.
2
 r0 () n Es (eV) Ep (eV)
2.360352 2 -5.25 1.20
ss sp pp pp
h (r0) (eV)

-2.038 1.745 2.75 -1.075
n
9.5 8.5 7.5 7.5
r ()

3.4 3.55 3.7 3.7

To convert the values into atomic units, divide all the lengths by the Bohr radius, aB = 0.5291772083
, and all the energies by the Hartree energy, EH = 27.2113834 eV.
2. Projection of Hopping Integrals
In the tight-binding model presented in the previous section, the electronic wave functions are
expanded in terms of the p orbitals along the Cartesian x, y and z axis, whereas the hopping integrals are
parameterized for p orbitals that are parallel or normal to the bonding directions. To construct the
Hamiltonian matrix elements, we need to decompose the Cartesian p orbitals into the bond-parallel and
bond-normal p orbitals.
PROJECTION OF S-P INTEGRALS
Consider the Hamiltonian matrix element, s|H|p , between the s orbital, |s, on one atom and one of

the p orbitals, |p ( = x, y, z), on another atom. Let d be the unit vector along the bond from the first

atom to the second. In the figure below, a is the unit vector along one of the Cartesian (x, y or z) axes.

We first decompose the p orbital along a into two p orbitals that are parallel and normal to d ,
respectively (see the figure below):

pa = a d pd + a n pn , (8)

where n is the unit vector normal to d within the plane spanned by d and a .

3
 Let be the angle between vectors d and a . Consider an arbitrary point in the 3D space, whose
polar angle from the d axis is . On this point, the value of the p basis function around the a axis is
given by
pa = cos ( )
= cos cos + sin sin
= cos pd + sin pn ,
" %
= a d pd + cos $ ' pn
#2 &
= a d p + a n p
d n

where we have used a trigonometric addition theorem to derive the second line from the first. //
The Hamiltonian matrix element is then given by

s H pa = s H a d pd + a n pn = a d hsp (r ) ,
( ) ( ) (9)

where hsp (r) is the Hamiltonian matrix element in Eq. (6) evaluated at atomic distance r. Note the
overlap between the |s and |pn orbitals is zero by symmetry. To obtain explicit formula in terms of px,
py and pz centered around the first second atom, let us introduce the directional cosines along the x, y
and z axes as d = (d x , d y , dz ) . Then

s1 H p2 x p1x H s2 d h (r )
x sp
s1 H p2 y = p1y H s2 = d y hsp ( r ) . (10)
s H p p1x H s2 d h (r )
1 2z z sp
In Eq. (10), the matrix elements between the p orbitals on the first atom and the s orbital on the second
atom are obtained simply by inverting the direction of the bonding unit vector, d , i.e., changing the
signs of the directional cosines.
PROJECTION OF P-P INTEGRALS
Consider two unit vectors a1 and a2 , each of which is one of the Cartesian unit vectors, i.e., x , y
and z , along the x, y and z axes, respectively. Let d be the unit vector along the bond between two
atoms.

4
 First, the p orbital along a1 can be decomposed into two p orbitals that are parallel and normal to d ,
respectively, as we did in Eq. (8):

p1 = a1 d pd1 + a1 n1 pn1 , (11)

where n1 is the unit vector normal to d within the plane spanned by d and a1 . (We define the origins
of pd1 and pn1 to be at the first atom.) The p orbital along a2 is decomposed in a similar manner.

Now the Hamiltonian matrix element between |p1 and |p2 is given by

p1 H p2 = a1 d pd1 + a1 n1 pn1 H a2 d pd 2 + a2 n2 pn2

( ) ( ) , (12)
= ( a d ) ( a d ) p
1 2 d1 H pd 2 + ( a1 n1 pn1 ) H ( a2 n2 pn2 )
where we have used the fact that the matrix elements between orthogonal p orbitals are zero by
symmetry. Note that
pd1 H pd 2 = hpp (r ) , (13)
and

(a n
1 1 pn1 ) H ( a2 n2 pn2 ) = ( a1 n1 ) ( a2 n2 ) pn1 H pn2
= ( a1 n1 ) ( a2 n2 ) ( n1 n2 ) hpp (r )
= (( a1 n1 ) n1 ) (( a2 n2 ) n2 ) hpp (r ) , (14)

( ( ))( ( ))
= a1 a1 d d a2 a2 d d hpp (r )

where we have used the relation,

( a1 n1 ) n1 = a1 ( a1 d) d ,
see the figure below.

Substituting Eqs. (13) and (14) in Eq. (12), we obtain

( )( ) ( ( ))( (
p1 H p2 = a1 d a2 d hpp (r ) + a1 a1 d d a2 a2 d d hpp (r ) , )) (15)

5
 To obtain explicit formula in terms of px, py and pz centered around the first and second atoms, let us
introduce the directional cosines along the x, y and z axes as d = (d x , d y , dz ) . Then

p H p p1x H p2 y p1x H p2 z
1x 2x

1yp H p 2x p 1y H p2y p 1y H p 2z
p H p p1z H p2 y p1z H p2 z
1z 2x

d 2 h + 1 d 2 h , (16)
x pp ( x ) pp d x d y ( hpp hpp ) d x dz ( hpp hpp )

= d y d x ( hpp hpp ) d y hpp + (1 d y2 ) hpp
2
d d ( hpp hpp )
y z
d d h pp 2
+ (1 dz2 ) hpp
z x ( pp ) dz dy (hpp hpp ) dhz pp
where hpp is a short-hand notation for hpp (r) .

In summary, the Hamiltonian matrix in the tight-binding model consists of 44 blocks, in which
each atomic pair is assigned a block:
j


si s j si p jx si p jy si p jz
pix s j pix p jx pix p jy pix p jz
H = i , (17)
piy s j piy p jx piy p jy piy p jz
piz s j piz p jx piz p jy piz p jz



where the diagonal blocks (i = j) are diagonal,
Es 0 0 0
0 Ep 0 0
0 0 Ep 0 , (18)

0 0 0 Ep
and the off-diagonal blocks (i j) are given by
" h d x hsp d y hsp dz hsp %
ss
$ '
$ d h d x2 hpp + (1 d x2 ) hpp d x d y ( hpp hpp ) d x dz ( hpp hpp ) '
x sp
$ '
$ d h d y d x ( hpp hpp ) d y2 hpp + (1 d y2 ) hpp d y dz ( hpp hpp ) '. (19)
$ y sp '
$ '
$ dz hsp dz d x ( hpp hpp ) dz d y ( hpp hpp ) dz2 hpp + (1 dz2 ) hpp '
# &
Though two atomic basis functions on different atoms are not exactly orthogonal, we will ignore
their overlaps and treat the 4N-function basis set as an orthonormal basis set.

6
References
1. P. Hohenberg and W. Kohn, Inhomogeneous electron gas, Physical Review 136, B864 (1964).
2. W. Kohn and L. J. Sham, Self-consistent equations including exchange and correlation effects,
Physical Review 140, A1133 (1965).
3. For a review, see M. C. Payne, M. P. Teter, D. C. Allan, T. A. Arias, and J. D. Joannopoulos,
Iterative minimization techniques for ab initio total energy calculations: molecular dynamics and
conjugate gradients, Reviews of Modern Physics 64, 1045 (1992).
4. A. P. Sutton, Electronic Structure of Materials (Oxford University Press, Oxford, UK, 1993),
Chapter 6.
5. W. A. Harrison, Electronic Structure and the Properties of Solids (Dover, New York, 1989).
6. J. C. Slater and G. F. Koster, Simplified LCAO method for the periodic potential problem,
Physical Review 94, 1498 (1954).
7. M. L. Cohen and J. R. Chelikowsky, Electronic Structure and Optical Properties of Semiconductors,
2nd Ed. (Springer-Verlag, Berlin, 1989).
8. I. Kwon, R. Biswas, C. Z. Wang, K. M. Ho, and C. M. Soukoulis, Transferable tight-binding
models for silicon, Physical Review B 49, 7242 (1994).

7
Appendix AHamiltonian Matrix
Consider the energy eigenvalue problem,
H = . (A1)

As shown in Eq. (5), we expand the wave function, , in terms of atomic orbitals,
N
= ci i , (A2)
i=1 {s, px , py , pz }

where i is the -th atomic eigenstates ( {s, px, py, pz}) of the i-th atom. Substituting Eq. (A2) into
Eq. (A1) and multiplying the both sides by i , we obtain

c i i H i = ci i i . (A3)
i i

Using the orthonormality of the basis set, i i = ii , Eq. (A3) becomes

H c = ci ,
i ,i i (A4)
i

where the Hamiltonian matrix is defined as

1
H i,i = i H i = dr * (r ri ) 2 +V (r ) (r ri ) . (A5)
2
In the 4N4N matrix in Eq. (17), the indices i and are combined as = 4(i - 1) + , where i {1, 2,
..., N} and {1s, 2px, 3py, 4pz}. With this indexing scheme, Eq. (A4) becomes an ordinary
matrix eigenvalue problem,

H c = c ! .
! (A6)

8
Appendix BTrigonometric Addition Theorem
In the proof of Eq. (8), we have used the trigonometric addition theorem,
cos ( ) = cos cos + sin sin , (B1)
which may be derived from the following identity,
exp (i( )) = exp (i ) exp (i ) . (B2)
The left-hand side of Eq. (B2) is
exp (i( )) = cos ( ) + isin ( ) , (B3)
whereas its right-hand side is
exp (i ) exp (i ) = "#cos ( ) + isin ( )$%"#cos ( ) isin ( )$%
. (B4)
= "#cos ( ) cos ( ) + sin ( ) sin ( )$% + i "#sin ( ) cos ( ) cos ( ) sin ( )$%

Comparison of the real parts of Eqs. (B3) and (B4) gives Eq. (B1).