Quantum Transport:

Atom to Transistor

Prof. Supriyo Datta

ECE 659
Purdue University

01.29.2003

Lecture 7: Schrödinger Equation: Method of Finite Differences

Ref. Chapter 2.2

Network for Computational Nanotechnology

 00:00
Retouch on Concepts

Hydrogen Atom Discrete Energy

• Last time, we discussed the
Levels
Schrödinger Equation:
∂Ψ − h2 2
ih = ∇ Ψ + U (r )Ψ

…
-q
∂t 2m
+q n=4
• We began with the simplest

En = -E0/n2
case, the Hydrogen atom. It
n=3
is the simplest because there
are no electron-electron
interactions n=2
• Experimentally discrete
n=1
energy levels are observed

01.29.2003
 Method of Finite 03:00
Differences: Lattice Points

• Today: How to solve the Schrödinger • First create a lattice for the 1 -D problem.
Equation numerically using the method
of finite differences. Width spacing ?n
a
• Start with the 1 -Dimensional
Schrödinger equation:
?
∂? h2 ∂2
ih =− ? + U (r ) ? 1 2 3 n-1 n n+1 N
∂t 2m ∂x 2
 ? ( x1 ) 
• In essence we want to convert the • ? (x,t) therefore ?( x )
Schrödinger Equation into a matrix becomes a column  2

equation. vector telling the  ? ( x3 ) 
 
value of ? at ? ( x ,t) →  M 
different points  ? ( xn ) 
at any given instant of time:  
 M 
? ( x )
 N 

01.29.2003
Method of Finite Differences: 07:00
Hamiltonian Matrix

• Similarly, the entire 1 -D Schrödinger • Formulating Schrödinger’s Equation in

Equation becomes: the following form:

? 1    ? 1   d?   − h2 d 2 
 i h =  ? 
dt  x = xn  2 m ∂x
d   = Η   M 
ih M 
2
 x= xn
dt      
 ? n     ? n 
+ [U (x ) ? (x , t )]x = x n
• How do we convert the Hamiltonian
operator into a matrix? • The second term on the right side
equation can be written down easily and it
• First try writing the matrix for U(x) and is U ( x n ) × Ψnfor any given point ‘n’ of the
− h d
then the matrix for 2 2 lattice. Note that ? n only depends on ? n
and this indicates that potential term U(x)
2
2 m dx appears on the diagonal of Hamiltonian
The total Hamiltonian should be a sum of only.
these two

01.29.2003
Method of Finite Differences: 14:00
Hamiltonian Matrix

• Thus:
• First try to write
d ? …
 ? 1  U ( x1 )  ? 1   dx 
d     M 
ih  M  =  O  
dt d ?  ? −?
? n   U ( xn )  ? n   dx  = n +1 n

n +1 2 a
• Now, how do we write the second
d ?  ? −?
derivative at a particular point?
 dx  = n n −1

n −1 2 a
Points in the Lattice Space • Then,
d?  d ? 
 dx  −
?n d 2 ?  n +1 2  dx  n −1 2
 2 
=
 dx  n a
? − 2? n+?
a = n +1 n −1

a2
x Which is the representation of a second
1 01.29.2003
2 3 n-1 n n+1 N derivative in the finite differences method
Method of Finite Differences: 20:52
Hamiltonian Matrix

• Generalizing the numerical expression for the second derivative we get:

− h2
ih
d
? n = [2 ? n − ? n −1 −? n +1 ] + U ( xn ) ? n
dt 2ma 2
t0 = h
∂
2ma so the Hamiltonian matrix ih ∂t {? } = [Η ]{? }
2
• Let now looks like …

 ? 1   2t0 + U ( x1 ) − t0 0 0 0 0 ? 1 
?   −t 2t0 + U (x 2 ) − t0 0 0 0 ? 
 2  0  2 
d  M   0 − t0 O O 0 0  M 
ih   =   
dt  ? n   0 0 O O O 0 ? n 
 M   0 0 0 O O − t0  M 
    
? N   0 0 0 0 − t0 2t0 + U ( x N ) ? N 
01.29.2003
 Eigen Values 27:30
Eigen Vectors

• What about the time derivative • And similarly, so does any linear
combination of basis functions
d? i.e. How do we calculate
ih
dt
{? (t)} given some initial
state {? (0)}? {Ψ (t )} = ∑ C α e −iE αt h
{α }
α

• Answer: Find the eigenvalues Ea and • What is not clear, is what the coefficients
eigenvectors {a} of the matrix [H]. Such C a are. This depends on the problem
that, [H]{a}= Ea {a}

• Using the eigenvectors as a basis set, by

substitution it can be shown that {? (t)} =
{a}e(-iEat)/h satisfies the matrix form of the
Schrödinger Equation But first step is to
find the eigen value E α
And the eigen vector { }α
01.29.2003
 35:34
Boundary Conditions

• What do we do when we get • Unlike the infinite -wall scenario, periodic boundary
near a boundary? conditions have a slightly different Hamiltonian matrix:

• In the case of infinite -wall

boundary conditions, such as the 2t0 + U (x1 ) − t0 L L L − t0 
infinite square well,  −t 2t0 + U (x2 ) L L L M 
-t0 ? 0 + (2t0 + U 1) ? 1- t0 ? 2  0 
is replaced by  M M O M 
(2t0 + U 1) ? 1-2 ? 2 [Η ] =  
 M M O M 
and
-t0 ? N-1 + (2t0 + U N) ? N - t0 ? N+1  M M O − t0 
 
by  − t0 L L L − t 0 2t 0 + U (xN )
-t0 ? N-1 + (2t0 + U N) ? N
Meaning, ? 0 and ? N+1 are
considered equal to zero • Sub-note: Most Hamiltonians are
normally Hermitian:
• For periodic boundary +
conditions, such as a particle on a • H = (H*)T = H
ring, we let
? 0 = ? N and ? N+1 = ? 1
01.29.2003
 PERIODIC VS. BOX 42:36
Boundary Conditions

• For periodic boundary conditions • Whereas for BOX boundary conditions

eigenvalues come in degenerate pairs the eigenvalues are non degenerate

Degenerate Eigenvalues Non-Degenerate Eigenvalues

E E

(eV)
(eV)

a a
20 40 60 20 40 60

• This is because the eigenfunctions for • This is because the eigenfunctions for the
the periodic case are: this case are:
sin (ka = (a2p)/L) and cos (ka x) sin (ka x)
with ka = (a2p)/L, a = 1,2,… with ka = (ap)/L, a = 1,2,…
01.29.2003
 Numerical vs. Analytical 46:47
Eigenvalues

• Finally note that numerical analysis yields

eigenvalues close to analytical result only
• Where the analytical result is given by Ea
= (h2a 2p2) / 2mL2
for low energies.
• This occurs because at very high
• For example a 100 point lattice in the
energies the wave function oscillates very
infinite square well yields the following…
fast and the wide lattice spacing does not
capture the whole story.
E Numerical vs.
Analytic
10 20 30 40

High Energy Wave: Infinite Square

Well

8 8
(eV)

a 0 L
20 40 60 80 100
01.29.2003