This action might not be possible to undo. Are you sure you want to continue?
INFINITE POTENTIAL WELL
To solve the TISE for an infinite potential well shown on the figure below (Fig. 6), we go
back to the step potential and examine the solution in region 2 when E<V
0
(the evanes
cent wave) of the form:
x x x
e
E V
E
e
k
k
x e A
i k
k
x
2 2 2
0
2
2
2
1
1
2
) 1 (
2 1
1
2
2
2
) (
2
) (
κ − κ − κ −
−
·
κ +
ψ →
κ +
· ψ .
In the limit ∞ →
0
V , it is obvious that 0 ) (
2
→ ψ x for any value of x≥0. Therefore, we
can say that in this limit, the wavefunction vanishes at the boundary of the infinite poten
tial step. We can generalize this by saying that the wavefunction vanishes at 2 / L x t ·
for the infinite well problem.
V(x)
x
L/2 L/2
∝ ∝
Figure 6. Infinite potential well.
Notes:
Ø The solution of the TISE for this type of potential constitutes a boundstate problem.
A significant difference between continuum and boundstate problems follows from
the fundamental difference between the nature of these states, i.e.
(a) the type of boundary conditions that they satisfy:
0 ) , (
2
÷ ÷ ÷ → ÷ ψ
t∞ → x
t x at all times è bound states
0 ) , (
2
÷ ÷ ÷ → ÷ ψ
t∞ → x
t x at any particular time t è unbound states
(b) Energy spectrum:
bound states è discrete energy spectrum
unbound states è continuous energy spectrum
Ø A major difficulty in solving the TISE for bound states is that we do not know the al
lowed quantized energies in advance. We must find these by solving for the Hamilto
nian eigenfunctions that obey the appropriate boundary conditions:
( )
∫
· ψ ψ
t∞ → → ψ
∞
∞ −
1 ) ( ) (
, 0
*
dx x x
x x
E E
E
Solution to the particle in the box problem:
Since the particle is free inside the box, we can write the general solution to the TISE in
this region as:
) sin( ) cos( ) ( kx D kx C Be Ae x
ikx ikx
+ · + · ψ
−
For this wavefunction to be a proper wavefunction, it must satisfy the two boundary con
ditions at x=L/2 and x=L/2. The one at x=L/2 leads to:
¹
¹
¹
'
¹
·
π −
· →
π
− · ·
·
π
· → π · ·
⇒ · +
K
K
, 3 , 2 , 1 ,
) 1 2 (
2
) 1 2 (
2
, 0
, 3 , 2 , 1
2
2
, 0
0 ) 2 / sin( ) 2 / cos(
n
L
n
k n
kL
D
n ,
L
n
k n
kL
C
kL D kL C
n
n
The top solution leads to odd parity wavefunctions ) (x
o
ψ , and the bottom one to even
parity wavefunctions ) (x
e
ψ , of the form:
¹
¹
¹
¹
¹
'
¹
]
]
]
π −
· ψ
]
]
]
π
· ψ
x
L
n
L
x
x
L
n
L
x
e
o
) 1 2 (
sin
2
) (
2
cos
2
) (
Important:
The discussion presented in this section suggests that the wavevector of the particle in an
infinite potential well is quantized, i.e. the spatial functions that satisfy the boundary con
ditions imposed by the potential exist only for discrete values of the wavevectors which,
in turn, leads to discrete energy spectrum:
1
2 2
2
2 2 2 2
2
2
E n n
mL
m
k
E
L
n
k
n
n n
·
π
· · →
π
·
h h
,
where
1
E is the energy of the lowest allowed energy level, also known as the ground
state. The list of allowed stationarystate energies is called energy spectrum.
D. FINITE POTENTIAL WELL
Ø Consider the ground state of an infinite potential well. The energy corresponding to
this state equals to
2
2 2
1
2mL
E
π
·
h
Now, let's look at the final well solution to the TISE shown schematically on Figure 7.
Due to the finite height of the well, there is a spilling of the wavefunction in the clas
sicallyforbidden region. To first order, the spilling of the wavefunction can be ex
plained as having a wider well with effective width λ>L. This, in turn leads to a
lower value for the ground state of a finite well when compared to the infinite well
solution. In the limit, when the barrier height approaches zero and λ÷∞, we therefore
arrive at the continuum problem.
V(x)
x
L/2 L/2
V
0
Ground state E
1
for
finite well
First excited state E
2
for
finite well
Figure 7. A sketch of the wavefunctions for the lowest two energy levels for a finite
well.
Ø An additional observation that follows from this figure is that the ground state wave
function remains to be an even function, i.e. the parity of the states does not change.
Therefore, the even and odd solutions of the TISE have the following form:
¹
¹
¹
'
¹
≥
< < −
− ≤
· ψ
κ −
κ
2 / ,
2 / 2 / ), cos(
2 / ,
) (
L x Ae
L x L kx B
L x Ae
x
x
x
e
,
¹
¹
¹
'
¹
≥ −
< < −
− ≤
· ψ
κ −
κ
2 / ,
2 / 2 / ), sin(
2 / ,
) (
L x Ae
L x L kx B
L x Ae
x
x
x
o
where
2
2
h
mE
k · and
( )
2
0
2
h
E V m −
· κ .
Ø Due to the symmetry of the potential, we do not need to impose four boundary condi
tions at the two boundaries. The two boundary conditions at one of the two bounda
ries are sufficient to determine the energy levels corresponding to the even and odd
parity functions. This can further be simplified by introducing the single boundary
condition on the logarithmic derivative
[ ] [ ]
0 0
) ( ln ) ( ln
) 2 ( ) 1 (
x x x x
x d x d
· ·
ψ · ψ
In the above expression, ) (
) 1 (
x ψ and ) (
) 2 (
x ψ are the general solutions in regions 1
and 2 for some piecewise constant potential, and
0
x represents the boundary between
the two regions. It is straightforward to show that the equality of the logarithmic de
rivatives at the matching point reduces to the two conditions we have used for step
potentials, namely continuity and smoothness of the wavefunction at the matching
point.
Ø The application of the logarithmic boundary condition at x=L/2 leads to the follow
ing results for the even and oddparity functions:
(a) Even parity function:
( )
,
`
.

· κ
↓
,
`
.

− −
,
`
.

−
·
,
`
.
 κ
− κ
,
`
.
 κ
−
2
tan
2
sin
2
cos
1
2
exp
2
exp
1
kL
k
kL
Bk
kL
B
L
A
L
A
(b) Odd parity function:
( )
,
`
.

− · κ
↓
,
`
.

−
,
`
.

−
·
,
`
.
 κ
− κ
,
`
.
 κ
−
2
cot
2
cos
2
sin
1
2
exp
2
exp
1
kL
k
kL
Bk
kL
B
L
A
L
A
Discussion:
The condition on the allowed wavevectors for the even parity functions that are solutions
of the TISE can be rewritten as:
) ( 1 ) tan(
2
2
ξ · −
ξ
β
· ξ f ,
where
2
0
2
2
h
mV L
· β and
2
2
2
h
mE L
· ξ .
Similar expressions can be obtained for the odd parity functions, i.e. in this case one
needs to solve the implicit equation
) ( 1 ) cot(
2
2
ξ · −
ξ
β
· ξ − f .
The solution of these implicit equations can be obtained by using either the graphical so
lution method (see the figures below) or using, for example, the NewtonRaphson
method. An important conclusion that follows from the graphical solution method is that,
in the limit of infinitelysmall height of the well ( 0 → β ), there will be only one solution
to the TISE and the energy of that level will coincide with the top of the well. This is also
clear from the graphical solution to the problem.
4
2
0
2
4
0 1 2 3 4 5 6
tan(ξ)
(β
2
/ξ
2
1)
1/2
ξ=0.5*L*(2mE/hb
2
)
1/2
L=7.5 nm, V
0
=0.4 eV
m=6x10
32
kg
meV E 22 . 57 18 . 1 · → · ξ
(a)
4
2
0
2
4
0 1 2 3 4 5 6
cot(ξ)
(β
2
/ξ
2
1)
1/2
ξ
meV E 4 . 217 3 . 2 · → · ξ
(b)
Figure 8. The even and the odd solutions of a finite well problem. There are two bound
states in the well, one corresponding to the even function and the other to the odd func
tion. Case (a) corresponds to the even function solutions, and case (b) corresponds to the
odd solutions.
The graphical solution results for the bound states are 57.22 and 217.4 meV. The nu
merical solution to the same problem, using the scheme described in details in later sec
tion gives E
1
=57.3 and E
2
=218.5 meV, which is in excellent agreement with the graphical
analysis. The shape of the wavefunctions corresponding to these two energy levels is
shown in Fig. 9a. Also shown in this figure is the solution to the 1D TISE for an energy
value that is not an eigenstate (Fig. 9b). It is important to note that, even though the
wavefunction is continuous, it does not satisfy the smoothness condition (the wavefunc
tion does not have continuous derivative).
0.1
0
0.1
0.2
0.3
0.4
0.5
1.5x10
4
1x10
4
5000
0
5000
1x10
4
1.5x10
4
20 15 10 5 0 5 10 15 20
V(x)
ψ
1
(x)
ψ
2
(x)
V
(
x
)
[
e
V
]
ψ
(
x
)
[
m

1
/
2
]
distance x [nm]
(a)
0.1
0
0.1
0.2
0.3
0.4
0.5
0.015
0.01
0.005
0
0.005
0.01
0.015
20 15 10 5 0 5 10 15 20
Distance x [nm]
V(x)
Ψ(x)
(b)
E=0.15 eV
not smooth
Figure 9. (a) The shape of the eigenfunctions corresponding to the bound states of a finite
well. (b) Solution of the 1D Schrödinger equation for energy that does not correspond to
an eigenstate.
For this problem, one can also plot the transmission coefficient T(E) for particle energies
E>V
0
=0.4 V. Following the approach described for a step potential, with little bit more
algebra, one arrives at the following expression for the transmission coefficient T(E)
( )
( )
1
2
2
2
2
2
1
2
2
2
1
sin
4
1 ) (
−
]
]
]
]
−
+ · L k
k k
k k
E T ,
where
2
0 2
1
) ( 2
h
V E m
k
−
· and
2
2
2
2
h
mE
k · .
The variation of the transmission coefficient with energy is shown in Fig. 10. From the
results shown, it is clear that the transmission coefficients equals one for the following
condition:
2 2
2
2 2
2
1014 . 0
2
n n
mL
E n L k
n
·
π
· → π ·
h
,
which gives E
1
=0.1014, E
2
=0.4056, E
3
=0.9126 and E
4
=1.622 eV. These values are con
sistent with the results deduced from Fig. 10.
0.7
0.75
0.8
0.85
0.9
0.95
1
0.0 0.5 1.0 1.5 2.0 2.5 3.0
T
(
E
)
Energy [eV]
Figure 10. Variation of the transmission coefficient with energy, for the quantum well
problem from Figs. 8 and 9.
Useful information:
A Professor at ASU (Physics and Astronomy Department) whose name is Kevin Schmidt,
has created a web site on which you can find his Visual Schrödinger solver. This Web
based applet numerically solves the onedimensional Schrödinger equation for a variety
of standard Hamiltonians and permits users to define their own potential functions and
rapidly display the results. The applet has been designed primarily as a pedagogical tool.
The first time user may go directly to the applet and, with the brief introduction, try a few
typical problems. The Web address for the site is:
http://fermi.la.asu.edu/Schroedinger/html/Schroedinger.html .
The tool calculates the energy levels and the corresponding wavefunctions for finite
wells.
E. TRIANGULAR WELL SOLUTIONS
When an electric field is applied to the surface of a ptype semiconductor, such as occurs
in a metaloxidesemiconductor under bias, an ntype inversion layer is produced at the
surface (see Fig. 11). When the bands are bent strongly, as in strong inversion, the po
tential well formed by the insulatorsemiconductor surface and the electrostatic potential
in the semiconductor can be narrow enough that quantummechanical effects become im
portant. The motion of electrons in the direction perpendicular to the surface is con
strained to remain within this potential well and, if the thickness of the well is compara
ble to the electronic wavelength, size effect quantization leads to widely spaced electron
energy levels. These energy levels are then grouped into subbands, each of which corre
sponds to a particular quantized level of motion in the direction perpendicular to the sur
face. The same quantization effect occurs in quantum wells, heterostructures and super
lattices. In the thermodynamic limit, the electrons are described by plane waves in the
plane parallel to the interface. In MOS devices, such quantum effects play an important
role even at room temperature, where the width of the inversion layer is of the same order
of magnitude as the thermal wavelength of the carriers. Therefore, it becomes necessary
to solve the Schrödinger equation for the subband energies and wavefunctions in order to
study transport properties of inversion layer electrons.
The periodic crystal potential in the bulk of semiconducting materials is such that,
for a given energy in the conduction band, the allowed electron wavevectors trace out a
surface in kspace. In the effectivemass approximation for silicon, these constant energy
surfaces can be visualized as six equivalent ellipsoids of revolution (Fig. 11), whose ma
jor and minor axes are inversely proportional to the effective masses. A collection of such
ellipsoids for different energies is referred to as a valley.
E
F
V
G
>0
Ε
11
Ε
12
Ε
13
Ε
14
Ε
21
Ε
22
zaxis [100]
(depth)
∆
2
band :
m
⊥
=m
l
=0.916m
0
, m

=m
t
=0.196m
0
∆
4
band:
m
⊥
=m
t
=0.196m
0
, m

= (m
l
m
t
)
1/2
∆
2
band :
m
⊥
=m
l
=0.916m
0
, m

=m
t
=0.196m
0
∆
4
band:
m
⊥
=m
t
=0.196m
0
, m

= (m
l
m
t
)
1/2
∆
2
band
∆
4
band
Figure 11. (left) Potential diagram for inversion of ptype semiconductor. (right) Con
stantenergy surfaces for the conductionband of silicon showing six conductionband
valleys in the <100> direction of momentum space. The band minima, corresponding to
the centers of the ellipsoids, are 85% of the way to the Brillouinzone boundaries. The
long axis of an ellipsoid corresponds to the longitudinal effective mass of the electrons in
silicon
l
m , while the short axes correspond to the transverse effective mass
t
m .
In this framework, the 1D Schrödinger equation for an electron residing in one of
these valleys is of the form
) ( ) ( ) (
2
2
2 2
z E z z V
z m
i i i
i
ψ · ψ
]
]
]
]
+
∂
∂
−
⊥
h
where
⊥
i
m is the effective mass normal to the semiconductoroxide interface of the ith
subband from either the unprimed or primed ladder of subbands, and
i
E and
i
ψ are the
energy level and the corresponding wavefunctions. For <100> crystal orientation the six
equivalent minima of the bulk silicon conduction band split into two sets of subbands.
The first set (∆
2
band) consists of the two equivalent valleys with inplane effective mass
m

=0.19m
o
and perpendicular effective mass m
⊥
=0.91m
o
. The second set (∆
4
band) con
sists of the four equivalent valleys with m

=
t l
m m =0.42m
o
and m
⊥
=0.19m
o
. The energy
levels associated with the ∆
2
band comprise the socalled unprimed ladder of subbands,
whereas those associated with the ∆
4
band comprise the primed ladder of subbands.
To calculate the subband structure for the MOS capacitor, one needs to find the
selfconsistent solution of the Schrödinger equation, and of Poisson's equation
[ ] ) ( ) ( ) ( ) (
) (
1
z n z p z N z N e
z z z
A D
− + − − ·
]
]
]
∂
ϕ ∂
ε ∂
∂
− +
where ) (z ϕ is the electrostatic potential, ) (z ε is the dielectric constant, ) (z N
D
+
and
) (z N
A
−
are the ionized donor and acceptor concentrations, ) (z n and ) (z p are the elec
tron and hole densities. Note that the potential energy term appearing in the Schrödinger
equation is related to the potential ) (z ϕ in the following manner: . ) ( ) ( const z e z V + ϕ − ·
The boundary conditions that one needs to impose are the following ones: V(0) = 0 and
V'(∞)→0. In the analysis presented here, we assume that z=0 corresponds to the SiO
2
/Si
boundary. The quantummechanically calculated electron density is then given by:
) ( ) (
2
z N z n
i
i
i
ψ ∑ · ,
where N
i
is the sheetcharge density corresponding to energy level E
i
, and N
s
is the total
sheet charge density that is calculated using
]
]
]
]
,
`
.
 −
+ ∑
π
· ∑ ·
T k
E E T k m
N N
B
i F
i
B i
i
i s
exp 1 ln
2

h
When the Poisson equation is formally solved for this particular problem, one finds the
solution
( )
]
]
]
∫
ψ − − ∑
ε
+
,
`
.

−
ε
· + ·
z
i
i
i
sc sc
A
i d H
dx x x z z N
e
W
z
z
W N e
V V z V
0
2
2 2
) (
2
1 ) ( ,
for W z ≤ , where W is the width of the depletion region. The first term represents the
contribution from the fixed space charges whereas the second term represents the contri
bution from induced charges in the spacecharge region (the socalled Hartree term). The
width of the depletion layer is calculated to be
I
d sc
eN
W
φ ε
·
2
,
where
d
eφ is the band banding shown in Fig. 12, given by
( ) ( ) ∑
ε
− ε + ε − ε + ε − ε · φ
i
av
i i
sc
o o F
bulk
F c d
z N
e
e
) (
2
.
In the above expression,
z
i
( av)
is the average distance of the electrons that reside in the i
th subband from the interface,
ε
c
is the conduction band edge, and
ε
o
is the energy of the
lowest subband.
depletion region width w
eΦ
d
potential energy
z
ε
c
ε
F
ε
c
ε
F
Figure 12. Schematic bandbanding near the semiconductorinsulator interface, showing
the nominal conductionband edge (thick line) and the corresponding bandbanding asso
ciated with the fixed depletion layer charge only (thin line).
To obtain selfconsistent solution to this problem is a rather formidable task for students
at this level. Because of this, in the rest of this section we will only describe ways to find
approximate solutions. The explanation for the numerical solution of the 1D Schrödinger
equation, without coupling to the 1D Poisson equation, is described in the next section.
The simulation tool SCHRED used for obtaining numerical selfconsistent results for
MOS capacitors with ptype substrates and metal or polysilicon gates, developed at Ari
zona State University, and donated to the Purdue Semiconductor Simulation Hub, can be
accessed using the following link:
http://punch.ecn.purdue.edu:8000/ .
Follow this link to examine the capabilities of SCHRED. You need to read the user man
ual before you try to run SCHRED.
(1) Airy functions
If the inversion layer contribution to the potential energy is negligible compared to the
depletion layer contribution, for small values of z ( ) W z << one can approximate the po
tential energy profile by
z eE z
W N e
z V
s
sc
A
·
ε
≈
2
) ( ,
where E
s
is the electric field at the semiconductorinsulator interface (z=0). This is called
the triangularpotential approximation. It leads to the Airy equation of the form
0 ) (
2
2
· ξ ξψ −
ξ
ψ
d
d
, where
3 / 2
2
2
3 / 1
2
2
2 2
,
`
.

−
,
`
.

· ξ
s
s
meE
mE meE
z
h
h h
.
The solutions of this equation are Airy functions Ai(ξ), and the zeroes of the Airy func
tions are
L , 3 , 2 , 1 ,
4
1
2
3
3 / 2
·
]
]
]
,
`
.

−
π
− · ξ n n
n
which occurs for z=0. This means that the energy levels can be calculated using
L
h
, 3 , 2 , 1 ,
4
1
2
3
2
3 / 2
3 / 1
2
·
]
]
]
,
`
.

−
π
,
`
.

· n n
eE
m
E
s
n
.
The eigenvalues E
i
are asymptotic values for large i, but they are amazingly close even for
the ground state n=1. The exact eigenvalues for the three lowest states have ) 4 / 1 ( − n re
placed by 0.7587, 1.7540, and 2.7525, respectively.
Example:
Calculate the energy levels for a triangular potential well by using the Airy function
method. Use
2 12
10
−
· cm N
s
,
2 11
10 5
−
× · cm N
depl
,
0
8 . 11 ε · ε
sc
and
0
91 . 0 m m · .
Plot the first five energy levels.
Solution:
The surface electric field equals to:
) / ( 10 3 . 2 ) (
7
m V N N
e
E
depl s
sc
s
× · +
ε
·
The Airy function method gives:
) ( ) 4 / 1 ( 96 . 78
4
1
2
3
2
3 / 2
3 / 2
3 / 1
2
meV n n
eE
m
E
s
n
− ·
]
]
]
,
`
.

−
π
,
`
.

·
h
,
i.e., we have: E
1
=65.18 meV, E
2
=114.7 meV, E
3
=155 meV, E
4
=190.6 meV and E
5
=223.1
meV. Note that the separation of the levels decreases because of the linear increase of the
well width. The wavefunctions corresponding to these five energy levels are plotted in
Figure 13.
0
0.1
0.2
0.3
0.4
0.5
2x10
4
1x10
4
0
1x10
4
2x10
4
3x10
4
0 5 10 15 20
P
o
t
e
n
t
i
a
l
e
n
e
r
g
y
[
e
V
]
W
a
v
e
f
u
n
c
t
i
o
n
s
[
1
/
m
]
Distance [nm]
ψ
1
(x)
ψ
2
(x)
ψ
3
(x)
ψ
4
(x)
ψ
5
(x)
V(x)
Figure 13: Eigenfunctions corresponding to the lowest five energy levels of a triangular
potential well. Also shown in this figure is the shape of the confining potential.
(2) Variational Approach
The triangularpotential approximation is a reasonable approximation when there is little
or no charge in the inversion layer, but fails when the inversion charge density is compa
rable to or greater than the depletion charge density W N N
I depl
· . At low temperatures
and moderately high inversion charge densities and (100) orientation of the surface, only
the lowest subband of the two equivalent valleys with longitudinal mass perpendicular to
the interface is usually occupied. In this case, a variational approach gives a good esti
mate for the energy of the lowest subband. One can approximate the wavefunction of the
lowest subband with some trial function. The trial function proposed by Fang and How
ard
is
2 /
2 / 1
3
2
) (
bz
o
ze
b
z
−
,
`
.

· ψ
where b is a parameter which is determined by minimizing the total energy of the system
for given values of the inversion and depletion charges. The total energy per electron, the
quantity that needs to be minimized, is then given by
,
`
.

ε
+
,
`
.

ε
−
ε
+
,
`
.

· + + ·
b
N e
b
N e
b
N e
m
b
V V T N E
sc
s
sc
I
sc
depl
z
i d
16
33
2
1 6
3
8
/
2
2
2
2
*
2 2
2
1
h
The first term represents the expected value of the kinetic energy of the electron, while
the second and third terms correspond to the average potential energies of the electron
interacting with the depletion and inversion charge. The factor 1/2 prevents double
counting of the electrons. After some algebra, one finds that the value of b that minimizes
the average energy per electron equals to
3 / 1
2
* 2 *
12
,
`
.

ε
·
sc
z
N e m
b
h
where
s depl
N N N
32
11
*
+ ·
The energy of the lowest state is found after a straightforward calculation to be
i d o
V V T + + · ε
and the inversion charge contribution to the potential energy (Hartree term) reduces to
( ) ∑
]
]
]
,
`
.

+ − −
ε
·
− −
i
bz bz
i
sc
i
bz ze e
b
N
e
z V
2
1
2 1
3
) (
2
F. Numerical solution of the 1D Schrödinger equation
There are various ways in which one can solve the eigenvalue problem numerically.
The
simplest one is the shooting method that will be explained in the context of the example
shown in Fig. 14 representing an arbitrary potential energy profile. For simplicity, one
usually assumes that the potential becomes infinite at some points
min
z z · and
max
z z ·
which means that the boundary conditions for the solutions of the Schrödinger equation
are
ψ( z
min
) · ψ( z
max
) · 0. The proper choice of these limit points depends upon the par
ticular problem that one is solving. For example, in the silicon inversion layer,
z
min
· 0,
whereas
z
max
can be any arbitrary point in the depletion layer that is far enough from the
interface, so that we can assume with certainty that the wavefunction has vanished.
The idea of the shooting method is as follows. If one is looking for a bound state,
it starts with some small trial eigenvalue ε . Upon integrating toward larger z from
z
min
,
one can generate a solution ψ
<
(z), which increases exponentially through the classically
forbidden region and then oscillates beyond the left turning point in the classically al
lowed region. The integration becomes numerically unstable if one continues the integra
tion past the right turning point, due to the admixture of the undesirable exponentially
growing solution. Therefore, for each energy value, one generates a second solution
ψ
>
(z), by integrating from
z
max
to some smaller values. To determine whether the energy
that one chooses at the beginning is an eigenvalue, one has to compare ψ
<
(z) and ψ
>
(z)
at some matching point
z
m
. Since both ψ
<
(z) and ψ
>
(z) satisfy a homogeneous equa
tion, their normalization can always be chosen so that the two functions are equal at
z
m
.
An eigenvalue is then signaled by equality of the derivatives at the matching point. If one
approximates the derivatives by their simplest finitedifference approximations, discussed
later in the text, an equivalent condition is that
f ·
1
ψ
ψ
<
(z
m
− h) − ψ
>
(z
m
− h)
[ ]
· 0 ,
where h is the stepsize. The quantity ψ can be either the value of ψ
<
at
z
m
, or the
maximum of ψ
<
or ψ
>
. If this condition is satisfied, then one has found the eigenvalue.
Otherwise, one needs to increase the energy and repeat the previously mentioned steps.
The use of a bisection method can be very effective in fast location of the eigenvalues.
V(z)
ψ(z)
z
min
z
max
z
m
z
z
ψ
<
(z) ψ
>
(z)
(a)
(b)
ε
Figure 14. (a) Schematic potential used in the discussion for the shooting method. (b)
Form of the solutions ψ
<
and ψ
>
at an arbitrary energy. The left turning point is used as
a matching point. When ε is an eigenvalue, the derivative is continuous at the matching
point.
In the remaining part of this section, we explain how one can numerically inte
grate the 1D Schrödinger equation. If some arbitrary function is known at some equidis
tant node points (as shown in Figure 15), one can use Taylor series expansion to represent
the first and second derivatives of this function in terms of the values of the function at
the node points.
h x 2
0
− h x −
0 0
x h x +
0
h x 2
0
+ h x 3
0
+
i 1 + i 2 + i 3 + i 1 − i 2 − i
node
Figure 15. Onedimensional equidistant mesh.
The procedure for obtaining the finitedifference representation of the first and second de
rivatives on a uniform mesh is outlined below:
• Use Taylor series expansion of an arbitrary function f(x):
i
i i
i
i i
x x
x x
x
f
h f f
x
f
h f f
h O
x
f
h x f h x f
h O
x
f
h x f h x f
∂
∂
− ≈
∂
∂
+ ≈
→
+
∂
∂
− · −
+
∂
∂
+ · +
−
+
·
·
1
1
2
0 0
2
0 0
) ( ) ( ) (
) ( ) ( ) (
0
0
In the above expressions h is the mesh size, and ) (
2
h O is the truncation error. On
the right, we have used a shorthand notation for the node values. Subtracting the
second expression from the first one gives the centered finite difference expression
for the first derivative at node point i:
h
f f
x
f
i i
i
2
1 1 − +
−
·
∂
∂
• Using the same arguments, we can also derive the finitedifference approximation for
the second derivatives:
2
1 1 1 1
'
2 / 1
'
2 / 1
2
2
2 1
h
f f f
h
f f
h
f f
h h
f f
dx
df
dx
d
dx
f d
i i i i i i i i i
i
i
− + − + − +
+ −
·
]
]
]
−
−
−
·
−
·
,
`
.

·
• Once we have the finitedifference approximation for the second derivative, we can
utilize this result to obtain the finitedifference approximation of the 1D Schrödinger
equation:
0 ) ( ) ( ) ( ) ( ) (
2
2
2
2
2
2 2
· ψ +
ψ
→ ψ · ψ +
ψ
− x x k
dx
d
x E x x V
dx
d
m
h
,
where
[ ] ) (
2
) (
2
2
x V E
m
x k − ·
h
.
A simple substitution of the previously derived result for the second derivative gives:
( ) 0 2
1
2 2
1
· ψ + ψ − + ψ
− + i i i i
k h .
• Having derived the finitedifference approximation for the 1D TISE, we can now
utilize the forward and backward integration schemes, to get
>
ψ and
<
ψ :
(a) forwardintegration scheme:
( )
1
2 2
1
2
− +
ψ − ψ − · ψ
i i i i
k h
(b) backward integration scheme:
( )
1
2 2
1
2
+ −
ψ − ψ − · ψ
i i i i
k h
• The initialization of the wavefunctions goes as follows: If x=0 (i=0) and x=x
max
(i=N)
are the boundary points of our integration domain, at which we can assume with cer
tainty that the wavefunction has vanished, then
0 , 0
0
· ψ · ψ
N
.
One also needs to use some small values at the neighboring node points (i=1 and
i=N1), such as
9
1 1
10
−
−
· ψ · ψ
N
to ensure that we get the nontrivial solution. When performing the numerical inte
gration from the left and from the right, to avoid overflow and underflow, one has to
renormalize the wavefunctions if they exceed some predetermined maximum value.
• The eigenvalue is signaled if, for a given energy E, the second derivative is continu
ous, i.e. the function
[ ]
< >
− −
ψ − ψ
ψ
·
1 1
1
m m
i i
i
h
f
goes to zero (this signals smoothness of the wavefunction). It is important to point
out that before one evaluate f, it is necessary to match the solutions at the turning
point i
m
, to ensure continuity of the wavefunction. The energy variation of f is shown
in the figure below. These results were obtained for a finite quantum well from Fig
ure 8 (L=7.5 nm, V
0
=0.4 V and m=6×10
32
kg). Notice that the zero's of f coincide
with the values for the energy levels that we obtained using the graphical solution
method.
Important note:
The graphical solution method gives only the eigenvalues for a finite square well,
whereas the numerical method gives us simultaneously the eigenvalues and the corre
sponding eigenfunctions for an arbitrary confining potential.
0.3
0.2
0.1
0
0.1
0.2
0.3
0 50 100 150 200 250 300 350 400
f
(
c
h
e
c
k
o
n
s
m
o
o
t
h
n
e
s
s
)
Energy [meV]
L=7.5 nm
V
0
=0.4 eV
m=6x10
32
kg
Figure 16: Energy dependence of the function f for a finite potential well, that checks the
smoothness of the eigenfunctions. The zero's of f denote the energy eigenvalues for the
finite potential well from Figure 8.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.