Quantum Mechanics Project I, Spring 2012

FY521 Project
By
Daniel Bgh Drasbk (090190)
&
Rasmus Rexbye Hansen (160889)
February 27, 2012
1
Contents
1 Introduction 3
2 The Schrodinger equation 4
2.1 The Stationary Schrodinger Equation . . . . . . . . . . . . . . . 4
2.2 The dimensionless Schrodinger Equation . . . . . . . . . . . . . . 7
2.3 Restrictions and geometric behavior of wave functions . . . . . . 9
2.4 Specic examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 The WKB approximation . . . . . . . . . . . . . . . . . . . . . . 15
3 Numerical investigations by the shooting method 17
3.1 Procedure of the shooting method . . . . . . . . . . . . . . . . . 17
4 Properties of stationary bound state wave functions and the
variational method 19
4.1 Properties of the Hamilton Operator . . . . . . . . . . . . . . . . 19
4.2 The variation method . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3 Analytic example of the variation method . . . . . . . . . . . . . 21
5 Numeric results from the variation principle 24
6 Summary 24
7 Appendix 25
.1 dimensionless SE . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
.2 The shooting method . . . . . . . . . . . . . . . . . . . . . . . . . 25
.3 Test functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2
1 Introduction
In this project we try analyze the dierent properties of the Schrodinger equa-
tion. In doing so, we try to obtain a good idea of some of the basic properties
of quantum mechanics. We start out by seeing how quantum mechanical sys-
tems dier from classical mechanics, and work our way through the essentials
of Schrodingers equation of motion. This is illustrated by a few examples. Fur-
thermore, we use dierent procedures to work our way around the numeric
diculties which arise, when attempting to investigate energies of a quantum
mechanical system.
3
2 The Schr odinger equation
2.1 The Stationary Schr odinger Equation
Let us rstly consider the reduced Plancks constant; . It is a natural constant
governing the relationship between the energy of a photon and its frequency.
The Planck relation is written as:
E = h
E being the energy, h = 2 and the frequency. The units for is [Js],
which corresponds to the units of angular momentum or the action integral
over the Lagrangian. From Einsteins Special Theory of Relativity, we know
the relativistic energy of an elastic collision, for particles with zero mass, can be
written as below, and from the classic relation between wavelength, frequency
and velocity, it is possible to obtain the De Broglie relation:
E =
_
(pc)
2
+m
2
c
4
= pc (1)
= c
E = h = pc
h = p
=
h
p
This relationship shows how the particle-wave duality of elementary particles is
related by the linear momentum(p) and the wavelength (). One can see, that
if the momentum (p = mv) is high, the wavelength becomes arbitrary small,
which is in consistence with the classical approach.
Schrodingers wave equation consists of two parts; a time dependent part and
a time independent part. Here we break up the equation of motion into two
parts, each governing their respective variable.
i
(r, t)
t
=

2
2m

2
(r, t)
r
2
+V (r)(r, t)
i

t
T(t)(r) =

2
2m

2
r
2
T(t)(r) +V (r)T(t)(r)
i
T(t)
t
1
T(t)
=

2
2m

2
(r)
r
2
1
(r)
+V (r)
Here we have separated the variables. When varying the dierent components
one sees that in order for this to still hold, each side must be constant (which
4
we will denote E), so:
i
T(t)
t
1
T(t)
=

2
2m

2
(r)
r
2
1
(r)
+V (r) = E
E = i
T(t)
t
1
T(t)
E =

2
2m

2
(r)
r
2
1
(r)
+V (r)
ET(t) = i
T(t)
t
E(r) =

2
2m

2
(r)
r
2
+V (r)(r)
Thus, we can obtain two ordinary dierential equations, which both can be
solved the trivial way. First we will consider the solution for the stationary
Schrodiner Equation (SE) with zero potential, then the solution of the time
dependent SE. The stationary SE will have no time dependence, since one is
only looking at how the wave packet behaves. The general solution will be:
E =

2
2m

2
(r)
r
2

2
(r)
r
2
=
2mE

2
(r)
2
(r)
(r) = Ae
ir
+Be
ir
Setting B = 0 is equivalent to a wave moving in the positive direction of r.
One must remember that the exponential solution really consists of cosine and
sine terms, specied by the restrictions on the specic wave equation. Thus, a
general solution to the stationary SE could be;
(r) = Ae
ir
Where the amplitude (A) is determined by normalization. Normalization en-
sures that the inner product of amounts to 1. Now, considering the solutions
to the time dependent part of the SE:
ET(t) = i
T(t)
t
(2)

T(t)
t
=
E
i
T(t)
T(t) = e
i
Et

We can combine the two solutions obtained for each part of the SE, to one
unique solution;
(r, t) = (r)T(t) = Ae
ir
e
i
Et

= Ae
i(r
Et

)
5
Now, let us look at the probability density in order to prove time independence,
but rst let us ensure that our wave equation is correct. We start by ensuring
normality;
_
(r)

(r)dr =
_
Ae
ir
Ae
ir
dr
= r|A|
2
= 1
|A| =
1

r
(r, t) =
1

r
e
i(r
Et

)
Now that we have normalized the wave function, we want to prove the time
dependency of the function, from the probability density:
|(r, t)|
2
= |(r)T(t)|
2
= (r)

T(t)

(r)T(t)
= (r)

_
iE

_
(r)
_

iE

_
= (r)

(r) = |(r)|
2
This means that the probabilities of states are all time independent.
When comparing two wave equations of dierent states, one nds that they will
be mutually orthogonal. We have left out this calculation here, but one would
nd that the amplitude of one state has to be zero, which means that no wave
function can exist in the same state. The property of orthonormality may be
written in terms of the Dirac delta function:

m
|
n
=
mn
=
_
0 if m = n
1 if m = n
This completeness of the possible states, and a clear denition of state, are dis-
cussed in detail later on, here it is only noted as a property of the wave function.
So, since the states of the wave function are orthonormal, the only thing gov-
erning the probability of a particle being in the dierent states, is the expansion
coecients. These coecients ensure both linearity and probability. The ex-
pansion coecients can be thought of as a probability amplitude of a state.
Together with its complex conjugate it gives the probability of a particle being
at a state, as shown above.
Another approach to determine linear independence is by the Wronskian Matrix.
We know that there exists two types of states. The bound state (E < V (),
E < V (+)) restricting the wave equation to a nite spatial region, such as
the stationary SE, and the scatter states (E > V () and/or E > V (+)),
which are not restricted.
For the scatter state the are two solutions, one for r < 0 and one for r > 0. For
the bound state there is only one solution, since the function at are equal.
6
We write the Wronskian for the scatter state as:
W(r) =

1

2
1
r
2
r

=
1

2
r

2

1
r
= 0

dW
dr
=
_

1
r

2
r
+
1

2
r
2
_

2
r

1
r
+
2

1
r
2
_
=
1

2
r
2

2

1
r
2
=
1
_

2mE

2
_

2
_

2mE

2
_

1
= 0
The same can be done for a bound state, and again we nd
dW
dr
= 0. Therefore
we can proclaim that it is linearly independent.
2.2 The dimensionless Schr odinger Equation
For us to make numerical calculations we need to have a dimensionless SE.
This is because a program like MatLab cannot calculate with the necessary
unit/dimensions. The standard time independent SE is given by:

2
2m
d
2
(r)
dr
2
+V (r)(r) = E(r)
This form of the SE is convenient to do analytically calculations. For us to nd
an dimensionless SE, we have to rewrite the variables to dimensionless variables,
which is done by dening the following:
x =
r
L
, v(x) =
V (xL)
E
, =
E
E

, (x) = (xL)

L
We now substitute this into the SE:

2
2m
d
2
(xL)
dxL
2
+V (xL)(xL) = E(xL)
And since we know that if you dierentiate with respect to a variable multiplied
by a constant, is the same as dierentiating with respect to the variable and
then afterwards multipling with the constant. Thus we can rewrite
d
2
(xL)
d(xL)
2
=
1
L

1
L
d
2
(xL)
dx
2
we therefore now have:

2
2mL
2
d
2
(xL)
dx
2
+V (xL)(xL) = E(xL)

2
2mL
2
d
2
(x)/

L
dx
2
+v(x) E

(x)/

L = E(x)/

1
E

2
2mL
2
d
2
(x)
dx
2
+v(x)(x) =
E
E

(x) = (x)
7
We now want the term
1
E

2
2mL
2
to be equal to 1. If it does, we would have the
dimensionless SE we have been trying to nd. For this to happen we can easily
see that E

=

2
2mL
2
. Thus we now have the relation between E

and L:
E

=

2
2mL
2
(-4)
And example of how the dimensionless SE could be used, could be a nite square
well potential, where the wells width is given by 2a. Here we have r = 2a and
we can not determine what L should be equal to. The easiest way is to set
L = a, and we therefore have x =
r
L
=
2a
a
= 2. If we write the dimensionless
SE we know everything except the potential function. But from what we know
we can determine its value at certain points:
v(x) =
_
_
_
v
0
if < x < 1
0 if 1 x 1
v
0
if 1 < x <
A sketch for the nite square well can be seen in the appendix:
We could have done the exact same sketch for the innite square well potential,
the only dierence would be that v
0
would be substituted by .
We know from theory that a harmonic potential can be described by V (r) =
1
2
m
2
r
2
If we would like to calculate the harmonic potential, we need to nd
v(x) in the dimensionless SE. We can do this in the following way:
V (r) =
1
2
m
2
r
2
V (xL) =
1
2
m
2
L
2
x
2
We know from what we postulated earlier, that we can describe v(x) = V (r)/E

.
We will further try to use as the unit for E

, since we know that it has the

same dimension as energy. We therefore have:
v(x) =
V (r)
E

=
m
2
L
2
x
2
2
=
x
2
L
2
m
2
We now substitute L
2
from our earlier denition of E

and afterwards (again)

put E

= :
v(x) =

2
2mE

x
2
m
2
=

2
x
2
m
2m
2

=
1
4
x
2
This means that instead of calculating a new potential in the dimensionless SE
every time, we can just use this value, as long as we know that the potential is
harmonic. The basic unit here is [E

] =
Another example on where we can use the same value for the potential every
time, is for the hydrogen atom. If we use the Bohr radius for the length unit.
The non-dimensionless SE would have the Coulomb potential, known from elec-
tromagnetism: V
c
(r) =
e
2
40
1
r
. This means that we now have:
L = a =
4
0
e
2

2
m
, V
c
(xL) =
e
2
4
0
1
xL
8
We can now nd the dimensionless potential for the hydrogenatom, using the
denition above as a dention of E

. We thereby get that v(x) =

2
x
. Here we
use the basic unit for the length as [L] =
40
e
2

2
m
Both E

and L are independent, but not of each other. This means that you
can chose E

> 0 or L > 0 but not both since they are related by;
E

=

2
2mL
2
You can go from the dimensionless SE back to the normal stationary SE by
using the following:
x =
r
L
, v(x) =
V (r)
E

, =
E
E

, (x) = (r)

L
By substituting this into the dimensionless SE you will come back to the ordi-
nary time independent SE.
2.3 Restrictions and geometric behavior of wave functions
In this section, we would like to examine the geometrical behavior of the wave
functions. We choose to use the dimensionless SE, which makes it easier to do
the analasys. We will rst examine whether

is continuous or discontinuous
in a point where v(x) is respectively continuous, and has a nite discontinuity.
This means that we should integrate the SE from x to x +:

d
2
(x)
dx
2
+v(x)(x) = (x)
_
x+
x
d
2
(x)
dx
2
+v(x)(x)dx =
_
x+
x
(x)dx

d(x +)
dx

d(x )
dx
=
_
x+
x
(v(x) )(x)dx
We know that this is continuous if lim
0
on the lefthand side is equal to zero,
which means that the two points are the same. It would not be continuous if
this does not hold, since the points will not meet.
We now want to examine three forms of potential:
v(x) is continuos
v(x) is discontinuos
v(x) described by v
0
(x)
In the rst case when 0 it is easy to see that the areal underneath the graph
goes toward 0. It is therefore trivial that

is continuous. In the second case,

we can see that for 0 the areal does not tend toward zero.
If we split up the integral to the following;
_
x
x
(v(x) )(x)dx +
_
x+
x
(v(x) )(x)dx
9
It becomes clear, that if we let goes to 0, we see that both the left and the
right expression equals 0 and therefore

is continuous.
In the third case we have the potential described by Diracs delta function. We
know the following about the Dirac delta function:
(x) =
_
0 if x = 0
if x = 0
Further we know that
_

(x)dx = 1 and that

f(x)(x a) = f(a)(x a) and
_

f(x)(x)dx = f(0)
We know that x takes on a value, at the point where there will be discontinuity
and therefore setting x = 0, the integration of the SE becomes:
_

(v
0
(x) )(x)dx =
_

v
0
(x)(x)dx +
_

(x)dx
= v
0
_

(x)(x)dx
_

(x)dx = v
0
(0)
_

(x)dx = v
0
(0)
The integral in the second last term is 0, due to the same argumentation as in
the rst two cases. We can therefore proclaim that

, is discontinuous as long
as (0) = 0.
We now want to examine for a potential being symmetric around x = 0, This
means that v(x) = v(x), so the SE(x) must equal SE(-x):
d
2
(x)
dx
2
= (v(x) )(x) = (v(x) )(x)
We can therefore deduce that (x) = (x). We now use that we know that:
_

dx = 1
_

| (x) |
2
d(x) = 1
_

| (x) |
2
d(x) = 1

| (x) |
2
d(x) = 1
_

| (x) |
2
dx = 1

2
_

| (x) |
2
dx = 1
2
_

dx = 1
So, must by all means equal 1, if this has to hold. Which leads to (x) =
(x); thus it either an even or odd function.
10
Rewriting the stationary SE in the form below, makes the geometric approxi-
mations easier too study;
d
2
(x)
dx
2
= (v(x) ) (x)
The second order derivative can be thought of as an innitesimal change within
the slope of the wave function (How

(x) curves, that is). This is of course

dependent of the states energy relation between v and . Below is a sketch of
how the function could look, for various values of v compared with .
There can be no stationary state with energy less than the minimal potential
energy. If this is the case, the energy state will tend to . This is illustrated in
the gure above, where the potential is set too zero and to be less than zero;
this results in a positive value for the wave function no matter the values of x,
since < 0 for a bound state.
It should be noted here, that v > and v < on the gure, are only a picture
of what happens, when the position dependent potential varies. One has to
remember that the potential is a function of x and the energy of the specic
state is a constant. Should > v = 0, it will correspond to a particle moving
freely without inuence.
A useful approximation, for understanding how the energy interaction works is,
that the more potential a function has, the harder it is to move it away from its
equilibrium point, or better yet; if the state of the wave equation has a high
potential, the harder it is to excite it from its ground state.
Another restriction on the wave function comes from Heisenbergs uncertainty
principle. It states that the more you know about a particles position, the less
you know about its momentum. There are two ways this can be illustrated:

p


2
and
E

t


2
The standard deviation of position and momentum are intertwined by this for-
mula, so if one knows the almost exact position, the standard deviation of
momentum will become proportionally large. A clever way around this uncer-
tainty is to, for example, trap the particle in a box of some specic size, then
11
assuming the box contains the wave, then it will be easier to avoid violating this
principle.
2.4 Specic examples
One way to illustrate the energy interaction, is by trapping the particle in a
potential well, by setting restrictions on the potential. This is done by stating
that the potential is innity, except in a small region where it will be zero. But
since this is a rather deep well, we would like to obtain a nite well and see how
the energies of the two dier. so we have;
v
0
(x) =
_
0 if 0 < x < a
v
0
if elsewhere
v

(x) =
_
0 if 0 < x < a
if elsewhere
Notice, that the wave function on the boundaries is zero. If this was not true, we
would not have trapped the entire wave function and the outcome possibilities
might be wrong. If one looks at the equation for a stationary state, it is easy
to see what happens if the potential is innite of nite. The oscillations of this
state is depend on the forms of energy at a specic point. Should the potential
be innite, we have v > , and so on. The graph above shows how part of the
wave functions behaves for the dierent values of and v.
For instance, one could dig a hole in the bottom of the quantum well, which
will decrease the potential of the wave function, and again we can consult the
graph for how the function would act, when comparing the potential energy at
a specic point, with the energys state. If we dug a hole in the quantum well,
it would lead to an increase in the

, where the hole was dug. This is because

if v(x) 0 we have:
d
2

dx
2
=
But remembering that < 0 for this bound state, we have
d
2

dx
2
> 0.
If we for instance, have an innite potential well with walls positioned at; x =
0 and x = 2, we can obtain the wave function by some of the restrictions
mentioned above. We know that v(x) = 0 in this interval, and that the wave
function must be zero at the end points.
d
2

dx
2
=

(x) = Asin(x) +Bcos(x)
(0) = (2) = 0 B = 0
(x) = Asin(

x)
12
We know that in the interval, L = {0 < x < 2}, and will be a -periodic
function, so we can break it into positive integers(n) in this interval:
(x) = Asin(L) = Asin(n)
L = n
=
n
2

2
L
2
L=2
=
n
2

2
4
(x) = Asin
_
n
L
x
_
Since the potential is zero here, the only energy can come from the momentum,
with the relation = p
2
. So for the product p, we kan obtain the dimensional
units;
p =

2mE = 2
=
2

and E =

2

2
2m
p =
2

_
2m

2
2m
= 2 = h
Which is the De Broglie relation, so the units of the product will be the same
as Plancks constant; [Js].
Now, assume we have a potential, that is zero everywhere, except in a sin-
gle point. This can be thought of as Dirac-delta function, with some initial
potential (v(x) = v
0
(x)). We know that for a bound state, we have < 0.
Substituting this into the stationary wave function above, one obtains:
d
2
(x)
dx
2
= (v
0
(x) +) (x)
l
2
(v
0
(x) +)
(x) =
_
_
_
Ae
l(x)x
for x < 0 B = 0
Be
l(x)x
for x > 0 A = 0
A for x = 0 A = B
Choosing the potential to be at x = 0, which we have done, is to be able
to eliminate every where else, since (x = 0) 0. Now we need to ensure
orthonormal amplitudes of the bound state. A lot of terms can be neglected,
since (x) is discontinuous at x = 0. We assume that (x) is symmetric around
zero, so below zero the wave function increases and above it decreases. This gives
us an opportunity to simplify the problem. Knowing that is discontinuous at
x = 0, we include a small real number () and let it tend to zero to be able to
13
evaluate the integral.
d
2

dx
2
= [v(x) ]

_
x+
x
d
2

dx
2
dx =
_
x+
x
[v
0
(x) +]dx

d
dx
(x +)
d
dx
(x ) =
_
x+
x
v
0
(x)
_
x+
x

Setting x = 0

d(0)
dx
+

d(0)
dx

= lim
0
_

_
+

v
0
(0) +
_
+

(0)
_
l(0) l(0) = v
0
(0)
2l = v
0
= l
2
=
_
v
0
2
_
2
The rst order dierentials, on the lefthandside, corresponds to the dierential
of the two solutions if x = 0. It is crucial to keep in mind that the potential
only exists at x = 0. Elsewhere the inclination and decent only depend on .
Now, consider the potential square well, with width 2a, the potential may be
written as;
V (x) =
_
0 for |x| > a
v
0
for a x a
We derived the solutions to the wave equation, to be the following:

even
(x) =
_
_
_
Fe
x
for x > a
Dcos(lx) for a > x > 0
(x) for x < 0

odd
(x) =
_
_
_
Fe
x
for x > a
Csin(lx) for a > x > 0
(x) for x < 0
Here F, D and C are arbitrary constants, with
_
2mE

2
and l
_
2m(E+V0)

2
.
From conditions of continuity and some algebra, it is possible to nd the con-
nection of allowed energies, between and l to be:
= ltan(la)
14
Now dening k la and k
0

a

2mV
0
, we can obtain an easier expression for
how the energy acts if is even or odd, which is represented below;

even
tan(k) =
_
k
2
0
k
2
1

odd
tan(k) =
1
_
1
k
2
0
k
2
The two points which have been marked by a red circle, are the two points in
which the equation for the even bound state energies are fullled. And therefore
we have boundstates in these points. The graph of the odd wave function have
been excluded in this report.
2.5 The WKB approximation
This type of approximation is useful when attempting to calculate solutions to
partial derivatives dependent on a small parameter. When used in quantum
mechanics the WKB technique can be applied to the stationary SE to obtain
approximate solutions to it. The general formula for this approximation is;
(x)
C
_
p(x)
e

p(x)dx
p(x)
_
2m[E V (x)]
15
Where V (x) is approximated to be constant, or slowly varying in comparison
to the wavelength. The constant C includes all constants that may arise from
the calculations. The momentum, is assumed to be real, as in the classic region,
so E > V (x). However might not be real. The approximation comes from
assuming that the amplitude (A) is slowly varying, so:
A

A
<<
p
2

2
.
Underneath we derive the WKB for the harmonic potential analytically. We
know that the potential is given by v(x) =
1
4
x
2
for the dimensionless SE. We
further know from theory that:
p L = n = nh

2

_
L
0
pdx = nh

2
So we may now make a semi classical approximation:
E =
p
2
2m
+V (r) p =
_
2m(e V (r))
We now put this into our integral and evaluate it to obtain:
_
L
0
pdx =
_
L
0
_
2m(E V (r))dx = nh

2
This is only valid for the non-dimensionless SE, with a potential equivalent to
that of a square well with an innite potential. For a dimensionless SE it states
the following:
p(x) =
_
v(x)
In classical physic we can nd the turning points x
+
and x

so in our case with

the harmonic potential we have that:
=
1
4
x
2
x

= 2

Thus we have the integral:

_
x+
x
_

x
2
4
dx =
_
_
x
2
+
4

x
2

4
=
1
2
_
x+
x+
_
x
2
+
x
2
In order to do this calculation, we must substitute,
x = sinx
+
dx = x
+
cosd
and plug it back in, to obtain:
_
2
0
x
2
+
_
1 sin
2
cosd =
_
2
0
x
2
+
cos
2
d =
x
2
+
2
_
2
0
d
x
2
+
1
2

2
=
x
2
+

4
16
And since we know that =
x
2
4
, we can rewrite the expression as:
x
2
+
4
=
So, by using the integral we determined in the start, we can nd the same value
that the WKB program does. That is;
=
_
n
1
2
_
= n
1
2
3 Numerical investigations by the shooting method
In this section we describe how to use the dimensionless SE to make numerical
calculations of the stationary SE. A Maple le was created for us by J. Boiden
Pedersen, in which the basic functions was already described and a library had
been made. So the general setup for our calculations was pre-made. We shall
described the numerical calculations for the following four potentials: Finite
Square Well, Harmonic Oscillator, Triangle Potential and Coulomb Po-
tential. Further more, we shall use the WKB approximation on some of these
potentials.
3.1 Procedure of the shooting method
First of we dene the potential as being V = 50 (1 Box(x, 1, 1)). Where
50 declares the depth of our well, while the Box command describe how wide
our well is. We start out by doing the shooting method to nd the interval in
which the ground state lies. The shooting method needs to know the potential,
the energy in which it should investigate, the maximum value of x, how many
calculations it must do and if it should look for an odd or even wave function.
If we set the energy to high we see that (x) and if we set it to low then
(x) and since this is not a possible solution we can discard these values
for our bound-state energy, but we now know in which interval to search.
Then using the AutoShooting method in maple; it does the exact same thing as
the shooting method, except it searches for the solution in a specied interval.
Its output is a graph with the mean energy of the wave function. If you write
true in the end of the input, one will also obtain the probability density of the
function. We now repeat these steps until we have the rst four excited states
and the ground state. We know that the ground state is an even function, thus
we give AutoShoot the command +1, the rst exited state is an odd and there-
fore we give the AutoShoot the command -1 etc. We can do the exact same
procedure for the harmonic oscillator, only we now have the potential function:
V =
x
2
4
.
The triangle potential is a little dierent; here we want to examine the asym-
metric innite triangle, this is a potential which is innity for x < 0 and
x
10
for
17
x > 0. The tricky part of this is, that we cannot calculate exactly with innities;
instead we can use some symmetry arguments to describe how the energy could
behave, from the symmetric innite triangle.
We know that at x = 0 the wave function should equal 0, for x > 0 it is exactly
the same as the symmetric innity triangle. Therefore, if we use innite triangle
energies as a reference, we know that we can only use those whose solution is
0 at x = 0, since the wave function should be continuous. It is therefore only
the rst and third excited state of the symmetric innite triangle, which are
solutions to the asymmetric innite triangle.
We can create a nite asymmetric triangle potential, in which the potential for
x < 0 is equal to 1 and
x
10
for x > 0, but we need to put a limit on this last
term, so it will not be able to exceed 1.
The fourth potential is the Coulomb potential. This potential also needs to be
analyzed in the right way, since for x=0 the potential is minus innity. There-
fore we need to tell Maple that it should not calculate any further than to minus
100, so the potential can not be less than -100. As well as in the asymmetric
triangle potential, we should only use the rst and third excited state, as actual
values for the Coulomb potential, for our ground state and our rst excited
state, respectively. This is because the Coulomb potential will be at x = 0.
Furthermore, it is worth noticing that the energies will all be negative. The
same procedure as the other potentials is then executed.
Here we have shown the dierent energies found and added some comments on
them. Further we have done a comparison between the results found from the
shooting method with the ones from the WKB:
When we did the calculations for the deviation of the harmonic potential, we
noticed that the deviation rose with the rising quantum number, but still the
18
deviation only increased a little. Therefore we can conclude that there is a
coherence between the WKB values and the shooting method values. The reason
for this can be seen in the WKB section where we make the calculations for this
WKB.
For the symmetric innite triangle potential, the deviation is again very low,
since it is approximated zero for three of the four values. So, again, the WKB
is a good approximation. It is the same with the asymmetric innite triangle
potential. But the WKB is not a good approximation for the Coulomb potential
since it has a large deviation. But still we saw that the approximation became
better with the rising quantum number. The Coulomb potential can be seen
as an analogy to the hydrogen atom; the the higher the quantum number, the
more excited the atom is and therefore the less it is bound to the nucleus.
We cannot do the same comparison with the WKB for the nal square well, since
we havent done any exact calculations for the innite square well potential, only
the nite square well. But again we can see that the deviation would have risen
with the rising quantum number, if we had done the calculations for the exact
energy. The values for the energy can be seen underneath.
The dierent graphs can be located in the appendix, with comments.
4 Properties of stationary bound state wave func-
tions and the variational method
4.1 Properties of the Hamilton Operator
The bound SE consists of smaller, mutually orthonormal, waves states and the
entire solution is the superposition of these states. This assumption of com-
pleteness of all states, yield dierent properties, which we will explore in this
section.
A state can be thought of as one possible outcome of a system. This can be
represented as a point in a phase space; here the phase space consists of all
possible values of momentum and position. These states are governed by a set
of dynamic variables, needed to describe the system fully. The representation
for a state can be written as a ket-vector in Hilbert space, containing all infor-
mation about the specic state. The points in the phase space are determined
by the observables of the system, usually represented as a self-adjoint operator,
known as Hermitian. These operators are what we can observe (such as position
or momentum), represented by a matrix acting on each vector in the Hilbert
space. So, knowing all possible states of a system, and the states probability,
19
one has the necessary means to describing a quantum system.
Suppose we have an Hermitian operator acting on the wave function, then,
corresponding to the operator, one would obtain dierent eigenvalues for a spe-
cic system. The eigenvectors/eigenfunctions will then be the solutions to the
dierent states of the system. Consider for instance the Hamiltonian operator
working on the bound SE;
(x) =

n
c
n

n
(x)
H
n
(x) =
n

n
(x)
Here the eigenfunctions are
n
(x) with their corresponding eigenvalues
n
. This
is a crucial and fundamental property of the Hamiltonian; when applied, it
singles out the dierent eigen-energies of each state.
Knowing that the set of eigenfunctions to the matrix H is complete, we can write
the solution in terms of expansion coecients and stationary states. However, if
each state is represented with an expansion coecient, governing the probability
of the specic state, will this also be so for the specic energy of the state? Let
us apply the Hamiltonian to the stationary SE (done here in Dirac notation),
to see how it works:
H | H| =

n
c
n

n
| H|

n
c
n

n
=

n
c

n
c
n

n
| H|
n

n
c

n
c
n

n
|
n
=

n
|c
n
|
2

n
= E
There is a lot of information hidden in this calculation. Most importantly; the
expectation value of the Hamiltonian is equivalent to the mean expectation value
of the systems energy. E is really nothing more than the superposition of all
individual eigenstates energies (
n
)and their probabilities (c

n
c
n
), for a system
described entirely by the wave function |(x).
Another important property of this calculation is, that the wave function can
be expressed as a quantum mechanical superposition of all states. For example,
if one has a die and throws it, then (without looking at it) one can express
the possible outcomes as a sum over the six outcomes (states) with a certain
possibility of being in each state (c
1
= ... = c
6
=
1
6
).
The expansion coecients are ensured to relate discrete probability to the sys-
tem. Just to be safe, let us look at what happens if we take the inner product
of two dierent states;
1 = (x)|(x) =

k
c
k

k
(x)|

n
c
n

n
(x)
=

n
c

k
c
n

k
|
n
=

n
c

k
c
n

k,n
k=n
=

n
c

n
c
n
=

n
|c
n
|
2
20
The squared expansion coecients sum to 1, as supposed to in statistics. We
started by looking at two dierent states (k, n), and saw that the inner product
of two linearly independent system states is a delta function, and that this delta
function may exclude one of the states.
4.2 The variation method
Knowing all this, we can formulate and upper bound for the ground state (
0
);

n
|c
n
|
2

n
|c
n
|
2

0
|H|
0
H
0
This is known as the variation principle. It states, that if one has any normalized
wave function, it is possible to nd a good approximation for the ground state,
since it will be the lowest possible expectation value of H. All other values will
correspond to the particle being in an excited state.
Then choosing dierent test functions for the wave equation, one starts to ap-
proximate the correct wave function. Then, seeing how they behave, on can
pick the most likely equation and nd the unknown parameter on which it is
based. Then by minimizing H, one obtains the minimal expectation value.
4.3 Analytic example of the variation method
Here we will observe a particle in a one dimensional well with the potential:
V (x) =
_
for x < 0
x for x > 0 > 0
We have the following three test functions:

1
= e
x
2
for x > 0

2
= x e
x
2
for x > 0

3
= x
2
e
x
2
for x > 0
For all three three test functions, we have that;
i
= 0 for x < 0 i.
We can resonate, that we must have an asymmetric innite triangle potential.
Since the wave function must be continuous ( = 0 at x = 0), we can discard

1
since it is not zero for x = 0. The best test-function must be
2
, since
it has the lowest energy. This is because the curve of
3
needs to be higher
than that of
2
, and since the curve and energy is proportional, we know that
E
2
< E
3
.
2
must therefor be the best test function, since you always hit
a higher energy with a test function than that of the wave function.
We will now show how to nd an expression for the mean kinetic energy and
the mean potential energy, as functions of and . We start by nding the
21
normalization factor, this is done in the following way:
We dene the wave function to be described as follows, where N is the normal-
ization factor: = N e
x
2
. We then state that the integral of multiplied
by its complexed conjugated, from minus innity to innity, should equal one.
But since we know that = 0 for x < 0, so we only have to evalute the integral
from zero to innity:
_

0

dx = 1
_

0

2
dx = 1 N
_

0
_
x e
x
2
_
2
dx = 1
N
_

0
x
2
e
2x
2
dx = 1 N

(2 1)!
1!
_
1
2

2
_
21+1
= 1
N

2
_
1
2

2
_
3
= 1 N
2

8
_
1

2
_
3
= 1 N =
8
3/2

The integration we did here was done by using the Gauss integral for even
exponents.
We are now ready to start the calculations for the kinetic energy, but instead
of multiplying this normalization factor from the beginning, we have chosen to
do it in the end instead. Therefore we have that:
K
N
=

2
2m
_

0

_
d
2
dx
2
_
dx =

2
2m
_

0

_
6xe
x
2
+ 4x
3

2
e
x
2
_
dx
=

2
2m
_

0
6x
2
e
2x
2
+ 4x
4

2
e
2x
2
dx

2
2m
__

0
6x
2
e
2x
2
dx +
_

0
4x
4

2
e
2x
2
dx
_
Gauss integral even
=

2
2m
_
6

2
_
1
2

2
_
3
+ 4 12

2
_
1
2

2
_
5
_

2
2m
_

3
2

3/2
_
1

2
_
3
+
3
2

5/2
_
1

2
_
5
_

2
2m
3
2
_

_
1

2
_
3
_
1
2
1
_
=
3
8

2
m
_

_
1

2
_
3
We shall now only multiply this with our normalization factor to obtain the
mean value for the kinetic energy:
K =
K
N
N =
3
8

2
m
_

_
1

2
_
3

8
3/2

3
2

2
m

3/2

=
3
2m

22
And now the calculations of the mean value of the potential energy:
V
N
=
_

0

[x] dx =
_

0
x
3
e
2x
2
dx
Gauss integral odd
=
1!
2
_
1

2
_
4
=

2
1
4
2
=

8
2
V =
V
N
N =

8
2

8
3/2

We can see here, that if we make large the kinetic energy will rise, while
the potential will fall. This makes sense since describes how fast the wave
function converges to zero. So, with large follows a larger curve of the wave
function and therefor also a higher kinetic energy, since these are proportional.
This also means that the wave functions position is more well dened and there-
fore the momentum must be proportionally undened; this relates to Heisenberg
uncertainty principle. If is small then our wave function will be more at and
therefore the probability of nding the particle further out is therefore larger,
hence the position will be undened, which will yield a larger potential energy.
We now want to determine the best estimate of the ground state, using
2
as
a test function. This is done by using of the Hamilton operator, since we know
that the energy of the ground state must be smaller or equal to minimum value
of the hamiltonian.
H = K +V =
3
2

m
+

Now to determine H
min
; this is done by taking the derivative of the Hamil-
tonian with respect to and setting this equal to zero:
dH
d
=
3
2

2
m

1
2

2
()
3/2
= 0
3
2

2
m
=
1
2

2
()
3/2
=
_

2m
3
2

_
2/3
Now we only have to substitute this into our denition of the Hamiltonian:
H
min
=
3
2

2
m
_

2m
3
2

_
2/3
+

_
_
_

2m
3
2

_
2/3
_
_
1/2
Which can be reduced to:
H
min
=
_
m
2
+ 1
_

_
6
2

2
m
_
1/3
23
5 Numeric results from the variation principle
Below we show a table over the dierent test-functions on the dierent poten-
tials, further we compare them with the energy values found by the shooting
method.
A graph of the best estimates can be seen in the appendix.
We can see here that the trigonometric test function (cosine) is the best for the
nite square well, the gaussian test function is not bad at all. Below we have
chosen to show the trigonometric test function together with the exact wave
function.
Furthermore, we can see that for the harmonic potential the gaussian test func-
tion is the best. In fact, it is the exact solution, which has to do with the shape
of the gaussian test function. This can also be seen on the following graphs.
At last we can see that all three of the test functions are good approximations,
but the gaussian test function is still the best, which can be seen here:
24
6 Summary
In this project we learned the basic properties of quantum mechanics. Starting
by analyzing the stationary SE and seeing how the propagation of it behaves,
then how to bring it into a dimensionless equation, to make it easier to work
with. After analyzing the geometric properties of the SE, the restrictions on
it, and working through some chosen examples, we saw how some useful ap-
proximation can come in handy. One of these approximations is the WKB
approximation, which we tried a hands on approach to, by analyzing some nu-
merical results. Lastly, we worked with the Hamiltonian operator, to get an
idea of how operators work and how these observables could be used. This was
also followed by some numerical calculations. All of this has created a solid
foundation on how energy interactions occur on a quantum scale.
25
7 Appendix
.1 dimensionless SE
Illustration of the potential well:
.2 The shooting method
In all the graphs, the blue line inidcates the potential, green is the probability
density and the black is the wave function.
The 4th excited state of a nite square well:
Here we can see that this is the fourth excited state, since we have four points
of interaction with the mean energy.
26
The third excited state of the harmonic potential:
Here we can see the third excited state of the harmonic potential. We can also
see that there is a larger possibility of nding the particle in the end-points.
A Gaussian wave packet in an innite symmetric triangle potential:
On this graph, we see the ground state for the symmetric innite triangle po-
tential. We can here see that the wave actually moves outside the potential, this
is not possible in classical mechanics, but in quantum mechanics it can. This is
one of the hard things to visualize.
27
The potential for a nite assymetric triangle:
Here we see the rst excited state in the asymmetric nite well. We cannot see
that at x=10 the potential goes no further then to V=1
The ground state of the asymmetric innite triangle:
Here we have the rst excited state of the symmetric innite triangle but as
we described in the shooting method theory it is also the ground state of the
asymmetric innite well if we only look at x > 0
28
The rst excited state of the Coulomb potential:
Here we see the second excited state of the coulomb potential. As well as in the
example with the asymmetric innite well we here only look at x > 0.
29
.3 Test functions
Square well potential with cosinus as testfunction
We have chosen only to show the cosine test function because it was the best
for the square well potential. Here the black line is the exact, while the red is
our test function.
The harmonic potential, with a Gaussian testfunction :
Here we have the Gaussian test function on the harmonic potential, we can
only see one line since the Gaussian test function is the exact solution for the
harmonic potential.
30
The symmetric triangle potential with a Gaussian test-function:
Here we have the gaussian test function in the symmetric triangle potential, we
see that it ts really well, but we are still able to distinguish the test function
from the exact function.
31