You are on page 1of 13

Notes on Quantum Mechanics

Salvatore Cardamone, 2014

Pre-requisites

1.1

Plane Waves

Consider a single-dimensional sinusoid moving along the x-axis with amplitude A0 . The
magnitude of the wave at position x and time t is given by the function
A(x, t) = A0 cos(kx t + )

(1)

where
k is the wavenumber, and is equal to 2/, being the wavelength of the wave.
k has units of radians per unit distance, an so corresponds to the rate at which
A(x, t) changes over a given distance.
is the waves angular frequency, and is equal to 2/T , T being the period of
the wave. T has units of radians per unit time, and so corresponds to the rate at
which A(x, t) changes over a given length of time.
is the phase of the wave, which corresponds to a shift of the wave along the
direction of propagation.
This formalism is easily generalised to a wave moving in a higher dimensional space by
treating the direction and wavenumber as vectors, termed a plane wave, such that
A(r, t) = A0 cos(k r t + )

(2)

where r is now a position vector corresponding to a point in R3 , and k is the wave vector.
Each component thus maps the rate at which A(r, t) changes over a given displacement
along the corresponding basis vector. The direction of the wave vector is typically in

the direction of propagation of the plane wave.


Mathematical manipulation of planes waves is greatly facilitated by use of the complex
exponential form, whereby the plane wave is described by
U (r, t) = A0 ei(krt+)

(3)

Expanding this in terms of sines and cosines


U (r, t) = A0 cos(k r t + ) + iA0 sin(k r t + )
U (r, t) = A(r, t) + iA0 sin(k r t + )
This form is simplified somewhat by modifying the amplitude of the wave to absorb the
phase factor, such that
U (r, t) = A0 ei(krt+) = A0 ei ei(krt) = U0 ei(krt)
Obviously, the plane wave is a real entity and we have generated a function which is
complex. However, once we have performed our calculations in the complex plane, we
are left with the trivial operation of taking the real quantity of U (r, t)
<(U (r, t)) = A(r, t) = A0 cos(k r t + )
As a final word on the topic, in the bulk of this document, solutions to the Schrodinger
Equation are by their very nature complex, and so the imaginary component in this
occurrence has not been introduced as a byproduct, but is in fact an inherent part of
the wave.

1.2
1.2.1

Fourier Analysis
The Fourier Series

Consider a function, f (x), which is periodic with period L. If f (x) obeys certain mathematical conditions, it may be expanded in a series of imaginary exponentials (and
thus, trigonometric functions), termed a Fourier Series. We proceed initially with the
complex exponential form, expanding our function
f (x) =

cn eikn x

(4)

n=

where kn = 2n/L. Multiplying each side by the complex conjugate of our exponential
(introducting the dummy variable p) and integrating over a single period of f (x), i.e.
between x0 and x0 + L

x0 +L

f (x)eikp x dx =

x0

x0 +L

Z
cn

ei(kn kp )x dx

x0

n=

The integral of the exponential function on the right hand side of the above allows
for a significant simplification upon decomposition into the corresponding trigonometric
functions, and realisation that the integral of the constituent sinusoids over the period
is equal to zero, since
Z

x0 +L


cos

x0

2n 2p

L
L

 
Z
x dx =

x0 +L


cos

x0

2(n p)x
L


dx = 0

with the same obviously being true for the sine function. However, if n = p, then the
right hand side simply becomes
x0 +L

dx = L
x0

allowing us to elegantly conclude that


Z

x0 +L

f (x)eikp x dx =

x0

an Lnp = ap L

n=

where we have introduced the Kronecker delta, pk . Therefore, the coefficients of the
Fourier Series are given by
an =
1.2.2

1
L

x0 +L

f (x)eikn x dx

(5)

x0

The Bessel-Parseval Relation

From (4), we may obtain

x0 +L

|f (x)|2 dx =

x0

x0 +L

f (x)f (x)dx =

x0

x0 +L

x0

|f (x)|2 dx =

x0

Z
X

n= p=

x0 +L

x0 +L

cp cn eikp x eikn x dx

n= p=

cp cn ei(kn kp )x dx =

x0

cp cn np L

n= p=

where the Kronecker delta arises in a manner equivalent to that shown in the previous
section. Subsequent simplification leads us to
1
L

x0 +L

|f (x)|2 dx =

x0

|cn |2

(6)

n=

which is termed the Bessel-Parseval Relation. Such a relationship proves to be extremely


useful if we consider the scenario where we have two functions, f (x) and g(x), each with
the same period, L, the Fourier coefficients of which are cn and dn , respectively. We
may subsequently generalise (6) such that
3

1
L
1.2.3

x0 +L

g (x)f (x)dx =

x0

dn cn

(7)

n=

The Fourier Transform

We begin this section with the statement that the fourier integral is a limit of a Fourier
series. Consider a function f (x) which is not necessarily periodic. We define fL (x) to
be a periodic function of period L which is equal to f (x) over the interval [L/2, L/2].
As we have shown, fL (x) may be expanded as a Fourier series
fL (x) =

cn eikn x

n=

where our Fourier coefficients are defined by (5), such that


1
cn =
L

x0 +L

ikn x

x0

1
fL (x)dx =
L

L
2

eikn x f (x)dx

L
2

We recall that the various kn were defined by kn = 2n/L, allowing us to form a


recursive relationship
2
1
kn+1 kn

=
L
L
2
which we substitute into our expression for the Fourier coefficients, and utilise this in
kn+1 kn =

our initial Fourier series, yielding


Z L

X
kn+1 kn ikn x 2 ikn x
e
fL (x) =
e
f (x)dx
2
L
n=
2
As L , (kn+1 kn ) 0, such that the summation over n is transformed into a
definite integral over the continuous variable k. Also note that in this limit, f (x)
fL (x), and so we are essentially approximating our unknow function f (x) by some
analytic periodic function fL (x). Re-establishing the limits of our integral in the above
equation to allow for L , we establish a function
1
F(k) =
2

eikx f (x)dx

(8)

eikx F(k)dk

(9)

which may be used to generate


1
f (x) =
2

where f (x) and F(k) are termed the Fourier Transforms of one another.

Fundamental Assumptions of Quantum Mechanics

2.1

Introduction

Quantum mechanics divides the world into two parts composed of a system and an
observer, which do not interact, except during the process of measurement. Quantum
mechanics predicts all the information an observer may obtain regarding a system, often represented in terms of a wavefunction. A measurement changes the information
an observer has about the system, and therefore changes the wavefunction of the system.
Applied to a system consisting of a single, structureless particle, the fundamental assumptions of quantum mechanics are;
1. The quantum state of a particle is characterised by a wavefunction, (r, t), which
contains all information regarding a system which an observer can possibly obtain.
2. (r, t) is interpreted as a probability amplitude of the particle, and so |(r, t)|2
gives a corresponding probability density. The probability that a particle is a time
t in an infinitesimal volume element dr is given by
dP (r, t) = C|(r, t)|2 dr

(10)

where C is some normalisation constant and P (r, t) is the corresponding probability density function. Neglecting particle destruction/ creation (as occurs with
photons in relativistic quantum mechanics), the particle must occupy some portion
of space, and so we are free to write

P (r, t)d r = 1

|(r, t)| d r

R3

2 3

R3

|(r, t)|d3 r <

R3

(11)

where the final statement implies that a wavefunction must be square-integrable.


3. The principle of spectral decomposition applies to the measurement of an arbitrary
physical quantity, A, such that;
(a) The result of a measurement belongs to a set of eigenvalues, {a}.
(b) Each eigenvalue is associated with an eigenfunction, a (r). If (r, t) = a (r),
then a measumrent of A at time t will yield the eigenvalue a.
(c) Any (r, t) can be expanded in terms of the eigenfunctions of A
(r, t) =

X
a

ca a (r)

(12)

The corresponding probability that a measurement at time t will yield the


eigenvalue a1 is subsequently given by
|ca |2
P (a1 ) = P 1 2
a |ca |

(13)

which is readily obtained when one implements the fact that the eigenfunctions for a mutually orthonormal basis.
(d) If a measurement of A yields a, then the wavefunction of the system immediately after measurement is given by a (r).
4. The Schr
odinger Equation describes the evolution of (r, t)

i~

~2 2
(r, t) =
(r, t) + V (r, t)(r, t)
t
2m

(14)

where the particle has a mass m and is subjected to the potential V (r, t).

2.2

A Free Particle

A free particle is not subjected to any forces, and so we may arbitrarily set V (r, t) = 0,
allowing for subsequent simplification of the Schrodinger Equation to
~2 2

(r, t) =
(r, t)
t
2m
Plane waves are potential solutions to the above, of form
i~

(r, t) = Aei(krt)

(15)

(16)

Finding the various derivatives of this function

(r, t) = i(r, t)
t
(r, t) = ik(r, t) 2 (r, t) = k 2 (r, t)
and recombining to form the Schr
odinger Equation
~(r, t) =

~2 k 2
(r, t)
2m

which constrains our choice of to


~k 2
2m
By use of the de Broglie relations, E = ~ and p = ~k, we obtain
=

E=

p2
2m

(17)

(18)

which is simply the kinetic energy of the particle, as we would expect given we assumed
no external potential.
We note that a plane wave represents a particle whose probability distribution is constant
throughout space
|(r, t)|2 = | (r, t)(r, t)|
|Aei(krt) Aei(krt) | = |A2 |
which is not a proper solution to the Schrodinger Equation as this is not square-integrable
by virtue of
Z

(r, t)(r, t)dr = |A|

dr =

Since the Schr


odinger Equation is linear, the principle of superposition applies, i.e. that
if a set of functions satisfy the equation, then a linear combination of them is also a
satisfactory solution. As such, a linear combination of plane wave solutions is also a
solution
(r, t) =

ak ei(krt)

(19)

recalling that we have the constraint that


~k 2
2m
It is hoped that the reader recognises (19) as being a three-dimensional equivalent to
k =

(4). As such, we are free to write (r, t) as a Fourier transform


Z

1
(r, t) =
(2)3/2

g(k)ei(krt) dk

(20)

R3

Such a wave function is called a three-dimensional wave packet, and may represent
any square-integrable function. For the sake of simplicity, we take the case of a onedimensional wave packet, whose wavefunction is given by
1
(x, t) =
2

g(k)ei(kxk t) dk

(21)

A Brief Example
We may gain some insight regarding the physical form of a wave packet by considering a superposition of a finite number of plane waves. In this case, we will take 3,




k
, and amplitudes 1, 12 , 12 , respectively.
with wavenumbers k0 , k0 k
2 , k0 + 2
Then, using these in (22), we obtain


1 i(k0 + k
g(k0 ) ik0 x 1 i(k0 k
x
x
)
)
2
2

+ e
(x, 0) =
e
+ e
2
2
2



g(k0 )
k
(x, 0) = eik0 x 1 + cos
x
2
2

(22)

By evaluation of the argument of the cosine, we see that |(x, 0)| is maximal when
x = 0. This is due to the fact that at this value of x, the three waves which form our
wave packet interfere constructively; as x moves away from x0 , the waves gradually
dephase with one another and being interfering destructively, leading |(x, 0)| 0
k
when the phase shift between eik0 x and ei(k0 2 )x is . We take the interval of

  x x 
x
the wave packet to span x0 x
2 , + 2 , and find the values of
2 , x0 + 2
x and k which satisfy this phase shift condition

k0 +

k
2

x k0 x

=
2
2

k0 x kx k0 x
+

=
2
4
2
kx = 4
By this, we see that the smaller the width, k, of the function |g(k)|, x must
complement this by increasing to satisfy the above equation, i.e. the more diffuse
the function |(x)| becomes (the distance between the two nodes of (x). As a
point of note, we see that (21) is periodic in x, and therefore there exists a series
of maxima and minima. Such a scenario is a result of a superposition of a finite
number of plane waves; in the limiting case where we take an infinite number of
plane waves, periodicity is lost over the real numbers, and only a single maximum
exists.
At some instance in time, say t = 0,
1
(x, 0) =
2

g(k)eikx dk

(23)

1
F[(x, 0)] = g(k) =
2

(x, 0)eikx dx

We assume g(k) to be some normally distributed function centred at k0 with width k,

and is given by g(k) = |g(k)|ei(k) . We further assume that k << k0 . We may expand
(k) in a Taylor series about the point k0 , i.e.

d
+ ...
(k) = (k0 ) + (k k0 )
dk k0
We are only interested in the function in a small interval about k0 , and so we discard
all second order and higher terms in the above expansion, leaving us with
Z
d
1
(x, 0) =
|g(k)|ei(k0 ) ei(kk0 ) dk eikx dk
2


Arbitrarily setting x0 = d
dk k0 , the above reduces to
1
(x, 0) =
2

1
|g(k)|ei(k0 ) ei(kk0 )x0 eikx dk =
2

|g(k)|ei(k0 ) eikx0 eik0 x0 eikx dk

Noting that we may add the term eik0 x eik0 x to our product of exponentials without
altering the function, and realising that this allows for a factorisation
Z
1
(x, 0) =
|g(k)|ei(k0 ) ei(kk0 )(xx0 ) eik0 x dk
2
Removing the exponentials which are not functions of k from the integral
Z
ei((k0 )+k0 x)

|g(k)|ei(kk0 )(xx0 ) dk
2
which is in a form more amenable to analysis. Primarily, we wish to reiterate that this
(x, 0) =

function is the equivalent of a Fourier series, in which we are essentially taking an infinite number of plane waves to recreate our wavefunction (x, 0). We initially consider
the case where |x x0 | is large, leading to the integrand oscillating substantially in the
interval of k. Qualitative analysis of this function reveals that the integral over k will
then lead to |(x, 0)| 0. In other words, if x is far-removed from x0 , the waves which
form (x, 0) rapidly become out of phase, and interfere destructively. Alternatively, if
x x0 , the integrand barely varies at all, and subsequent integration yields |(x, 0)|
tending towards its maximum value.
We are now in a position to allude to a particularly important relationship, which
will form the majority of the next section. When x moves away from x0 , |(x, 0)|
decreases. We see that |(x, 0)| actually equals zero when the integrand undergoes a
single oscillation over the interval k, i.e. when the exponent is equal to one
k(x x0 ) = 1
(which is equivalent to saying the position is equal to the wavelength). Denoting the
width of the wave packet by x, we see that the above relation gives a lower bound to
the exponent
9

xk 1

(24)

which is nothing more than a classical relation between the widths of two functions
which ae Fourier transforms of one another.

2.3

The Heisenberg Uncertainty Principle

The inequality in (23) proves to be extreme importance in quantum mechanics, and


allows us to derive some physical restrictions to what may be known about a system.
Given a plane wave of the form ei(k0 x0 t) , by use of the de Broglie relations (E = ~0
and p = ~k0 ), we see that the energy and momentum of the system are well-defined and
unrestricted, in that k0 and 0 may take any real value. We may view this in a different
manner by spectral decomposition; if we have a particle defined by the wavefunction
(x, 0) = Aeikx
then in saying the particle has a well-defined momentum p is equivalent to saying that
eikx characterises a momentum eigenstate with eigenvalue p = ~k. Since k may take
any real value, there are an infinite number of eigenvalues which one may realise upon
measurement of the state; as in classical mechanics, all values of the momentum are
physicall realisable.
However, consider (22), in which (x, 0) occurs as a linear superposition of momentum
eigenfunctions, with coefficients g(k). This allows us to interpret |g(k)|2 as the probability of finding a system in the eigenstate eikx , and so is a probability density for the
eigenstates of the system. The probability, dP (k), of obtaining an eigenvalue in the
interval [~k, ~(k + dk)] is therefore given by |g(k)|2 dk. Rewriting (22) in terms of the
momentum
(x, 0) =

1
2~

ipx/~

(p)e
dp

and implementing the Bessel-Parseval relation (6), we obtain


Z

|(x, 0)|2 dx =

|(p)|
dp

Denoting the value of the integral to be C, we see that


1
|(x, 0)|2 dx
C
which is the probability of the particle being found in the interval [x, x + dx] at t = 0.
dP (x) =

In an entirely equivalent manner


dP (p) =

1
|(p)|2 dp
C
10

which is the probability of the particle having a momentum in the interval [p, p + dp].
Now, returning to our inequality (23), we rewrite
xp ~

(25)

where p = ~k and represents the width of (p).


Therefore, consider a particle
whose position probability is defined by some region x about x0 , so that x represents the uncertainty in our knowledge in the position of the particle. Then, if one
were
to measure the
i momentum of the particle at the same time, a value in the interval
h
p
p
p0 2 , p0 + 2 , the uncertainty in the momentum being given by p. We may
therefore interpret (24) as the impossibility of being able to simultaneously decrease the
product of the uncertainties in both the position and the momentum of a particle below
~, i.e., we cannot know both the position and the momentum to an arbitrary degree of
accuracy. This relation is known as the Heisenberg Uncertainty Relation.
There is no analogous phenomenon within classical mechanics; the limitation expressed
in (24) arises from the fact that ~ is not equal to zero; it is only because ~ is extremely
small by macroscopic standards that we do not encounter the physical consequences
of this inequality on the macroscopic scale. Reiterating what was said in the previous
section, there is nothing inherently quantum mechanical about (24) as it merely expresses
a general property of Fourier transforms, numerous examples of which exist within
classical physics, e.g. we cannot simultaneously know the position and wavelength of
an electromagnetic wave with infinite accuracy. As such, Heisenbergs Relation does
not arise from some strange physical phenomenon, it is merely a byproduct of some
well-defined mathematics which manifests as some bizarre principle when applying to
physical reality.

2.4

Time-Evolution of a Wave Packet

A plane wave ei(kxt) propagates through space with a Phase Velocity, V (k), given by

k
We may draw physical insight from this by considering the case of an electromagnetic
V (k) =

wave, the phase velocity of which in a vacuum is independent of k and equal to the
speed of light, c. In a dispersive medium, on the other hand, the phase velocity of a
wave is moderated by the index of the medium, n(k), such that
V (k) =

c
n(k)

Recall (17), which allows us to write


V (k) =

~k
=
k
2m

11

which we see is similar to the phase velocity of a wave in a dispersive medium. When the
waves which form a superposition have unequal phase velocities, we shall see that the
velocity of the maximum point of the wave packet, xmax (t), is not given by the average
phase velocity
~k0
0
=
(26)
k0
2m
as one may expect. We begin by attempting to understand a qualitative description of
what happens once a wave packet is allowed to evolve with respect to time. Consider the


k
, angular
superposition of three plane waves, with wavenumbers k0 , k0 k
2 , k0 + 2




1 1
,

+
,
velocities 0 , 0
and
amplitudes
1,
,
respectively.
Then,
for
0
2
2
2 2
arbitrary t


1 i[(k0 + k
g(k0 ) i(k0 x0 t)
1 i[(k0 k
x(0
t]
x(0 +
t]
)
)
)
)
2
2
2
2
e
+ e
(x, t) =
+ e
2
2
2



g(k0 )
k

(x, t) = ei(k0 x0 t) 1 + cos


x
t
2
2
2
Thus, we see that the maximum point of |(x, t)| is given when the argument of the
cosine is equal to zero, i.e., when
k

x=
t
2
2

xmax (t) =
t
(27)
k
which differs from the value given in (25). This deviation from expectation has a physical origin which may be demonstrated by an analysis of the constituent waves. At time
x = 0, t = 0, we see that all three waves will be at their maximum value and so reinforce
constructively to give xmax . The three waves will emanate from this point, and dephase
as we move along x, owing to their differing angular velocities. However, we have just
seen that the phase velocity is dependent upon k, and so the maximum of the wave
k
2

characterised by k0 +

will propagate along the x-axis quicker than the other two.

As such, after a given amount of time, the maxima from the three waves which were

initially dephased will gradually become in phase as the k0 + k
wave catches up with
2
the other two.
We can arrive at the same conclusion based on (??). We see that to go from (x, 0) to
(x, t), all we need to do is change the distribution g(k) to g(k)ei(k)t , thus
1
(x, t) =
2

|g(k)|e

i((k)(k)t) ikx

1
dk =
2

|g(k)|ei(k) eikx dk

where we have simply used the identity (k) = (k) (k)t and expressed g(k) in polar
form. We differentiate this function with respect to k and evaluate at k0 to yield
12




d
d
d
=

t
dk k0
dk k0
dk k0


Recalling that we earlier defined d
dk k0 = x0 , and substituting in



d
d
d

t
=
x

= x0 + VG (k0 )t
0
dk k0
dk k0
dk k0
where we have defined

d
= VG (k0 )
dk k0

(28)

which we term the Group Velocity, and represents the velocity of the peak of the wave
packet. We may readily find the functional form of the group velocity by differentiating
(17) with respect to k
~k
d ~k 2
=
= 2V (k0 )
dk 2m
m
This relationship proves to be extremely important, as it enable us to recover the classical
VG (k0 ) =

situation for a free particle. In this case, uncertainty plays no role due to ~ being
negligible at the macroscopic level, and so it is perfectly reasonable to speak of a well
defined particle momentum, p0 , corresponding to a well-defined position, xmax (t). In
this case, the velocity is given by v =

p0
m,

which is implied by the above formula. In the

cases where x and p may both be considered negligible, the maximum of the wave
packet moves like a particle which obeys the laws of classical physics.

13

You might also like