Derivation of the Heisenberg Uncertainty Principle

Andre Kessler
April 13, 2010
We start off with our generic wave function (x, t). This can be written as
(x, t) = A cos (t kx)
Now, since the angular frequency is defined to be = 2 and the wave number is de2
fined to be k =
, we can substitute these quantities into the equation to get (x, t) =

2
h
A cos (2t
). Additionally, we know that E = h and = . Substituting these

p
2
1
quantities into the equation, factoring, and replacing
with , we get an alternate form
h
~
of the wave function:
1
(x, t) = A cos (Et px)
~
Theorem 0.1. (Fouriers Theorems)
1. Any periodic waveform can be represented by a discrete series of sines and cosines:
f (x) =

ai cos (i t ki x) + bi sin (i t ki x)

i=1

2. Any waveform can be represented as a continuous sum of sines and cosines (Fourier
transform):
Z
f (x) =
A(k) cos (t kx) + B(k) sin(t kx) dx

Luckily, for a lot of waveforms, you can get away with just one sine or cosine and thus
just one set of coefficients. Also, it turns out that the Fourier conjugate of a very localized
waveform will be spread out. Thus, if position and momentum (or energy and time, etc) are
Fourier conjugates, and if you know the position to a high degree of accuracy, then you dont
know the momentum very well and vice versa. Heisenberg found out that you can quantify
~
this as xp , and that is what we will derive.
2
1

Since (x) is a waveform, we can write it as a Fourier series:

Z
g(k) cos kx dk
(x) =

or
(x) = F[g(k)]
How do we find the coefficients? We should expect the classical momentum to be the
average value, and other values to be less probable - in other words, some sort of probability
2
distribution. Specifically, the normal distribution, meaning probability Pk = Ak e(kk0 )/(2k ) .
This implies that
p
p
2
g(k) = Pk = Ak e(kk0 )/(4k )
Now, consider the square of the norm of the wavefunction, ||2 . This is a probability
distribution; it tells you where the particle is most likely to be. Like earlier, we want the
middle value to be the classical position. Like earlier, it happens to be a normal distribution.
If we let x0 be the most likely position for the particle, then a normal distribution of the
2
2
positions is Px = Ax e(xx0 ) /(2x ) .Now, we will denote the envelope of a function f by fenv .
Since Px happens to be the upper envelope of |(x)|2 , (x)env happens to be
p
p
2
2
(x)env = Px = Ax e(xx0 ) /(4x )
Z
Z
|(x)|2 dx = 1, because the sum of all the probabilities
Px dx =
Additionally,

must be 1. This means we can find the coefficients Ax by simply normalizing the normal
distribution; this gets us
1
Ax =
x 2
Now, we state that (x) = Fg(k) in the following way:
Z
(x) =
g(k) cos kx dk

and since (x) and g(k) are Fourier conjugates the following is true as well:
Z
g(k) =
(x) cos kx dx

Substituting our results above gives

p

(kk0 )2 /(4k2 )

Ak e

1
2
2
p e(xx0 ) /(4x ) cos kx dx
x 2

(1)

Since the function (x) cos kx is even, its integral from negative infinity to infinity is simply
twice its integral from zero to infinity. The integral of that function is well known and can
be found in integral tables:

Z
b22
a2 x2
e 4a
e
cos bx dx =
2a
0
In the case of (x), first we note that we may as well choose x0 = k0 = 0, and we also have
1
1
a2 = 2 a =
and b = k. Lets see how our integral from (1) looks now:
4x
2x

Z
p
1 x2 /(4x2 )
2 k2 /(4( 412 ))
k2 /(4k2 )
x
cos kx dx =   e
Ak e
=2
e
x
0
2 1
2x

p
The Ak and
2

2
  are equal (?), so when we cancel that we get
1
2x

k2
4 2
k

=e

k2 !
1
4
2
4x

Now we take the natural log of both sides and were left with
k 2
k 2


=
1
4k2
4 42
x

This simplifies to k2 =

1
or
4x2
k x =

But we know that the wave number k =

1
2

p
, so
~
x p =

~
2

Note that this is true only if the probability distribution is normal. If it isnt p x will be
greater, as the normal distribution turns out to have the minimum possible product. This
means that we can write our more general physical law as
x p

~
2

or as
xp

~
2