Consider Eq. 2.50, the Fourier transformation of the
square pulse of width . Plots of the wavenumber spectra for pulses
of widths and are shown in
Fig. 2.3. We see that a narrower central peak in the
spectrum corresponds to a wider pulse.
If we define the width , called the wavenumber
bandwidth of the spectrum to be half the width of the central peak,
or the distance between and the first zero of the spectrum, we
have
(2.75)
Hence the spatial and wavenumber widths are inversely
proportional. The spatial width is called the
coherence length.
The comparing Fig.'s 2.4 and
2.5, we see the same kind of
relationship between a Gaussian pulse and its Fourier
transformation. Taking the width of a Gaussian to be its standard
deviation, we have
and from Eq. 2.62
(2.76)
or
(2.77)
Without a formal proof,2.5 drawing from our two examples, we arrive at an
approximate relation,
(2.78)
The direct analogy between wavenumber/position and
frequency/time transformations leads to another bandwidth limit,
(2.79)
Pulse length and duration place lower limits on wavenumber and
frequency bandwidths. The reverse is also true.
Bandwidth limits are of practical importance. For example, they
can be used to estimate the minimum length and duration of pulses
traveling in a medium with a limited bandwidth. You will work with an
example in homework.
In the context of quantum mechanics, bandwidth limits become
uncertainty principles (with
and
). However, we do not call them uncertainty principles
in the context of classical waves, and we should not. Why not?