In Section 3.2 we have assumed that
generic fluctuating quantities such as
and
and
have a white noise power spectrum
for
.
In section 4.1
we are going to suggest that this white noise
assumption is approximately true for any
fluctuating quantity
that comes
from a normal deformation (the term `normal'
will be defined there).
Obviously, the goodness of the `white noise assumption'
in the two cases mentioned
is related to the chaoticity of the system, and should be
tested for particular examples.
This has been done so for the cavity of Fig. 3.2
(see [14], and Figs. 3.4 and 4.2).
This cavity is an example of a `scattering billiard'
and so exhibits strong chaos (no marginally-stable orbits) [40].
If the motion is not strongly chaotic we may get a
that decays like a power law (say
with
) rather than an exponential
[40,78,54,37] (time of crossover to algebraic
decay is discussed in [53]).
In such case the universal behavior
is modified: we get
behavior for
at small frequencies (
diverges),
signifying faster-than-diffusive energy spreading in Eq.(2.8)
[37].
The stadium is an example where such a complication
may arise: an ergodic trajectory can remain in the
marginally-stable `bouncing ball' orbit family
(between the top and bottom edges)
for long times, with a probability
scaling as 
[78,54,53].
Depending on the choice of
this
may manifest itself in .
For example, in Figs. 2.5 and 3.7
the deformations P and Gp respectively
both
involve a distortion confined to the upper edge,
and the resulting sensitivity to the bouncing ball orbit
leads to large enhancement of
the fluctuations intensity
,
and is suggestive of singular behavior for small 
.
However in the same system the deformations W2 and G,
which are zero on the upper and lower edges,
show no such enhancement--the band profile is flat as
and the deviation from strong chaos is masked.
If the billiard has a mixed phase space (which is the generic case), then the integrable component does not contribute to diffusive energy spreading. Proposals have been made to account for this via a phase-space volume factor [158,149].