Classically, most attention has been paid to
the integers generated by this algorithm, which make up the continued
fraction itself. However, Gauss was apparently the first to study the
other part of this algorithm, which we present as the following map
from to
, called
the Gauss map [18] (see
FIGURE 1):
We use the notation ``mod 1'' to mean taking the fractional part. In terms of the Gauss map G, our algorithm then becomes
Note that the jump discontinuities occurring at (for each positive integer
n) may all be removed by mapping onto the circle with the transformation
. After this is done, we see that the Gauss map (
) is a map of the circle onto the circle, and may be
pictured on a torus, as in
FIGURE 2.
Click here to examine a
Maple program for graphing arbitrary
on the torus.
We make the following observation: if we represent a point in the
interval by its continued fraction,
,
then a simple induction shows that
,
, and so on. This makes a connection
between the Gauss map and the
shift map of symbolic dynamics [8].
We will not explore this connection further here, but we note that
the shift maps normally studied are slightly different than the Gauss map,
in that here the size of the numbers in the list being ``shifted'' is
not bounded.
Click here for more
details on the induction.
An analogy is illuminating: if we think of our space as a circular hoop
with the origin at one point O on the hoop, our initial point as a
dimensionless bead on the hoop, and the Gauss map as taking the bead from
its current position clockwise past O at least once to its next position
on the hoop, then the integers are the number of times the bead passes
O on the ith iteration (in general the maximum such number is called
the ``winding number'' of the map, and here this is obviously infinite),
and the
are the coordinates of the bead on the hoop once
it comes to rest. If the bead comes to rest close to the origin on one
side, with a small
, then on the next iteration it will be pushed
many times around the hoop. If it comes to rest close to the origin on the
other side, with a
close to 1, then it will only go past the
origin once on its next iteration. We may think of the bead as being
pushed around the circle, with the strength of the push being inversely
proportional to the distance measured counterclockwise from the point O.