The previous section was concerned with transformations obtained by composing a series of inversions in circles. These transformations are now known as Möbius transformations, and thanks to Poincaré, we know they have a particularly simple algebraic formula (which I won't tell you). Instead, I will list the basic types of Möbius transformations that can occur
Some of these transformations may be familiar. We shall limit ourselves to an illustration of the last and most beautiful transformation, the spiralling one, with an intrepid walker in the movie loxowalkB.mpg. We have also put in a few `guardrails' for his safety.
Here we are already showing a more general idea of `tiling'; where the tile is Dr. Stickler and the group is generated by a single loxodromic transformation.
In general, the study of kleinian groups begins with the selection of the generating transformations and then examines their collective action on geometric objects. Peter Doyle discovered (in a slightly different form) an amazing symmetry of loxodromic elements acting on a simple configuration of tangent circles. We shall phrase it in the form of a visual theorem in Figure 23.
|
|
The theorem means that any four such circles that comes with one
pairing transformation
automatically has another
. The group
generated by these transformations can be applied to the original
starting circles to develop a beautiful packing of circles in the
plane. If the four circles are all the same size, we simply get the
usual hexagonal packing of the plane. But if the circles are of
different sizes, we get something a lot more beautiful: a Doyle
Spiral. The transformations
and
will be loxodromic spiralling
element whirling the circles all around and through each other.
In general, the circles will overlap; but if we solve delicate
algebraic equations for certain special crystalline algebraic numbers,
we will achieve what Ken Stephenson dubbed a coherent Doyle
spiral (see [#!BeardonDubejkoetal1994!#]). In fact, the example in
the theorem is just such a coherent packing, as we see by applying
repeatedly in Figure 24. There is a hidden
structure in that 16 applications of
is the same as one
application of
. Thus, the group presentation has relations
and . Beardon, Stephenson, and Dubejko proved there is
exactly one such Doyle spiral for any similar such relation .
This is a relatively uncomplicated group, and yet if we have the technique of finite
state automata it's convenient for enumerating precisely the
group elements. When we do so, the complete Doyle spiral looks like
Figure 25. Here we change colors each time we apply
, and we leave the color alone when we apply
. The color
emphasizes both spiral transformations.
It is a delicate calculation to find the transformations for a given integer pair that produce the Doyle Spiral. The next figures show the (17,43) spiral and the (10,101) spiral.
If you've seen some color visualizations of functions of a complex variable, these may seem familiar. The spirit of using circle packings to approximate conformal functions is explained in Ken Stephenson's book Introduction to Circle Packing, and indeed the Doyle packings converge to a certain well known function. But that theorem is just a little off in the future.
The idea of interlocking families of spirals appears in nature as the phenomenon of `phyllotaxis', visible in pine cones, sunflower seed arrangements, shells, etc., although in nature the integers are almost always consecutive Fibonacci numbers.
We'll end this section with some of Jos Leys' striking visualizations of the coherent Doyle packings.