
FRACTAL.DOC
Lee Daniel Crocker [73407,2030]

April 18, 1989

This paper describes several things that are lumped together under the name
"fractal".  It is not a treatise on any of the subjects involved, but I hope
it answers the basic question, "What is a fractal?" for those interested.  I
apologize to experts in the field for omitting some favorites.


FRACTAL CURVES AND SURFACES

How long is the coast of Great Britain?

This is a simple question, and methods for answering it come immediately to
mind: step a yardstick along the coast end over end, counting the number of
flips of the yardstick needed to circle the island.  Now imagine doing the
same thing with a 12-inch ruler.  It is obvious that this time the answer
will be a bit larger, because the short ruler can measure more precisely the
fine details of the coastline your yardstick skipped.  If you use a 1-inch
ruler, your measurement will be larger still.

Imagine doing this with a perfect circle.  As your ruler gets shorter, you
inscribe within the circle polygons with more and more sides, each of which
comes closer to the circle itself.  Although the measured length gets larger
and larger, it converges toward a single value--the circle's circumference.

Every mathematician is familiar with continuously increasing values that
nonetheless converge to a limit.  As the ruler with which you measure the
coast gets smaller, you might expect its measured length similarly to
converge to a limit of its "true length".  This, in fact, never happens.
The length of the coast increases without bound as your ruler shrinks.
There is no such value as the "true length" of a coastline.

The coastline is a fractal curve.  What makes it a fractal is that it is
infinitely detailed as one goes deeper into its folds.  No matter how much
you magnify it, you only see more and more detail.  This is the essence that
unites all of the fractal types discussed here--each has infinite detail no
matter how much it is magnified.

One can easily simulate this with a computer.  Start with a straight line
drawn across the screen.  Pick a few points along the line and push them up
or down by a small random amount.  Now do the same for each segment of the
line thus created; then with each sub-segment from this pass, and so on,
until the limit of the computer's resolution is reached.  The line now will
be jagged and convoluted in a random and intricate way, and will look very
much like a coastline.

Computer artists use similar procedures to create realistic simulations of
natural phenomena like coastlines.  Extending the process to surfaces
instead of lines yields a method for creating complex terrain that very
closely resembles natural landscapes.  The most realistic computer-generated
scenes of mountains and clouds are created with similar methods.


ITERATED FUNCTION SYSTEMS

A "function" is mathematical formula that takes input values and produces
corresponding output values by some rigidly defined rule.  A simple example
is the function "2 * x + 5" (The * symbol is for multiplication).  When
given the input value 2, it produces the output value 9.  With input 3 it
produces the output 11.

What happens when you feed a function a starting input value, then continue
to feed the resulting output value back in as the next input?  For this
function, you get a series of numbers increasing without bound.  The value
1 yields 7, 7 yields 19, 19 yields 43, and so on through 91, 187, 379, and
upward.  This function is not very interesting.

Things get interesting when two variables are involved.  A pair of variables
can represent coordinates of a point on a plane.  With one such point as a
starting value, a function of the two variables will produce another point
as output.  This point will, in turn, produce a third, and so on.  Usually,
this procedure will eventually fall upon a point that produces itself as
output, thereby breaking the chain.  This point is called an "attractor".

Sometimes instead of settling to a single point, the function will settle
onto a set of points in a small area.  This area is called a "basin of
attraction".  If the function settles into a small set of points that
produce a loop, (e.g., point 1 yields point 2, point 2 yields point 3, point
3 yields point 1) these points are called an "orbit".

Take, for example, this set of equations:

     x2 = 1 + y1 - 1.4 * (x1^2)    (The ^ symbol stands for exponentiation)
     y2 = 0.3 * x1

Put x1 and y1 in as input values and take x2 and y2 as outputs.  Then place
these values back into the equation as its next inputs.  Here are the first
few values starting with x1=0 and y1=0:

     0,            0
     1,            0
    -0.4,          0.3
     1.076,       -0.12
    -0.7408864    -0.036

These values seem random and unordered.  It is hard to imagine them settling
down to a single point, and indeed they do not.  Neither do they settle into
an orbit of a small number of points or into a basin.  The remarkable thing,
though, is that they are not as random as they appear.  This is not apparent
until you plot thousands of iterations of this function on a graphic
display.  The points begin to take a distinct horseshoe-like shape with many
folds.  The points all stay within this curved area, but no point is ever
visited twice.  This is called a "strange attractor".

This particular curve is called the Attractor of Henon, after the
mathematician who discovered it.  Its horseshoe-shaped lines (actually
collections of single points) fold back on each other many times, but never
touch.  If you magnify any part of the curve that appears to be a simple
line, you will discover that it is made of many separate lines, each of
which is in turn made of other separate lines.  The more you magnify, the
more detail you see--this curve is a fractal.


SETS IN THE COMPLEX PLANE

This is a special case of iterated functions of two variables.  I discuss it
separately because the results are so startlingly different and because the
most popular type of fractal graphic--the Mandelbrot Set--is here.

Complex numbers were created to fill a few voids in algebra that real
numbers couldn't handle.  A complex number consists of two values called its
"real part" and its "imaginary part".  It can therefore represent a point on
a plane called the "complex plane".  Complex numbers have special rules for
addition, multiplication, and other functions similar to the methods for 
ordinary real numbers.  In fact, if the imaginary part of a complex number
is 0, it can be treated as an ordinary real number.  For example, the
complex number [2,0] multiplied by [3,0] produces the product [6,0].  [2,1]
multiplied by [3,2], however, produces the product [4,7].

If you perform a function on each point in some section of the complex plane,
you can place each of these points into a certain set based on the result of
the function for that point.  By assigning a color to every point in a given
set, you create some wonderfully intricate designs, which--as I am sure you
expected by now--get more and more detailed as you zoom in.

The most famous of these is the Mandelbrot Set.  For each complex point
[x0,y0], set [x1,y1] = [x0,y0].  Calculate [x2,y2] = [x1,y1]^2 + [x0,y0],
then feed the resulting [x2,y2] back into the formula as [x1,y1].  For
example, starting with the point [2,3], the next point is [2,3]^2 + [2,3] =
[-3,15].  The next point is [-3,15]^2 + [2,3] = [-214,-87].  This particular
point continues to get farther and farther from [0,0], but points nearer to
[0,0] tend to stay close by.

Now count the number of iterations of the formula it takes to produce a
point outside a certain boundary, usually [-4,-4] to [4,4].  Some points
never escape this boundary, so you must specify an iteration limit at which
to give up.  The points that never escape are said to be in the Mandelbrot
Set proper.

The other points bordering the set can be colored based on the number of
iterations it took each of them to escape the boundary.  Subsets of the
complex plane colored this way produce fascinating images, which are more
and more detailed the closer you look.  This, again, is why it is a fractal.


SELF-SIMILARITY

It is common to find small parts of a fractal shape that resemble the whole.
The Mandelbrot Set in particular contains tiny regions that closely (but not
exactly) resemble the whole set, and those regions contain still smaller
regions that resemble them.

Many shapes in nature exhibit this property as well; the branches of a tree,
the fronds of a fern.  Some think that in this way, a structure as complex
as a tree can be described with very little information.  Trees, ferns, and
all other living things store information in DNA, and its capacity for
information is limited.  The information storage capacity of a computer is
limited too, and computers can exploit the property of self-similarity to
reduce the resources needed for storing such structures.

It is possible, for instance, to store the basic shape of a fern along with
a few instructions for replicating this shape at different scales in
different positions, and to use this information to recreate a detailed
image of the fern using less storage than traditional methods might require.
Such iterated functions are used to represent compactly many of the shapes
of nature.  "Pure" mathematical shapes such as circles and squares are not
well suited to this kind of storage, but many natural shapes are.  This
process is often seen discussed as "fractal image compression".


FURTHER READING

There are many books available on fractals and related subjects.  Here are
some of the most popular:

"Chaos: The Making of a New Science", James Gleick

    A good (much better than mine) layman's introduction to the subject.
    Not much detail for the more serious student, though.

"The Fractal Geometry of Nature", Benoit Mandelbrot

    The classic reference by the creator of the term "fractal" and the
    example that bears his name.  Pretty technical, but good reading and
    wonderful illustrations.

"The Science of Fractal Images", Barnsley, et al.

    The most detailed discussion of computer algorithms available for the
    various types of fractals.  More wonderful illustrations.

"The Beauty of Fractals", Peitgen and Richter

    Another book filled with beautiful pictures and mathematics.  Many
    examples of different fractals with the equations that generated them.

I am sure there are many more good books on the subject.  Feel free to add
to this list as this paper is passed around.

I owe special thanks to Bert Tyler and Timothy Wegner whose program
FRACTINT renewed my interest in this field.  FRACTINT runs on many MSDOS
personal computers and is available from the Graphics Support Forum on
CompuServe.  Type GO PICS from any "!" prompt while on CompuServe to gain
access to this forum.  Call 800-848-8199 for information about CompuServe.

