Fractals

The Koch Snowflake

3
4
= Length
1 = Length
|
.
|

\
|
=
3
4
2
Length
First iteration
After
2 iterations
After 3 iterations
|
.
|

\
|
=
3
4
3
Length
After n iterations
|
.
|

\
|
=
3
4
n
Length
= =
|
.
|

\
|

3
4
Length
After iterations

(work with me here, people)

The Koch snowflake is six of these put together to form . . .
. . . well, a snowflake.
Notice that the perimeter of the Koch snowflake is infinite . . .
. . . but that the area it bounds is finite (indeed, it is
contained in the white square).
The Koch snowflake has even been used in technology:
Boston - Mar 13, 2002
Fractal Antenna Systems, Inc. today disclosed that it has
filed for patent protection on a new class of antenna arrays
that use close-packed arrangements of fractal elements to
get superior performance characteristics.
Fractal Tiling Arrays -- Firm Reports
Breakthrough in Array Antennas
Can you name the movie?
Each of the six sides of the Koch snowflake is
self-similar: If you take a small copy of it . . .
. . . then dilate by a factor of 3 . . .
. . . you get four copies of the original.
But self-similarity is not what makes the Koch snowflake
a fractal! (Contrary to a common misconception.)
After all, many common geometric objects exhibit
self-similarity. Consider, for example, the humble
square.
If you take a small square . . .
. . . and dilate by a factor of 2 . . .
. . . then you get 4 copies of the original.
A square is self-similar, but it most certainly is not a fractal.
If you take a small square . . .
. . . and dilate by a factor of 3 . . .
. . . then you get 9 copies of the original.
Let k be the scale factor.

Let N be the number of copies of the original that you get.
Note that for the square, we have that:
2 log = N
k
N
k
=
2
Or in other words, we have:
Lets compute N
k
log
for some other shapes.
Line segment
Original
Dilated
k = scale factor = 2
N = number of copies of original = 2
1 log = N
k
Triangle
Original Dilated
k = scale factor = 2
N = number of copies of original = 4
2 log = N
k
Cube
Original Dilated
k = scale factor = 2
N = number of copies of original = 8
3 log = N
k
Shape
Square 2
Line segment 1
Triangle 2
Cube 3
N
k
log
What does
N
k
log
tell us about a shape?
Thats right:
N
k
log
tells us the dimension of the shape.
(Note that for this to make sense, the shape has to be
self-similar.)
So for a self-similar shape, we can take
N
k
log
to be the definition of its dimension.
(It turns out that this definition coincides with a much more
general definition of dimension called the fractal dimension.)
Now lets recall what k and N were for one side of the
Koch snowflake:
k = scale factor = 3
N = number of copies of original = 4
... 261 . 1 4 log log
3
~ = N
k
So each side of the Koch snowflake is approximately
1.261-dimensional.
Thats what makes the Koch snowflake a fractal the fact that
its dimension is not an integer.
Even shapes which are not self-similar can be fractals. The
most famous of these is the Mandelbrot set.
The Cantor Set
Georg Cantor
Start with a line segment of length 1.
Now cut away the middle third.
Then cut away the middle third of each remaining piece.
] 1 , 0 [
1
=
C
Iterate.
] 1 ,
3
2
[ ]
3
1
, 0 [
2
=
C
. . . . . .
|
.
|

\
|
|
.
|

\
|
= ] 1 ,
9
8
[ ]
9
7
,
3
2
[ ]
3
1
,
9
2
[ ]
9
1
, 0 [
3

C
|
|
.
|

\
|
|
.
|

\
|
=

] 1 ,
3
1
3
[ ]
3
2
3
,
3
2
[ ... ]
3
1
,
3
2
[ ]
3
1
, 0 [
1
2
1
2
2 2 1 1 n
n
n
n
n n n n
n
C

The Cantor set is whats left after youre finished cutting.
In other words:

=
=
1 n
n
C
C = Cantor set
We can ask several questions about the Cantor set, such as:
What is its cardinality?
What is its length? Indeed, does the concept of length
apply to it?
Between any two points in the Cantor set, can you
find another point in the Cantor set?
Is it complete?
What is its fractal dimension?
If you take the Cantor set . . .
. . . and dilate by a factor of k = 3 . . .
. . . then you get N = 2 copies of the original.
... 63 . 0 2 log log
3
~ = N
k
So the Cantor set is approximately 0.63-dimensional.
Diehard
The Sierpinski Carpet
Start with a square of side length 3, with a square of side
length 1 removed from its center.
perimeter = 4(3) + 4(1)
area =
1 3
2 2

Think of this shape as consisting of eight small squares, each

of side length 1.
area =
|
.
|

\
|

3
1
8
1 3
2
2 2
From each small square, remove its central square.
perimeter =
3
1
4 8 1 4 3 4 + +
Iterate.
Iterate.
1 4 3 4 + perimeter =
1 3
2 2

area =
perimeter = perimeter =
3
1
4 8 1 4 3 4 + +
|
.
|

\
|

3
1
8
1 3
2
2 2
area =
Iterate.
perimeter =
|
.
|

\
|
+ + +
3
1
4
8
3
1
4 8 1 4 3 4
2
2
area =
|
.
|

\
|

|
.
|

\
|


3
1
8
3
1
8
1 3
2 2
2
2
2 2
Iterate.
|
.
|

\
|

|
.
|

\
|

|
.
|

\
|


3
1
8
3
1
8
3
1
8
1 3
3 2
3
2 2
2
2
2 2
area =
|
.
|

\
|
+
|
.
|

\
|
+ + +
3
1
4
8
3
1
4
8
3
1
4 8 1 4 3 4
3
3
2
2
perimeter =
Iterate.
Like the Cantor set, the Sierpinski carpet is whats left after
youre finished removing everything.
In other words, its the intersection of all the previous sets.
...
3
1
4
8
3
1
4
8
3
1
4 8 1 4 3 4
3
3
2
2
+
|
.
|

\
|
+
|
.
|

\
|
+ + + perimeter =

=
|
.
|

\
|
+ =
0
3
8
4 3 4
n
n
=
...
3
1
8
3
1
8
3
1
8
1 3
3 2
3
2 2
2
2
2 2

|
.
|

\
|

|
.
|

\
|

|
.
|

\
|


area =
(
(

+
|
.
|

\
|
+
|
.
|

\
|
+ +
=
...
9
8
9
8
9
8
1
3
3 2
2
|
|
|
|
.
|

\
|

=
9
8
1
3
1
2
= 0
So the Sierpinski carpet has an infinite perimeter but
it bounds a region with an area of zero!
Weird
Your turn: compute the fractal dimension of the Sierpinski carpet.
89 . 1 8 log
3
~
The Sierpinski carpet has a 3-dimensional analogue called
the Menger sponge.
Its surface area is infinite, yet it bounds a region of zero volume.
The fractal dimension of the Menger sponge is:
73 . 2 20 log
3
~
The Mandelbrot Set
Julia set generator