I originally described the Julia/Mandelbrot algorithm as throwing a point around in a spin drier. We have Z
Z*Z, which doubles the angle and squares the range from the origin. The de Moivre theorem deals with that. After that comes a shift along a vector (the
"tumble" as in a tumble-drier) which I will tackle later.
My first test was to see what would happen if the angle simply grew by a fixed amount at each iteration. I explored that above - and there were no Fractals.
Then I questioned what might happen if the angle had the angle added to it the first time, then twice the angle the second time, then three times the angle the third time. The result would be 1 + 2 + 3 + 4 ..... n . That is
SUMMA N.
For a start, I created an image with no tumble. It should be a circle, and it is. That tests the program.
There is a tinge of colour around the edges, but nothing like as much as in the circle I showed previously.
Now I add the
tumble as a Mandelbrot vector. It
tumbles at each iteration to an extent equal to the range of the pixel from the origin, and parallel with it. The Mandelbrot vector is, after all, the pixel vector.
There are slight tufts appearing. The Fractals are dawning. I decided to choose the tuft on the left for enlargement:
A disappointment. The pattern has a slight curve to it, but there are no exciting shapes.
The program had given one multiplication by the pixel
Z, two multiplications, three multiplications and so on. This led to
(n*(n+1))/2 multiplications. It is an n-squared exponentiation of Z.
That same program can easily be converted to use the previous result, instead of the pixel
Z. So we get
Z2 followed by two multiplications of Z
2 by Z
2 giving
Z6, followed by three multiplications of Z
6 by Z
6 giving
Z24, followed by four multiplications of Z
24 by Z
24 giving
Z120.
The numbers 2, 6, 24 and 120 are
factorials.Numbers less than one shrink instantly towards the attractor at the origin. Numbers above 1 fly rapidly away. The picture is in colour, but we perceive only black-and-white because the transition from many iteration to few is so abrupt.
With the Mandelbrot
tumble added, the whole set appears. But the bits of black in the air on the left suggest that the pattern is breaking up. I explored the area on the left.
The fierce arithmetic of a factorial exponentiation of Z leads to the points being flung around the complex plane and hitting the traps too soon. The fine, feathery detail has been "singed" off. Where with Julia/Mandelbrot we had "sea-horses", here we only have Ss.
This is what happens to the arithmetic.
Consider Z
Z*Z+Z
0. We get
Z4+2Z4+Z02+Z0, and at each iteration more binomials of this kind as can be discerned from Pascal's triangle. I am interested in just the first term, the
Z4.
With linear exponentiation, beginning at 4, we get:
4
8
16
32With n-squared (specifically summa-n) exponentiation:
4
16
128
2048With exponential exponentiation (Julia/Mandelbrot):
4
16
256
65536With factorial exponentiation:
4
64
16777216
1.3292279 times 1036So Fractals seem to be dawning somewhere near the n-squared region and reach their perfection, high noon, somewhere near the exponential exponentiation of Julia and Mandelbrot. By the time the exponentiation becomes factorial the chaos mathematics is too extreme, and the sun is setting on the Fractals.
It is a pity. I had hoped to have found something new and glorious. Nevertheless, there are other approaches to consider......
Until later.
Charles