The boundary between the black and white regions in the figure below would touch every point on a square if the generating algorithm were repeated ad infinitum.
As a topological torus (top-left edge abutting bottom-right, and top-right to bottom left), the black fractal pattern is identical to the white fractal pattern, but for a 180 degree rotation and small shift. The boundary between the black and white regions is then connected, forming a closed curve (on a 2-D surface embedded in a 3-D space) that separates the two regions.
The boundary is like other space-filling curves, such as the Peano Curve, Hilbert Curve, and Z-order curve, and can be generated by an L-System (Lindenmayer System, a form of iterative string replacement). These curves, and some other space-filling curves, can also divide a square and can be closed on a topological torus. With the Peano and Z-order curves the two regions on a topological torus can also be identical in all respects except for a 180 degree rotation. I suspect there are many 2-D Lindenmayer System figures that have this property. But, curiously, it appears to me that the Hilbert curve can't be closed such that there are two identical regions -- on a square, cylinder or torus.
Comments
u call this SFC ?!!
re Space filling curve 1 Fri, 2007-08-31 19:08 — Mark Dow --
hi Mark Dow
with all due respect i cannot see a curve or a path.
fractal ?
may be.
dendrite ?
probably.
is a dendrite curve ?
-
signed:
lsystems_fractals
curve and boundary
Space-filling curve? Yes.
Note that I refer to the boundary between these regions -- NOT to the black or white regions -- in the infinite limit (where the black an white regions become infinitessimally thin). Yes, the black and white regions are not continuous curves (can't be traced without backtracking over the same coordinates) but are dendritic, in the same way that the set of points not on a Hilbert or Peanno curve are dendritic.
Fractal? Absolutely, in the exactly the same sense that a Hilbert or Peanno curve is fractal.
Note that this image will be repeated as each quadrant of the next generation (the pattern is strictly self-similar), and the boundary (in the infinite limit) is nowhere differentiable.
Post new comment