Southeast-Asian J. of Sciences Vol. 2, No. 2 (2013) pp.

131-140

SOME EXAMPLES OF FINITE TYPE

FRACTALS IN THREE DIMENSIONAL
SPACES

Mai The Duy

University of Finance-Marketing (UFM)
Ho Chi Minh City, Viet- nam
e-mail: vnenlighten@yahoo.com

Abstract

By choosing the contraction functions in the Interated Function Sys-

tem we extend the construction from two dimensional spaces to three
dimensional spaces to build self-similar sets in 3-spaces. We also extend
the neighbor map concept for the attractors which have difference sizes of
sub-pieces to make them became finite type. Some interesting examples
of self similar sets in three dimensional space are given.

1 Introduction
A Fractal in general, is a rough or fragmented geometric shape that can be split
into parts, each of which is a reduced-size copy of the whole. This essential
property is called self-similarity. A fractal usually has Hausdorff dimension
which is greater than its topological dimension. Now with the aid of computer
programs, fractal geometry has recently grown and is continuing to grow and
we can visualize the beauty of many of the images that they have discovered.
Self-similar sets are a class of fractals which can be rigorously defined and
treated by mathematical methods. In 1981 Hutchinson rigorously defined self-
similar sets by this equation

F = f1 (F ) ∪ f2 (F ) ∪ ... ∪ fm (F ),

Key words: contraction functions, self similar sets, fractal geometry, Hausdorff dimension,
Neighbor maps.

131
132 Some examples of finite type fractals in three dimensional spaces

where fi , i = 1, ..., m are contracting maps (IFS) on Rd . Hutchinson proved

that for given maps there is exactly one compact non-empty set, F, which fulfils
the equation. This set F is called fractal or attractor of IFS.
Neighbor maps, which can be considered as a representation of relative po-
sition of pairs of non-empty intersecting sub-pieces. They were first introduced
by C.Bandt and S. Graf[3]. We extend the definition of Bandt on neighbor
maps and neighbor graphs. Two sub-pieces are neighbors if they intersect each
other and the relation between their sizes must be the same with the relation
between the sizes of the pieces on the original fractal. We also extend fractal
constructions from two-dimensional space to three-dimensional space. In nat-
ural, many things has the fractal structure such as fern leaf, cloud, mountains.
We try to find the self similar or self affine structure of those things. From
the geometric point of view, the interesting self-similar sets is the self-similar
sets that have the sub-pieces just touching or have exact overlap. So we have
to enlarge them to see whether they are Cantor sets or they have overlap. We
using a very strong tool that is the neighbor maps to discover that fact.

2 Magnify fractals - The neighbor maps

Figure 1: Using the neighbor maps we can magnify infinitely any self-similar
sets

Fractals are sets or entities that look the same under magnification. Small
pieces of such a set are similar to the whole set. Such sets are ”self-similar.”
To obtain a interesting structure in the self-similar sets, it is often required
that overlaps between pieces are sufficiently thin or just touching, which is
expressed by the open set condition:

Definition We say that the IF S {f1 , ..., fm} satisfies the open set condition
(OSC), if there exists an open set V such that

m
fi (V ) ⊂ V and fi (V ) ∩ fj (V ) = ∅, ∀i = j ∈ {1, ..., m}.
i=1
Mai The Duy 133

Figure 2: The relative position of sub-pieces.

We call such open set V a feasible open set of the fi , or of F .

The OSC controls the overlap of the sub-pieces Fi of fractal F . If an IFS
satisfies the open set condition then the Hausdorff dimension and the self-
similarity dimension of the attractor coincide.
However, it is not easy to check the open set condition. In 1992 Bandt and
Graf introduced an algebraic equivalent for OSC [3]. We take some notations,
let fi : Rd → Rd contractive
n similarities with contraction factor r, i ∈ I =
{1,
∞ n..., m} and F = i=1 fi (F ), I n := {(ui )i=1,...,n | ui ∈ I ∀i = 1, ..., n}, I ∗ :=
n=1 I , for u := u1 ...un ∈ I , define fu := fu1 ◦ ... ◦ fun and Fu := fu (F ).
n

Given an IFS {f1 , ..., fm}, for each u, v ∈ I ∗ , u = u1 u2 ... and v = v1 v2 ..., where
uk , vk ∈ I, k ∈ N. Let N = {h = fu−1 fv | u, v ∈ I ∗ , u1 = v1 }. The algebraic
formulation of OSC reads as following Theorem

Theorem [3] The iterated function system {f1 , ..., fm} satisfies the open set
condition if and only if there exists δ > 0 such that  h − id > δ, for all
h ∈ N.
The norm in this theorem is the norm on affine maps, which can be  g :=
A  +|b| if g = Ax + b, where
 A = max{ Ax  |x ∈ Rd with  x ≤ 1}.
In 2001, Bandt [4] described an algorithm deciding on separation, when all the
contraction factors are equal to r. The algorithm is as followed: Starting with
identity map id, we applied the automorphism
hij (g) := fi−1 .g.fj , i, j = 1, · · · , m, and i = j
Repeat this process with the obtained maps belonging to a neighborhood U of
id until all the maps run out of U . The reality of this algorithm is confirmed
by following proposition:
134 Some examples of finite type fractals in three dimensional spaces

Proposition [4, Lemma 4.1] Given similarities fi = rAi (x + ai ), where r ∈

(0, 1), and Ai are orthogonal matrices. Let U be the neighborhood of id in the
space of similarities defined as
(1 + s)c
U := {sBx + b | |b| ≤ } where c = max |ai |.
1−r i∈{1,··· ,m}

Then the complement of U is mapped into itself by each hij .

Neighbor maps, which can be considered as a representation of relative
position of pairs of non-empty intersecting sub-pieces. They were first intro-
duced by C.Bandt and S. Graf[4]. We take the sierpinski gasket to explain the
neighbor maps. Suppose that our sirpinski gasket have three sub-pieces in the
origin (see Figure 2) and the IFS have three contraction functions {f1 , f2 , f3 }.
The relative position of sub-pieces F1 and F2 were represented by the map
g = f1−1 f2 (F ), The relative position between sub-pieces F1 and F2 are the
same with the relative
 position between the fractal F and its image f1−1 f2 (F ).
Suppose that F1 F2 have very small gap (they do not touching when the
contraction factors r are smaller than 0.5, let take r = 0.499). Because the
relative position between sub-pieces F1 and F2 are the same with  −1the relative
position between the fractal F and its image f1−1 f2 (F ) then F f1 f2 (F ) also
has a gap. Because  F is larger than F1 then the gap of F f1−1 f2 (F ) is bigger
than the gap of F1 F2 . When we go to the next level; the relative position
between F12 and F21 are the same with the relative position between the fractal
−1
F and f12 f21 (F ). We continue other deeper level we have the relative position
between F12222 and F21111 are the same with the relative position between the
−1
fractal F and f12222 f21111 (F ). So we can imagine that we enlarge F12222 equal
to F so the gap between F12222 and F21111 also to be enlarged that we can see
in the Figure 1. It leads us to understand that when we apply the Bandt’s
algorithm this example then the algorithm will never stop because the gap go
bigger and bigger, that out of the neighborhood U and never come back to hij .
Bandt and Graf have give the definition of neighbor map for the maps have the
same contraction  factors. Given f1 , ..., fn a neighbor map is an element of the
set {fu−1 fv | Fu Fv = ∅, u = v ∈ I n }.
We extend the definition of Bandt on neighbor maps and neighbor graphs.
Two subpieces are neighbors if they intersect each other and the relation be-
tween their sizes must be the same with the relation between the sizes of the
pieces on the original fractal. A type is a standardized relative position of two
intersecting pieces of the fractal.

Definition: We say two pieces Fu and Fv are neighbors if

rk rk
Fu ∩ Fv = ∅, u, v ∈ I ∗ , u1 = v1 , min ≤ ru,v ≤ max .
k,l∈I rl k,l∈I rl

where u = u1 u2 ...., v = v1 v2 .... and ru,v is the contraction factor of the con-
tracting map fu−1 fv , ru,v = rruv .
Mai The Duy 135

With above definition two pieces Fu and Fv are neighbors if they intersect
each other and their sizes can be comparable that they are not big different.

Definition: A neighbor map is an element of the set

V = {fu−1 fv | Fu ∩ Fv = ∅, Fu and Fv are neighbors }

3 The neighbor graph-Finite type Fractals

The two pieces fu and Fv are neighbors, and the type of that relation between
fu and Fv are presented by the neighbor map h = fu−1 fv . Every fractal that
has a finite type neighbor map is called a finite type fractal. A type is a
standardized relative position of two intersecting pieces of the fractal and the
neighbor graph will show the relation between types. More detail we have

Definition: The neighbor graph G = (V, E) of an IF S {fi }m

i=1 is given by the
sets
V = {fu−1 fv | Fu ∩ Fv = ∅, Fu and Fv are neighbors },
rk rk
E = {(g, h, ij) ∈ V × V × I 2 | h = fi−1 gfj min ≤ rh ≤ max }
k,l∈I rl k,l∈I rl

rj
where i, j ∈ I, rh = ri ru,v .

The fractals has finite type neighbors is finite fractals. If a self-similar set
is a finite type fractal then the Bandt’s algorithm will stop after sometime and
we get the number of types. In the following part of this paper we give many
new examples of finite type fractals in three dimensional space.

4 Some three-dimensional examples - The choice

of IFS
Until now there are just a few examples of fractals in Rd with d ≥ 3. We have
two examples are well-known: the Menger sponge and the fractal tetrahedron
[19] and recently, the three-dimensional twindragon [1] and the fractal octahe-
dron, the three -dimensional modification of Sirpinski’s triangle [6]. Rendering
the pictures in this paper we use new solfware packages from Russia for three
dimensional fractals [14, 15, 16] . So that the fractal objects in three dimen-
sional space can be visualized. That could be one of the reasons make the
research on three-dimensional self-similar fractals has been increasing.
In this paper we turn fractal in plane to fractal in space by two ways: The
first way is we change the number of functions in IFS (Example 6), and the
second way is to keep the number of subpices, it means keep the number of the
functions in the IFS of both versions 2D and 3D (Examples 1 up to 5).
136 Some examples of finite type fractals in three dimensional spaces

In plane in each function if IFS we usually using one rotation matrix, when
we go to space we have to combine many rotation. In this paper we use three
3π
rotation matrices M ,M  ,M  where M is the rotation by in the X-axis
2
3π 3π
combine with the rotation by in the Y-axis and M  is the rotation by
2 2
π
in the X-axis combine with the rotation by in the Y-axis, more exactly :
2
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
0 1 0 0 −1 0 0 0 1
M = ⎣0 0 1⎦ , M  = ⎣ 0 0 1⎦ , and M  = ⎣−1 0 0⎦
1 0 0 −1 0 0 0 −1 0

Example 1. The simplest 3D fractal is the rectangular cuboid

√ √
3
F = [0, n] × [0, 3 n] × [0, n2 ]
√
with the IF S: {fk (x) = rM x + (k − 1)}, k = 1, 2, ..., n and r = 1/ 3 n

Example 2 (See Figure 1 ). The rectangular cuboid canbe made by different

sizes of subpieces: √ √
3 3
F = [0, 2] × [0 2] × [0, 4]
The IFS has three functions: √ √
f1 (x)
√ = rM x, f2 (x) = r 2 M 2 x+(1, 0, 0) and f3 (x) = r 2 M  x+(1, 3 2, 3 4) ,
r = 1/ 3 2

Figure 3: The rectangular cuboid in the Example 2

Example 3 (See Figure 4 ). The spiral fractals made of three equal sizes of
√ The IFS: f1 (x) = rM x − v,f2 (x) = −rM x − v, f3 (x) = rM x where
subpieces.
r = 1/ 3 3, v = (1, 0, 0). About the type of spiral fractals we can see the
Mai The Duy 137

Figure 4: The spiral fractal

Figure 5: Eight face neighbor types of the spiral fractal

Figure 5 and count the numbers of sub-pieces around F . When neighboring

pieces meeting in a single point or a line are neglected we can see 8 face neighbor
types. Other versions of spiral fractals with different sizes of sub pieces can see
in the Figure 6

Figure 6: The spiral fractal with three different sub-pieces

Example 4 (See Figure 6 ). The spiral fractals made of three different sizes
The IFS: f1 (x) =
of subpieces. √ √ rM x√+ v,f2 (x) = r 2 M 2 x, f3 (x) = −r 2 M 2 x + v
where r = 1/ 2, v = (−1/3, 2/3, 3 4/3) .
3 3

Example 5 (See Figure 7 ). The new Menger sponse uses only 4 contraction
138 Some examples of finite type fractals in three dimensional spaces

Figure 7: The new Menger sponse with 4 contraction functions in IF S

functions in The IFS: f1 (x) = rM x + v,f2√ 0, 0) , f3 (x) =

(x) = rM x + (2, √
2 2  2 2 
r M x + (1, 0, 0) , f4 (x) = r M x + (1, 0, 2/ 3) where r = 1/ 3 3.
3

Figure 8: The golden dodecahedron fractal

Figure 9: The 50 fixed points of the functions in IFS of the golden dodecahedron

Example 6 (See Figure 8 ). The golden dodecaheron fractal has 50 contraction

functions in IFS fk (x) = δ(x − ak ) + ak if k = 1, ..., 20, and fk (x) = δ 2 (x −
ak ) + ak , if k = 21, ..., 50, where δ is the golden ratio and ak are the vertex
points on the dodecahedron and the midpoints of the lines which connect that
vertex points (see Figure 9 ).
When cutting the golden dodecahedron fractal we have slices as we can see
in the Figure 10 and Figure 11. My friend, Ruediger suggests me that the
Mai The Duy 139

Figure 10: Two sides of cutting slides of the golden dodecahedron

Figure 11: There is a hole in the center of the golden dodecahedron fractal.

golden dodecahdron fractal contains plane segments and in the center there is
a hole. We can implies that holes exists almost everywhere, but it need tobe
proved so there is much left to explore about this golen dodecahedron fractal.

References
[1] C. Bandt and M. Mesing , Fractals of finite type, Banach Center,Publications 84 (2009),
131-148.
[2] C. Bandt, Self-Similar Sets 5. Integer Matrices and Fractal Tilings of Rn, Proc. Amer.
Math. Soc. 112 (1991), 549-562.
[3] C. Bandt and S. Graf, Self-similar sets 7. A characterization of selfsimilar fractals with
positive Hausdorff measure, Proc. Amer. Math. Soc 114 (1992), 995-1001.
[4] C. Bandt, Self-similar measures, Ergodic Theory, Analysis and Efficient Simulation of
Dynamical Systems (ed. B. Fiedler), Springer (2001), 31-46
[5] C. Bandt and N.V. Hung , Fractal n-gons and their Mandelbrot sets, Nonlinearity 21
(2008), 26532670.
[6] C.Bandt, M. Duy and M. Mesing, Three-Dimensional Fractals, Math-
ematical Intelligencer, Volume 32 (2010), Number 3, 12-18(web sites :
http://www.springerlink.com/content/w200578285042037/ )
[7] M. Duy, Some self-similar constructions in two and three dimensions and their neigh-
bor geometry, dissertation, Greifswald university 2011 (web sites: http://ub-ed.ub.uni-
greifswald.de/opus/volltexte/2011/993/)
[8] M.F. Barnsley , Fractals Everywhere, 2nd ed., Academic Press,1993.
140 Some examples of finite type fractals in three dimensional spaces

[9] J.E. Hutchinson , Fractals and self-similarity, ndiana Univ. Math.J. 30 (1981), 713747.
[10] K.J. Falconer , “Fractal Geometry. Mathematical Foundations and Applications”, Wiley
1990.
[11] H. Jones and A. Campa , Fractals Based on Regular Polygons and Polyhedra, N.M.
Patrikalakis (ed.) Scientific Vizualization of Physical Phenomena, Springer, New York,
1991, 299-314.
[12] B.B. Mandelbrot, “The Fractal Geometry of Nature”, Freeman, San Francisco, 1982.
[13] K.A. Roth, Julia sets that are full of holes, Math. Intelligencer 30, No. 4 (2008), 51-56.
[14] “Chaoscope software”, Chaoscope software developed by N. De-
sprez,http://www.chaoscope.org/.
[15] “IFS Builder 3d software”,created by Kravchenko Alexei and Mekhontsev Dmitriy, grad-
uates of Novosibirsk State University (NSU), http://fractal.nsu.ru/builder3d.
[16] “Fractracer software”,created by Mekhontsev Dmitriy, http://fractracer.com/
[17] R. Darst, J.A. Palagallo and T.E. Price, Fractal tiling in the plane, Mathematics Mag-
agine, 71 (1988), 22-23.
[18] D. Broomhead, J. Montaldi and N. Sidorov, Golden gaskets: variations on the Sierpinski
sieve, Nonlinearity 17 (2004), 1455-1480
[19] http://www.mathpaint.blogspot.com