Rose-Hulman Undergraduate Mathematics Journal

Volume 7 Article 9
Issue 1

On A Sequence of Cantor Fractals

Mohsen Soltanifar
K.N. Toosi University of Technology, soltanifar@sina.kntu.ac.ir

Follow this and additional works at: https://scholar.rose-hulman.edu/rhumj

Recommended Citation
Soltanifar, Mohsen (2006) "On A Sequence of Cantor Fractals," Rose-Hulman Undergraduate Mathematics
Journal: Vol. 7 : Iss. 1 , Article 9.
Available at: https://scholar.rose-hulman.edu/rhumj/vol7/iss1/9
 On A Sequence Of Cantor Fractals
Mohsen Soltanifar

Department Of Mathematics, Faculty of science

K.N. Toosi University of Technology
P.O.Box: 16315-1618, Tehran, Iran
Email: soltanifar@sina.kntu.ac.ir

April 3, 2006

Abstract
In this paper we discuss some topological and geometrical prop-
erties of terms in a sequence of Cantor fractals and the limit of the
sequence in order to obtain an exact relation between positive real
numbers and Hausdorff dimensions of fractals of Euclidean spaces.

1 Introduction
Georg Cantor(1845-1918), the founder of axiomatic set theory, studied many
interesting sets. He was very interested in infinite sets, in particular, those
with strange properties. In 1883, he published a description of a set, called
Cantor set in his own honor. The P∞Cantor set C is the set of all x ∈ [0, 1]
an
with the ternary expansion x = n=1 3n , and an = 0, 2 for all n ∈ N. This
set is one of the classical examples of fractals in the fractal geometry and is
usually referred to as Cantor fractal.
In this paper, we replace the base 3 by an arbitrary odd positive integer
other than 1, with even coefficients in expansion. For every positive integer
s, let:
∞
( )
X an
Γ(s) = x ∈ [0, 1] : x = , q = 2s + 1, an = 0, 2, 4, · · · , (q − 1)
qn
n=1

2000 Mathematics Subject Classification: 28A80.

Key words and Phrases: Cantor fractal, limits of sequences, Hausdorff dimension.

1
 We call the set above a Cantor fractal of the middle q 0 th and of the
order s. We show that for every positive integer s, the Γ(s) has the same
properties as Cantor fractal C = Γ(1).

2 Basic Topological and Geometrical Results

There is another useful way to define Γ(s), see e.g.[1],[2]. To define Γ(s) in
the unit interval [0, 1], let q = 2s + 1 for s = 1, 2, 3, · · ·; then let us consider
the affine maps:

Ts,i (x) = ( 1q )x + ( 2i
q) i = 0, 1, · · · , s

Now let Is,0 = [0, 1] and let n = 1, 2, · · ·, then we define Is,n inductively as :

Is,n = si=0 Ts,i (Is,n−1 )

S

Since Ts,i ’s are continuous and closed, they take each closed subinterval
of [0, 1] in to a closed subinterval of [0, 1], each map Ts,i sends (s + 1)n−1
disjoint closed intervals to (s + 1)n−1 disjoint closed intervals, hence, their
union which is Is,n is the disjoint union of (s + 1)n disjoint closed intervals
Is,nj = [as,nj , bs,nj ], each has a length of ( 1q )n , so the total length of Is,n
is ( s+1 n
q ) . By constructing Is,n ’s for each n S = 1, 2, · · · and i = 1, 2, · · · , n,
we have Ts,i (Is,n−1 ) ⊆ Is,n−1 , hence Is,n = si=0 Ts,i (Is,n−1 ) ⊆ Is,n−1 this
implies :

Is,0 ⊇ Is,1 ⊇ Is,2 ⊇ · · ·

and each Is,n is a closed set and thus a compact subset of [0, 1]. Since this
collection has the finite intersection property and [0, 1] is compact, they have
a nonempty intersection, then Γ(s) is defined as:

Γ(s) = ∞
T
n=0 Is,n

Since Γ(s) is an intersection of closed subsets of [0, 1], it is a bounded and

closed subset of [0, 1], thus by using the Heine-Borel Theorem in R, Γ(s) is
compact for every s = 1, 2, · · · .

In the first theorem, we obtain an explicit formula for Γ(s).

Theorem 1. Let Γ(s) be a Cantor fractal of the middle q 0 th and of the

order s, then

2
 Sqm−1 −1 Ss
Γ(s) = [0, 1]− ∞ qk+(2r−1) qk+2r
S
m=1 k=0 r=1 ( qm , qm ) (2.1)

for every s = 1, 2, 3, · · ·.

Proof . Before we introduce the two lemmas that are needed to prove this
theorem, let us have the following notation:
? =I
Is,n s,0 − Is,n
for n = 0, 1, 2, · · · and s = 1, 2, · · · .
Lemma 1. With the above definitions and notations, the equality
? = I?
S Ss ?
Is,n s,1 ( i=0 Ts,i (Is,n−1 ))

holds for every n = 1, 2, · · · and s = 1, 2, 3, · · ·.

Proof. By the definition of S ? :

Is,n
Is,n = Is,0 − Is,n = Is,0 − ( si=0 T
? Ts,i (Is,n−1 )) = si=0 (Is,0 − Ts,i (Is,n−1 )) =
T
s ? )) = si=0 (Is,0 − ?
T
Ti=0 (Is,0 − Ts,i (Is,0 − S Is,n−1 S(TSs,is (Is,0 ) − ?Ts,i (Is,n−1 ))) =
s ? ?
i=0 ((Is,0 − Ts,i (Is,0 )) Ts,i (Is,n−1 )) = Is,1 ( i=0 Ts,i (Is,n−1 )). 

Lemma 2. With the above definitions and notations, the equality:

? =
Sn Sqm−1 −1 Ss qk+(2r−1) qk+2r
Is,n m=1 k=0 r=1 ( qm , qm )
holds for n = 1, 2, · · · and s = 1, 2, · · ·.

Proof. We prove the assertion by induction. It is clear that it holds for

n = 1. Let it hold for positive integer n − 1, then:

? = I?
S Ss ?
Is,n s,1 ( i=0 Ts,i (Is,n−1 )
?
S S s S n−1 S(2i+1)q m−1 −1 Ss qk+(2r−1) qk+2r
= Is,1 ( i=0 m=1 k=2iqm−1 r=1 ( q m+1 , qm+1 ))
2iq m−1 −1 qk+(2r−1)
? ( si=1 n−1 s
, qk+2r
S S S S S S
= (Is,1 m=1 k=(2i−1)q m−1 r=1 ( q m+1 q m+1
)))
Ss Sn−1 S(2i+1)qm−1 −1 Ss qk+(2r−1) qk+2r
i=0 m=1 k=2iq m−1 r=1 ( q m+1 , qm+1 )
?
S Sn−1 Sqm −1 Ss qk+(2r−1) qk+2r
= Is,1 ( m=1 k=0 r=1 ( q m+1 , qm+1 ))
Sn Sqm−1 −1 Ss qk+(2r−1) qk+2r
= m=1 k=0 r=1 ( qm , qm ). 

Now by virtue of the Lemma 2, we prove Theorem 1:

3
Γ(s) = ∞
T T∞ ?
S∞ ?
n=0 Is,n = n=0 (Is,0 − Is,n ) = Is,0 − n=0 Is,n =
S∞ S qm−1 −1 Ss qk+(2r−1) qk+2r
[0, 1]− m=1 k=0 r=1 ( qm , qm ) for every s = 1, 2, 3, · · ·. 

In the next theorem we study some remarkable and interesting topolog-

ical properties of these fractals whose proof shows that description of these
sets can be made without using digit expansion viewpoint.

Theorem 2. The Γ(s) is

(i) nowhere dense
(ii) totally disconnected
(iii) perfect
(iv) uncountable
(v) with Lebesgue measure zero
for every s = 1, 2, 3, · · ·.

Proof. (i) To prove the nowhere densness of Γ(s) it is sufficient to show that
it does not contain any open interval (γ, δ) of [0, 1]. By using Archimedean
property, there exists positive integer m such that q −m < δ−γ 2q , which im-
qk+1
plies (2q)q −m + γ < δ. Now let k0 = min{k ∈ Z0+ | γ < q m }, hence,
qk0 +2 qk0 +2
γq m < qk0 + 1. If δ < qm , then (2q)q −m +γ < qm or (2q − 1) + γq m <
qk0 +2
qk0 + 1. This contradicts what k0 represents. Hence qm ≤ δ and thus
( qkq0m+1 , qkq0m+2 ) ⊆ (γ, δ). Now by using the equation (2.1) we conclude that
Γ(s) does not contain ( qkq0m+1 , qkq0m+2 ) which, in turn, proves that Γ(s) does
not contain (γ, δ).
(ii) Suppose x1 , x2 ∈ Γ(s) such that x1 < x2 . Since Γ(s) does not con-
tain the interval
T (xS T there exists x1 < y < x2 such that y ∈
1 , x2 ), so / Γ(s).
Now (Γ(s)T [0, y)) (Γ(s) (y, 1]) T is a disconnectedness of Γ(s) such that
x1 ∈ Γ(s) [0, y) and x2 ∈ Γ(s) (y, 1].
(iii) Since Γ(s) is closed, it is sufficient to prove that every element of
Γ(s) is its limit point. For every s = 1, 2, 3, · · · and n = 0, 1, 2, · · · put
S(s+1)n
Ls,n = j=1 {as,nj , bs,nj } and consider the set L(s) = ∞
S
n=1 Ls,n . By con-
sidering the definition of L(s), it is obvious that L(s) ⊆ Γ(s). Next we
prove that L(s) is dense in Γ(s). Let x ∈ Γ(s) and  > 0, there exists
positive integer n such that q −n < . Since x ∈ Is,n , there exists a unique
1 ≤ j ≤ (s + 1)n such that x ∈ Is,nj . Now Is.nj ⊆ (x − , x + ) implies that
as,nj , bs,nj ∈ (x − , x +T), thus Γ(s) clusters at x, so it is perfect.
(iv) Let x ∈ Γ(s) = ∞ n=0 Is,n , so x ∈ Is,n for every n ∈ N. Further more,
S(s+1)n
Is,n = j=1 Is,nj where Is,nj ’s are disjoint closed intervals. Thus, there

4
exists a unique positive integer 1 ≤ jn ≤ (s + 1)n such that T x ∈ Is,njn for
n ∈ N (in fact Is,1j1 ⊇ Is,2j2 ⊇ · · · ⊇ Is,njn ⊇ · · · and x = ∞ n=1 Is,njn ).
Now by virtue of division algorithm for each jn there exist unique integers
kn , rn such that jn = (s + 1)kn + rn where rn = 0, 1, · · · , s. Thus for each
x ∈ Γ(s) there exists a unique sequence {rn }∞ n=1 where rn = 0, 1, · · · , s for
all n ∈ N. Hence, if we put Ws = {0, 1, · · · , s}, then there exists a func-
tion Φs : Γ(s) → WsN defined by Φs (x) = {rn }∞ n=1 . Next an application of
Cantor’s intersection theorem shows the surjection of Φs , this completes the
proof.
(v) By considering the construction of Γ(s), it is clear that for every
n ∈ N, Γ(s) ⊆ Is,n . So m(Γ(s)) ≤ m(Is,n ); nevertheless, the sequence
s+1 n ∞
{( 2s+1 ) }n=0 has zero limit, thus m(Γ(s)) = 0. 

Before starting to state and proof the next Lemma, we should use a
notation. We use the symbols ind, dimH to represent the topological and
Hausdorff dimensions of sets, respectively. According to the Mandelbrot
definition, the following Lemma shows that the Γ(s) is an example of frac-
tals.

Lemma 3.
(i) indΓ(s) = 0
ln(s+1)
(ii) dimH Γ(s) = ln(2s+1)
for every s = 1, 2, 3, · · ·.
T
Proof. (i) We consider the set Λs = {Γ(s) Is,nj |n = 0, 1, 2, · · · and j =
1, 2, · · · , (s + 1)n }. The Λs constitutes an open base for Γ(s) with clopen
elements. This shows that indΓ(s) = 0.
(ii) This statement is a special case of Example 4.5 of [3]. 

For any inquisitive reader, a discussion on the convergence of the se-

quence of fractals {Γ(s)}∞
s=1 may be interesting . We consider lim Γ(s) in
s→∞
the sense employed in [4].

Theorem 3. The lim Γ(s) does not exist.

s→∞

Proof. First, Suppose that s = 1, 2, 3, · · ·, then there exists a minimum odd

positive integer s0 such that s0 ≥ s. An application of induction shows that

5
:
Tu s0
⊆ [0, 1] − ( 2u+1
t=s0 It,1 , 2u+1−s
2u+1 ) for u = s0 , s0 + 1, · · ·
0

Hence, ∞ {0, 1} so ∞
T T
t=s0 It,1 = t=s It,1 = {0,
S 1}.TNow by definition of Γ(t),
we conclude that t=s Γ(t) = {0, 1}, thus ∞
T∞ ∞
s=1 t=s Γ(t) = {0, 1} or

lim inf Γ(s) = {0, 1} (2.2)

s→∞

Secondly, suppose q = 2s + 1(s = 1, 2, · · ·). T By using induction,

T∞ it is clear
that I( q−1 ),2n ⊆ I q2 −1 n = 0, 1, 2, · · ·, so ∞ I
n=0 ( q−1
),2n ⊆ n=0 I q2 −1
2 ( 2 ),n 2 ( 2
),n
2
which implies Γ( 2 ) ⊆ Γ( q 2−1 ),
q−1
thus by virtue of induction, we obtain:

2
q −1 2m −1
Γ( q−1
2 ) ⊆ Γ( 2 ) ⊆ · · · ⊆ Γ(
q
2 ) ⊆ ··· (2.3)
for q = 3, 5, 7, · · ·.
m
S∞ S∞ q 2 −1
This implies s=1 Γ(s) = s=1 Γ( 2 ) for m = 1, 2, 3, · · ·. Next for each
2m
s = 1, 2, 3, · · ·, there exists a positive integer m such that s ≤ 3 2−1 so
m
S∞ S∞ q 2 −1
) or ∞ Γ(t) ⊇ ∞
S S S∞
S∞t=s Γ(t) ⊇ Γ(
t=1T 2 t=s s=1 Γ(s); thus t=s Γ(t) =
∞ S ∞ S ∞
s=1 Γ(s), hence, s=1 t=s Γ(t) = s=1 Γ(s) or
S∞
lim sup Γ(s) = s=1 Γ(s) (2.4)
s→∞

Now by using (2.2)and (2.4), we have lim inf Γ(s) 6= lim sup Γ(s). This proves
s→∞ s→∞
the desired result. 

As we saw in the previous theorem, the lims→∞ Γ(s) does not exist; how-
ever, we wish to discuss some properties of lim inf s→∞ Γ(s) and lim sups→∞ Γ(s).
To begin with, by virtue of equation (2.2) the discussion about lim inf s→∞ Γ(s)
is obvious and in particular it is not a fractal. Secondly, for the set lim sups→∞ Γ(s),
we have:

(i) density. The lim sups→∞ Γ(s) is dense in [0, 1]. To prove this, let
x ∈ [0, 1] − lim sups→∞ Γ(s) and  > 0 be given. There exists an odd posi-
tive integer q such that 3q < , hence there exists at least one odd positive
integer j so that as,1j ∈ (x − , x + ), since as,1j ∈ lim sups→∞ Γ(s), we

6
conclude the desired result.
(ii) nonperfectness. Since lim sups→∞ Γ(s) is a countable union of sets
with Lebesgue measure zero so it has Lebesgue measure zero. If it is perfect
in unit interval, then it has Lebesgue measure one which contradicts the
previous result.
(iii)noncompactness. The nonperfectness of lim sups→∞ Γ(s) implies that
it is not closed so, by using Heine-Borel theorem we conclude that it is not
compact(However it is locally compact).
(iv)total disconectedness. For each s = 1, 2, 3, · · ·, the [0, 1] − Γ(s) is
an open dense subset T∞ of [0, 1], hence by virtue of the Bair’s theorem, in
the version of [5], s=1 ([0, 1] − Γ(s)) = [0, 1] − lim sups→∞ Γ(s) is a dense
subset of[0, 1]. Now let x1 , x2 ∈ lim sups→∞ Γ(s) such that 0 ≤ x1 <
x2 ≤ 1, hence thereT exists Sx1 < y < x2 so T that y ∈/ lim sups→∞ Γ(s)
so (lim sups→∞ Γ(s) [0, y)) (lim sups→∞ Γ(s) (y, 1]) isT a disconnected-
ness of lim sups→∞T Γ(s) so that x1 ∈ lim sups→∞ Γ(s) [0, y) and x2 ∈
lim sups→∞ Γ(s) (y, 1].
(v)The limsups→∞ Γ(s) is a fractal. Notice that for s = 1, 2, 3, · · ·,
Γ(s) ⊆ lim sups→∞ Γ(s) and ind Γ(s) = 0, so by virtue of [6], we conclude
that ind(lim sup Γ(s)) = 0. Next, we have:
s→∞

dimH (lim sup Γ(s)) = sup (dimH Γ(s)) = 1

s→∞ 1≤s<∞

Hence, by comparing ind and dimH for this set, the proof is complete.

Notice that most fractals in the real line have Hausdorff dimension less
than 1 and every subset of the real line which contains an open subset has
Hausdorff dimension 1 . Here we constituted an example of a fractal in the
real line with Hausdorff dimension 1 which does not contain any open subset
of the real line. In fact, there is a continuum fractals of this type.
2m
In the proof of Theorem 3, we obtained the sequence of fractals{Γ( q 2−1 )}∞ m=0
for q = 3, 5, 7, · · ·. We want to find a relation between these sequences
m
q 2 −1 ∞
and the main sequence{Γ(s)}∞ s=1 . Since {Γ( 2 )}m=0 is a subsequence of
{Γ(s)}∞ s=1 , so all its elements have the same properties that we proved for
∞
elements of {Γ(s)}s=1 . Furthermore, with respect to equation (2.3) we have
2m m
q 2 −1
limm→∞ Γ( q 2−1 ) = ∞
S
m=0 Γ( 2 ). So a similar discussion that we had
2m −1
about lim sups→∞ Γ(s) still holds for the limm→∞ Γ( q 2 ), in particular, it
is a fractal with Hausdorff dimension 1.

7
3 The Hausdorff dimension Theorem
In this section we intend to obtain an exact relation between Hausdorff
dimensions of fractals of Euclidean spaces and positive real numbers. In the
previous section we showed that there are denumerable fractals with Haus-
dorff dimension 1.This result is a special case of the next Lemma.

Lemma 4. For any positive integer n, there is a continuum fractals with

Hausdorff dimension n in n-dimensional Euclidean space.

Proof . Let {qn }∞

n=1 be the increasing sequence of prime numbers, for any
2m
n ∈ N the set limm→∞ Γ( qn 2 −1 ) is a fractal with Hausdorff dimension 1.
Now for any non-empty subset A of N put:
m
S qn 2 − 1
F = n∈A ( lim Γ( )) (3.1)
m→∞ 2

All fractals of type (3.1) have all the properties of lim sups→∞ Γ(s) and
in particular, the Hausdorff dimension 1. So the uncountability of sub-
sets of N proves the Lemma for case n = 1. Next, each F is a Borel set
and has topological dimension zero, so for any n ≥ 2 taking nthQCartesian
products i=1 Fi of pairwiseQdistinct above fractals with dimH ( ni=1 Fi ) =
Qn
P n n
i=1 dimH (Fi ) = n and ind( i=1 Fi ) = 0 gives the desired result. 

In [7], it has been proved that for each 0 < r < 1, there is a con-
tinuum fractals in R with Hausdorff dimension r. Furthermore, for each
positive integer n, there is a continuum fractals with Hausdorff dimension r
in Rn (r < n). To obtain a little more exact result, we end our work with:

Theorem 4 (The Hausdorff dimension Theorem). For any real r > 0,

there is a continuum fractals with Hausdorff dimension r in n-dimensional
Euclidean space((−[−r]) ≤ n).

Proof. It is sufficient to prove the theorem for any non-integer r > 1. Since
we have r = [r] + (r) by virtue of Lemma 4 and [7] there are continuum frac-
tals with Hausdorff dimensions [r] and (r) in R[r] and R, respectively, say
G1 and G2 . Taking the Cartesian products G1 × G2 completes the proof.

8
4 Acknowledgements
I would like to thank my supervisor Dr H. P. Masiha for his invaluable
guidances. Furthermore I would like to thank Dr H. Haghighi who has
helped me with this paper in many ways. Finally, I wish to thank the Edi-
tor in chief for helpful comment.

References

[1] R. Kraft, Intersections of thick Cantor sets, Mem. Amer. Math. Soc. 97
(1992), no. 468
[2] J. Li, F.Nekka, Intersections of triadic Cantor sets with their translates.
(2). Hausdorff measure spectrum function and its introduction for the clas-
sification of Cantor sets. Chaos, Solitons & Fractals. 19(2004), no 1. 35-46
[3] K. J. Falconer. Fractal Geometry-Mathematical Foundations and Appli-
cations, wiley, chichester, 1990
[4] J. H. Dshalalow. Real Analysis-An Introduction to the Theory of Real
Functions and Integration, Chapman & Hall/CRC, Boca Ration, Florida,
2001
[5] C. C. Pugh. Real Mathematical Analysis, Springer-verlage, New-York,
Inc, 2002
[6] G. A. Edgar. Measure,Topology and Fractal Geometry, Springer-verlage,
New-York, 1992
[7] S and V.Sharapov. Dimensions of Some Generalized Cantor Sets, pre-
print:
(http://classes.yale.edu/fractals/Frac And Dim/cantorDims/CantorDims.html)