logo

Cantor Set 📂Topology

Cantor Set

Definition

g5ZtX.png

Let I=[0,1]C1=[0,13][23,1]C2=[0,132][232,332][632,732][832,1]Cn=[0,13n][23n,33n][3n33n,3n23n][3n13n,1] \begin{align*} I =& \left[ 0, 1 \right] \\ C_{1} =& \left[ 0, {{1} \over {3}} \right] \cup \left[ {{2} \over {3}} , 1 \right] \\ C_{2} =& \left[ 0, {{1} \over {3^2}} \right] \cup \left[ {{2} \over {3^2}}, {{3} \over {3^2}} \right] \cup \left[ {{6} \over {3^2}}, {{7} \over {3^2}} \right] \cup \left[ {{8} \over {3^2}} , 1 \right] \\ &\vdots \\ C_{n} =& \left[ 0, {{1} \over {3^n}} \right] \cup \left[ {{2} \over {3^n}}, {{3} \over {3^n}} \right] \cup \cdots \cup \left[ {{3^n-3} \over {3^n}}, {{3^n-2} \over {3^n}} \right] \cup \left[ {{3^n - 1} \over {3^n}} , 1 \right] \end{align*} .

C:=n=1Cn\displaystyle C := \bigcap_{n=1}^{\infty} C_{n} is called the Cantor Set.

Theorem

  • [1]: C={xI  x=0.x1x2,xi{0,2}}C = \left\{ x \in I \ | \ x= 0.x_{1} x_{2} \cdots , x_{i} \in \left\{ 0,2 \right\} \right\}
  • [2]: CC is an uncountable set.

Explanation

The Cantor set is defined simply but appears as an important example in real analysis.

Theorem [1] means that every element of the Cantor set can be expressed using only 00 and 22 in its ternary expansion. No need to prove it specifically, but if you think about it, it isn’t too hard to understand.

Moreover, adding up the lengths of all intervals results in limn(23)n=0\displaystyle \lim_{n \to \infty} \left( {{2} \over {3}} \right)^{n} = 0, which is quite interesting because it is an uncountable set. This uncountability is usually proven by applying Cantor’s diagonal argument to theorem [1], showing that there is no one-to-one correspondence between CC and the set of natural numbers using ternary representation, which doesn’t provide much learning in the process.

I’ve summarized a more topology-oriented proof for reference.

Proof

[2]

Part 1.

It is trivially true that CRC \subset \mathbb{R}, thus it can be a metric space. Since CC is defined as an intersection of closed intervals, it is a closed set.

Heine-Borel Theorem: For ERnE \subset \mathbb{R}^{n}, a necessary and sufficient condition for EE to be compact is that EE is bounded and closed.

Meanwhile, since C[0,1]C \subset [0,1], it is bounded, and by the Heine-Borel theorem, CC is a compact space. Being compact in a metric space is equivalent to being complete and totally bounded, hence, CC is a complete metric space.

Baire Category Theorem: Every complete metric space is a Baire space.

By the Baire category theorem, CC is a Baire space.


Part 2.

For any cCc \in C and ε>0\varepsilon>0, B(c,ε)B(c, \varepsilon ) contains at least one point that is not cc.

This is described as CC having no isolated points, which is mathematically expressed as (C{c})B(c,ε) \left( C \setminus \left\{ c \right\} \right) \cap B(c, \varepsilon ) \ne \emptyset Therefore, removing just one point from the whole, C{c}C \setminus \left\{ c \right\} is a dense set in CC.


Part 3.

Assuming that CC is a countable set, it could be represented by C={cn  nN}C = \left\{ c_{n} \ | \ n \in \mathbb{N} \right\}.

Definition of a Baire space: For every dense open set sequence {On}n=1\left\{ O_{n} \right\}_{n=1}^{\infty}, n=1On\displaystyle \bigcap_{n=1}^{\infty} O_{n} makes the space a Baire space if it is dense.

Since Part 1 showed that CC is a Baire space and Part 2 showed that C{cn}C \setminus \left\{ c_{n} \right\} is a dense set, n=1(C{cn})\displaystyle \bigcap_{n=1}^{\infty} \left( C \setminus \left\{ c_{n} \right\} \right) is dense. However, n=1(C{cn})=\displaystyle \bigcap_{n=1}^{\infty} \left( C \setminus \left\{ c_{n} \right\} \right) = \emptyset and C\overline{ \emptyset } \ne C, so it is not a dense set. This contradiction means that CC is an uncountable set.