logo

In a Metric Space, Compact Implies Closed and Bounded 📂MetricSpace

In a Metric Space, Compact Implies Closed and Bounded

Theorem

A compact subset KK of a metric space (X,d)(X, d) is bounded and a closed set.

Description

The converse does not generally hold. In Euclidean spaces, the converse holds.

Proof

Boundedness1

By contradiction, assume that KK is not bounded. Since being compact in a metric space is equivalent to being sequentially compact, KK is sequentially compact.

Sequentially Compact

A metric space XX being sequentially compact means that for every sequence {xn}\left\{ x_{n} \right\} in XX, there exists a convergent subsequence {xnk}\left\{ x_{n_{k}} \right\} that converges to a point in XX.

Since KK is unbounded, there exists an unbounded sequence {ynM}\left\{ y_{n} \in M \right\} satisfying d(yn,b)>nd(y_{n}, b) \gt n for a fixed bXb \in X. As KK is sequentially compact, this sequence has a converging subsequence, which also must be unbounded due to the property {yn}\left\{ y_{n} \right\}. However, in a metric space, a converging sequence is bounded, which leads to a contradiction. Thus, the assumption is wrong, and KK is bounded.

Closedness2

Let KK be a compact subset of the metric space XX. To show that KcK^{c} is open, it suffices to prove that every point in KcK^{c} is an interior point. Let us now consider pKcp \in K^{c}. And let VqV_{q} and WqW_{q} respectively be neighborhoods of radius smaller than 12d(p,q){\textstyle \frac{1}{2}}d(p,q) for pp and qq. VV is a neighborhood of pp, but we use the index qq because the radius of pp is determined by the distance to qq whenever a particular qKq\in K is given for a fixed pKcp\in K^{c}. If the text is not clear, see the diagram below.

5F3CC9892.png

Now consider {Wq}qK\left\{ W_{q} \right\}_{q\in K}. This forms an open cover of KK. By assumption, since KK is compact, there exists some q1,,qnq_{1},\cdots,q_{n} that satisfies the equation below.

KWq1Wqn=W K \subset W_{q_{1}}\cup \cdots \cup W_{q_{n}}=W

And let’s call it V=Vq1VqnV=V_{q_{1}}\cap \cdots \cap V_{q_{n}}. Then, VV is still a neighborhood of pp. Also, because we initially chose the distances VqV_{q}, WqW_{q} carefully, we know that VW=V\cap W=\varnothing holds. Therefore, since WW is an open cover of KK, the following is true.

VKc V \subset K^{c}

Since every pKcp\in K^{c} always has a neighborhood contained in KcK^{c}, every pKcp \in K^{c} is an interior point of KcK^{c}. Therefore, KcK^{c} is an open set, and KK is a closed set.


  1. Erwin Kreyszig, Introductory Functional Analysis with Applications (1978), p77 ↩︎

  2. Walter Rudin, Principles of Mathmatical Analysis (3rd Edition, 1976), p37 ↩︎