When Does a Connected Set Meet the Boundary of a Set?

Title: When Does a Connected Set Meet the Boundary of a Set? Category: mathematics

A connected set does not necessarily meet the boundary of every set. The standard theorem is conditional: if a connected set meets both a set $A$ and its complement, then it must meet the boundary of $A$. This applies in $\mathbb{R}$ with the usual topology, and more generally in any topological space.

The precise statement

Let $X$ be a topological space, let $A\subseteq X$, and let $C\subseteq X$ be connected. If

$$C\cap A\ne\varnothing \qquad\text{and}\qquad C\cap (X\setminus A)\ne\varnothing,$$

then

$$C\cap\partial A\ne\varnothing,$$

where the boundary is

$$\partial A=\overline{A}\cap\overline{X\setminus A}.$$

Here, “connected” means that $C$, with its relative topology, cannot be written as the union of two disjoint, nonempty relatively open sets. The set $C$ need not itself be an interval or be open or closed.

Proof by contradiction

Suppose instead that $C\cap\partial A=\varnothing$. A point outside the boundary lies either in the interior of $A$ or in the interior of its complement. In symbols,

$$X\setminus\partial A=\operatorname{int}(A)\cup\operatorname{int}(X\setminus A),$$

and these two interiors are disjoint. Therefore every point of $C$ belongs to one of them, since $C$ avoids the boundary.

Moreover, $C$ meets both interiors. For example, take a point of $C\cap A$. Because it is not on the boundary, it is not in $\overline{X\setminus A}$, so it lies in $\operatorname{int}(A)$. The same reasoning shows that a point of $C\cap(X\setminus A)$ lies in $\operatorname{int}(X\setminus A)$.

Consequently,

$$C=\bigl(C\cap\operatorname{int}(A)\bigr) \cup \bigl(C\cap\operatorname{int}(X\setminus A)\bigr)$$

is a union of two disjoint, nonempty sets. Each is relatively open in $C$, because it is the intersection of $C$ with an open set in $X$. This separates $C$, contradicting its connectedness. Thus $C$ must intersect $\partial A$.

Example in the real line

Let $A=(-\infty,2)$ and $C=[0,4]$. The set $C$ is connected, it contains points in $A$ (such as $1$), and it contains points outside $A$ (such as $3$). The boundary of $A$ is $\{2\}$, and indeed

$$C\cap\partial A=[0,4]\cap\{2\}=\{2\}.$$

In the special case where $C=[a,b]$ is an interval and $A$ is an open subset of $\mathbb{R}$, the result says that if one endpoint or point of the interval is in $A$ and another is outside, the interval contains a boundary point of $A$. The general proof does not require interval structure; connectedness is enough.

Why the hypotheses matter

If $C$ does not meet both sides of $A$, the conclusion can fail. For example, take $A=(0,1)$ and $C=\{2\}$. A singleton is connected, but it does not meet $\partial A=\{0,1\}$ because it lies entirely outside $A$.

Connectedness also matters. For a direct disconnected example, take $A=(-1,1)$ and $C=\{0,2\}$. Then $C$ meets both $A$ and its complement, while $\partial A=\{-1,1\}$ and $C$ misses that boundary. This shows why the connectedness assumption cannot be dropped.

Common mistakes

• Omitting the two-sided condition. Connectedness alone does not force intersection with an arbitrary boundary. You must know that $C$ meets both $A$ and $X\setminus A$.
• Treating “connected” as synonymous with “interval.” In $\mathbb{R}$, connected subsets are intervals, but the theorem is valid in general topological spaces where there is no interval description.
• Using the boundary as if it were always part of $A$. A boundary point belongs to exactly one of $A$ and its complement; it may lie in either, and the boundary need not be contained in $A$. (If $A$ is closed, it is contained in $A$.) The proof uses the boundary’s separating role, not membership in $A$.

The same argument is often summarized by saying that a connected set cannot pass from a set to its complement without crossing the set’s boundary.