Real-Line Topology Proof Mastery

From Local Openness to Global Structure

The previous tutorial emphasized a useful order of choices: first locate a point or interval, then use openness to make room around it. We now use that local reasoning to reveal the global structure of an arbitrary open subset of the real line. Although an open set may be given as a union of many intervals, its points can be grouped into maximal open intervals that do not overlap.

The proof has two distinct tasks. First, we must show that the maximal intervals exist and cover the open set without overlap. Then we must show there are at most countably many of them. Keeping these tasks separate is a useful proof technique: establish the geometric structure first, and only then use the countability of the rationals to count its pieces.

Definition: Let $U\subseteq\mathbb{R}$ be open. A maximal open interval in $U$ is a nonempty open interval $I$ such that $I\subseteq U$, and there is no open interval $J\subseteq U$ with $I\subsetneq J$. The maximal open intervals in $U$ will be called its interval components.

“Maximal” means that the interval cannot be enlarged while remaining inside $U$; it does not mean that the interval has the largest possible length among all intervals in $U$. For example, a short interval inside $U$ may be maximal if every attempt to extend it would leave $U$. The empty open set has no interval components.

Grouping Points by the Intervals Between Them

For $x,y\in U$, say that $x$ and $y$ are related when every point between them also belongs to $U$. Formally, write $x\sim y$ if the closed interval with endpoints $x$ and $y$ is contained in $U$. This condition captures the idea that there is no gap in $U$ between the two points.

Theorem (Interval-Component Decomposition): Every open set $U\subseteq\mathbb{R}$ is the union of a pairwise disjoint, at most countable collection of maximal open intervals.

Proof. If $U=\varnothing$, the collection is empty and the claim holds. Suppose $U\ne\varnothing$. We first verify that $\sim$ is an equivalence relation on $U$. It is reflexive because $x\in U$ implies $[x,x]\subseteq U$, and it is symmetric by the definition. For transitivity, suppose $x\sim y$ and $y\sim z$. The interval with endpoints $x$ and $z$ is contained in the union of the intervals with endpoints $x,y$ and $y,z$. Both of those intervals lie in $U$, so $x\sim z$.

For $x\in U$, let $C_x=\{y\in U:y\sim x\}$ be its equivalence class. This class is an interval: if $y_1,y_2\in C_x$ and $y_1<z<y_2$, then the intervals joining $x$ to $y_1$ and $x$ to $y_2$ together contain $z$ and the entire interval joining $x$ to $z$. Thus $z\sim x$, so $z\in C_x$.

The class $C_x$ is also open. Fix $y\in C_x$. Since $U$ is open, choose $\varepsilon>0$ such that $B_\varepsilon(y)\subseteq U$. If $z\in B_\varepsilon(y)$, the interval with endpoints $y,z$ lies in $B_\varepsilon(y)$, hence in $U$. The interval with endpoints $x,z$ is contained in the union of the intervals with endpoints $x,y$ and $y,z$; both lie in $U$. Therefore $z\sim x$, and $B_\varepsilon(y)\subseteq C_x$. This proves that $C_x$ is open.

Each equivalence class is therefore a nonempty open interval contained in $U$. It is maximal: if an open interval $J\subseteq U$ meets $C_x$, choose $w\in J\cap C_x$. For any $y\in J$, the interval joining $w$ to $y$ lies in $J$, and the interval joining $x$ to $w$ lies in $U$. Together they contain the interval joining $x$ to $y$, so $y\in C_x$. Hence $J\subseteq C_x$. In particular, no open interval in $U$ can properly contain $C_x$.

Equivalence classes either coincide or are disjoint, and every point of $U$ belongs to its own class. Thus these open intervals are pairwise disjoint and their union is $U$. To prove countability, enumerate the rationals as $q_1,q_2,\ldots$. Each class is a nonempty open interval, so it contains a rational by the Density of the Rationals theorem. Assign to each class the first $q_n$ in the enumeration that it contains. Distinct classes are disjoint, so they receive distinct indices. There can therefore be at most countably many classes. $\square$

The proof also identifies the role of each hypothesis. Openness makes each class open, while density of the rationals supplies a distinct label for each disjoint interval. The interval structure of the real line ensures that related points form an interval rather than a more complicated set.

Worked Example: Merging Overlapping Intervals

Let $U=(-3,1)\cup(0,4)\cup(6,8)$. The first two intervals overlap, and their union is $(-3,4)$: every point greater than $-3$ and less than $4$ belongs to at least one of them, while neither interval contains points outside that range. The third interval is disjoint from $(-3,4)$. Thus $$ U=(-3,4)\cup(6,8). $$

These two intervals are maximal in $U$. Extending $(-3,4)$ past either endpoint would include points not in $U$; extending $(6,8)$ past either endpoint would do the same. They are therefore the interval components. In particular, the original three intervals are not the components: overlapping intervals must be grouped together.

Why the Decomposition Is Unique

The theorem gives a specific collection, not merely some way to cover $U$ by disjoint intervals. To see why the maximality condition matters, suppose $I$ and $J$ are maximal open intervals contained in $U$ and that they intersect. Their union is an open interval: two intersecting intervals have no gap between their points. Since $I\cup J\subseteq U$, maximality forces $I=I\cup J$ and $J=I\cup J$, so $I=J$. Thus distinct maximal intervals cannot overlap.

Theorem (Uniqueness of the Interval Components): The collection of maximal open intervals contained in $U$ is the unique collection of pairwise disjoint maximal open intervals whose union is $U$.

Proof. Let $\mathcal{I}$ be any pairwise disjoint collection of maximal open intervals contained in $U$ whose union is $U$. Take any maximal open interval $C\subseteq U$. Choose $x\in C$. Since $\mathcal{I}$ covers $U$, there is an $I\in\mathcal{I}$ with $x\in I$. The intervals $I$ and $C$ intersect, so their union is an open interval contained in $U$. By maximality of both, $I=C$. This holds for every maximal interval $C$, so $\mathcal{I}$ is exactly the collection of interval components. $\square$

Worked Example: The Real Line with the Integers Removed

The integers form a closed set by the Locally Finite Sets Are Closed theorem: every bounded interval contains only finitely many integers, so the integers are locally finite. Therefore $U=\mathbb{R}\setminus\mathbb{Z}$ is open. For each integer $n$, the interval $(n,n+1)$ lies in $U$.

Every real number that is not an integer lies between two consecutive integers, so these intervals cover $U$. They are pairwise disjoint, and none can be enlarged within $U$, because its endpoints are integers. Hence the interval components are exactly $(n,n+1)$ for $n\in\mathbb{Z}$. This example has infinitely many components, but they can be counted using the integers, in agreement with the theorem.

Finite Endpoints Belong to the Boundary

The interval components also reveal where an open set can stop. A component may extend indefinitely in one direction, but any finite endpoint must sit just outside the open set. Points of the component approach that endpoint, so it is also in the closure of the open set.

Theorem (Finite Endpoints of Components): If $C$ is an interval component of an open set $U\subseteq\mathbb{R}$, then every finite endpoint of $C$ belongs to $\partial U$.

Proof. Let $a$ be a finite endpoint of $C$. Points of $C$ can be chosen arbitrarily close to $a$, so every neighborhood of $a$ meets $C\subseteq U$. Thus $a\in\overline{U}$. We claim that $a\notin U$. If $a\in U$, openness gives an interval around $a$ contained in $U$. That interval meets $C$, and its union with $C$ is an open interval contained in $U$ that extends past $a$. This contradicts the maximality of $C$. Hence $a\notin U$. Since $U$ is open, $\operatorname{int}(U)=U$; by the Boundary as Closure Minus Interior identity, $\partial U=\overline{U}\setminus U$. Therefore $a\in\partial U$. $\square$

Worked Example: Intervals Accumulating at Zero

Consider $$ U=\bigcup_{n=1}^{\infty}\left(\frac{1}{2n+1},\frac{1}{2n}\right). $$ Each interval is open. For successive indices, the interval for $n+1$ lies strictly to the left of the one for $n$, since $$ \frac{1}{2n+2}<\frac{1}{2n+1}. $$ Thus the displayed intervals are pairwise disjoint. Each is a maximal open interval in $U$: the gap between successive intervals contains points not in $U$, and extending past either endpoint also leaves $U$. These are therefore exactly the components.

The components accumulate toward $0$, but $0\notin U$. In fact, every neighborhood of $0$ meets $U$, because the left and right endpoints of the $n$th interval tend to $0$. Since $U$ is open, $0$ is not an interior point of $U$, and hence $0\in\partial U$. This illustrates that a boundary point can be approached by infinitely many distinct components; the decomposition does not require the components to stay a fixed positive distance apart.

A Reliable Proof Routine

When asked to describe an open subset of the real line, avoid trying to guess all its intervals at once. The equivalence-class proof gives a systematic route from individual points to the full decomposition.

Connect points with no gap.
Relate two points when the entire interval between them lies in the open set.

Verify the relation.
Check reflexivity, symmetry, and transitivity before treating its classes as groups.

Prove each class is an open interval.
Use the interval condition for the interval property and openness of the set for a neighborhood around each class point.

Establish maximality and countability separately.
Use the interval condition to rule out enlargement, then assign a rational to each disjoint component to count them.

A common mistake is to treat any interval cover as a component decomposition. Overlapping intervals may merge into a larger interval, as in the first example; only maximal intervals give the unique pieces. Another mistake is to assume that countability follows just because the set lies in $\mathbb{R}$. The proof needs a specific label for each piece, and rational density supplies one while disjointness ensures that no label is used twice.

Check Your Understanding

Use the definitions and proof techniques in this tutorial to answer the following questions.

Why does the relation “the interval between the two points lies in $U$” satisfy transitivity?
Where does the openness of $U$ enter the proof that an equivalence class is open?
Why can two distinct maximal open intervals contained in $U$ not intersect?
How does the density of the rationals give a countability argument for the interval components?
Why must a finite endpoint of a component lie outside $U$, and why does it nevertheless lie in $\overline{U}$?