Metric Spaces Are Hausdorff Spaces

Every metric space is a Hausdorff space. If a topological space is not Hausdorff, it cannot arise from a metric.

The property of being a Hausdorff space ensures that for any two distinct points, there exist two disjoint neighborhoods separating them.

In simpler terms, this means that in a metric space (a space where distances between points are well-defined), we can always find open sets that separate any two distinct points.

Note. The Hausdorff property must hold for every pair of distinct points in the space, not just for some.

A Practical Example

Consider the Euclidean plane \(\mathbb{R}^2\) with the standard Euclidean distance, denoted \(d(x, y)\), where \(x = (x_1, x_2)\) and \(y = (y_1, y_2)\) are points in the plane. The distance is defined as:

$$ d(x, y) = \sqrt{(x_1 - y_1)^2 + (x_2 - y_2)^2}. $$

In this context, \(\mathbb{R}^2\) with the Euclidean distance forms a metric space.

Metric spaces like \(\mathbb{R}^2\) satisfy the Hausdorff property, meaning we can "separate" any two distinct points using disjoint open sets.

Take two points \( A = (x_1, y_1) \) and \( B = (x_2, y_2) \) in \(\mathbb{R}^2\) with \(A \neq B\).

The Euclidean distance between \(A\) and \(B\) is \(d(A, B) > 0\) because the points are distinct.

Now, choose a radius smaller than half the distance between \(A\) and \(B\):

$$ r = d(A, B) / 2 $$

Construct two circles with radius \(r\):

  • \( U = \{ P \in \mathbb{R}^2 : d(P, A) < r \} \), the set of points within \(r\) of \(A\),
  • \( V = \{ P \in \mathbb{R}^2 : d(P, B) < r \} \), the set of points within \(r\) of \(B\).

Notice that \(U\) and \(V\) do not overlap:

$$ U \cap V = \varnothing $$

This happens because every point in \(U\) is closer to \(A\) than \(B\), and every point in \(V\) is closer to \(B\) than \(A\).

Since this reasoning applies to any two distinct points \( A \) and \( B \) in \(\mathbb{R}^2\), we can conclude that \(\mathbb{R}^2\) with the Euclidean distance satisfies the Hausdorff property.

Thus, \(\mathbb{R}^2\) is a Hausdorff space.

Example 2

Consider the set \(\mathbb{R}\) of real numbers with the finite complement topology.

In the finite complement topology, a set \(U \subseteq \mathbb{R}\) is open if \(U\) is empty (\(\varnothing\)) or if its complement \(\mathbb{R} \setminus U\) is finite (i.e., it contains only finitely many elements).

In other words, a set is open if it includes "almost all" points of \(\mathbb{R}\), except for a finite number.

Now, take two distinct points \(x, y \in \mathbb{R}\).

Try to find disjoint open sets containing \(x\) and \(y\).

Let \(U\) be an open set containing \(x\). To be open in the finite complement topology, the complement of \(U\) (\(\mathbb{R} \setminus U\)) must be finite, meaning \(U\) must contain "almost all" points of \(\mathbb{R}\), except for a finite number.

Similarly, let \(V\) be an open set containing \(y\), with its complement \(\mathbb{R} \setminus V\) also finite.

Since both \(U\) and \(V\) contain "almost all" points of \(\mathbb{R}\), their intersection \(U \cap V\) cannot be empty because both sets must include infinitely many shared points.

Note. For example, take \(x = 1\) and \(y = 2\). Try to "separate" \(x\) and \(y\) using disjoint open sets in the finite complement topology.

  • Define an open set \(U\) containing \(x = 1\). For \(U\) to be open in this topology, it must include all points of \(\mathbb{R}\) except a finite number. For example, exclude \(y = 2\) and a few other nearby points. Thus, \(U\) will contain almost all real numbers: $$ U = \mathbb{R} \setminus (2-\epsilon, 2+\epsilon) $$
  • Define an open set \(V\) containing \(y = 2\). Similarly, \(V\) must include all points of \(\mathbb{R}\) except a finite number. For instance, exclude \(x = 1\) and a few nearby points: $$ V = \mathbb{R} \setminus (1-\epsilon, 1+\epsilon) $$

Because of how open sets are defined, both \(U\) and \(V\) include "almost all" points of \(\mathbb{R}\), so their intersection \(U \cap V\) cannot be empty. It will contain infinitely many points: $$ U \cap V = \mathbb{R} \setminus \left[(2 - \epsilon, 2 + \epsilon) \cup (1 - \epsilon, 1 + \epsilon)\right] \ne \emptyset $$ This shows that in \(\mathbb{R}\) with the finite complement topology, it is impossible to find disjoint open sets that separate \(x = 1\) and \(y = 2\).

Since we cannot find disjoint open sets \(U\) and \(V\) around two distinct points \(x\) and \(y\), the space \((\mathbb{R}, \text{finite complement topology})\) is not a Hausdorff space.

As a result, this topological space cannot be represented by a metric space.

Proof

Consider two distinct points \( x \) and \( y \) in a metric space \((X, d)\).

Since they are distinct, the two points are separated by some positive distance \(\varepsilon > 0\).

Define two open spherical neighborhoods, or open balls, around \(x\) and \(y\) that do not overlap.

Let these neighborhoods be \( U \) and \( V \):

  • \( U \) is the set of points less than \(\varepsilon/2\) away from \(x\),
  • \( V \) is the set of points less than \(\varepsilon/2\) away from \(y\).

We now show that \(U\) and \(V\) do not overlap.

Assume, for contradiction, that there is a point \(z\) belonging to both \(U\) and \(V\). This would mean:

  • \( z \) is within \(\varepsilon/2\) of \(x\),
  • \( z \) is within \(\varepsilon/2\) of \(y\).

By adding these distances and using the triangle inequality, the distance between \( x \) and \( y \), \( d(x, y) \), would be less than \(\varepsilon\):

$$ d(x,y) < d(x,z) + d(z,y) < \epsilon/2 + \epsilon/2 = \epsilon $$

This contradicts our initial assumption that \( d(x, y) = \varepsilon\).

Thus, no point can belong to both \(U\) and \(V\).

This proves that in a metric space, we can always find disjoint open sets \( U \) and \( V \) around \( x \) and \( y \).

Therefore, every metric space is Hausdorff.

And so on.

 
 

Please feel free to point out any errors or typos, or share suggestions to improve these notes. English isn't my first language, so if you notice any mistakes, let me know, and I'll be sure to fix them.

FacebookTwitterLinkedinLinkedin
knowledge base

Metric Topology