In the Principles of Mathematical Analysis, 3rd Ed. by Rudin, he mentions in a Theorem 2.43 on page 41 that every non-empty perfect set $P$ in $\mathbb{R}^{k}$ is uncountable.

In the proof, he assumes contrariwise that the points of $P$ are denumerable and labelled $\overline{x_{1}}, \overline{x_{2}}, \overline{x_{3}}, ...$ . He then constructs a sequence of neighborhoods $\{V_{n}\}$ by induction as follows:

Let $V_{1}$ be the neighborhood of $\overline{x_{1}}$, defined as $V_{1} = \{\overline{y} \in \mathbb{R}^{k} : |\overline{x_{1}} - \overline{y}| < r\}$, and the closure of $V_{1}$ denoted $\overline{V_{1}}$ defined $\overline{V_{1}} = \{\overline{y} \in \mathbb{R}^{k} : |\overline{x_{1}} - \overline{y}| \leq r\}$.

Now, suppose $V_{n}$ has been constructed, with $V_{n} \cap P \neq \emptyset$. Then, there exists $V_{n+1}$ such that:

  1. $\overline{V_{n+1}} \subset V_{n}$ (the closure of $V_{n+1}$ is a proper subset of $V_{n}$)
  2. $\overline{x_{n}} \notin \overline{V_{n+1}}$
  3. $\overline{V_{n+1}} \cap P \neq \emptyset$.

The third condition ensures our induction hypothesis holds and allows for the construction to continue.

Now, let $K_{n} = \overline{V_{n}} \cap P$. Since $\overline{V_{n}}$ is closed and bounded, it must be compact as well (proved earlier). Since $\overline{x_{n}} \notin K_{n+1}$ by construction, no point of $P$ is in $\cap_{1}^{\infty}K_{n}$. Since $K_{n} \subset P$, this implies that $\cap_{1}^{\infty}K_{n} = \emptyset$. This fact in turn contradicts a corollary of an earlier theorem that such an intersection must be non-empty.

Where my understanding falls short is in the statement that is in bold. The contradiction relies on it. Now, since since $P$ is perfect, $P$ is closed and every point of $P$ must be a limit point of $P$. Further, since $P$ is non-empty, it must have infinitely many points (or else, it cannot have any limit points). Now, when I take a point $x_{n+1}$ inside the neighborhood $V_{n}$ of $x_{n}$, and construct a neighborhood $V_{n+1}$ in a manner that doesn't include $x_{n}$, $V_{n+1}$ still contains infinitely many points; it just doesn't contain $x_{n}$. So, when I continue this induction, doesn't each subsequent neighborhood $V_{m}$ have points in it? And, when I take an infinite intersection as suggested, it needn't contain the specific labelled points at the centre of the neighborhood; but it still does contain infinitely many points.

Where am I going wrong?

For some visual context, I am visualizing my argument as a set of circles with each successive circle (denoting $V_{n}$) drawn within the previous one but drawn so that it doesn't contain the centre of the previous one.


1 Answer 1


You are making things too complicated. $\overline {x_n} \notin K_{n+1}=\overline {V_{n+1}} \cap P$ by property 2). If $\overline {x_i}$ is a point of $P$ which belongs to every $K_n$ then $\overline {x_i} \in K_{i+1} \subset \overline {V_{i+1}} $ again contradicting 2).

  • $\begingroup$ I see your point, sir. Could you explain then where my circles analogy fails? It is true that if I consider some point $x_{i}$, then, $K_{i+1}$ must not contain it by construction. However, $K_{i+1}$ will still have an infinite number of points in common with its superset, right? $\endgroup$ Nov 27, 2020 at 9:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.