The Timpson Lyceum seal

The Timpson Lyceum · General Topology (Point Set Topology)

Limit Points & Boundaries – Lecture 11 (Topology)

Overview

This lecture develops the topological notion of a limit point and uses it to give a clean characterization of the closure. We then port the calculus idea of a limit into general topology through convergent sequences, prove the prototype result lim⁡n→∞1n=0\lim_{n\to\infty} \tfrac{1}{n} = 0 two ways (topologically and with the ε\varepsilon-style definition on R\mathbb{R}), and connect limit points to sequences in standard Rn\mathbb{R}^n. The lecture closes with Hausdorff spaces, where limits are unique, and a first look at the boundary of a set. The recurring theme is that limits, continuity, derivatives, and integrals are all special cases of ideas that topology states in full generality.

Key Concepts

Recap (interior, closure, Theorem 2.5).

Limit point. Let A⊆XA \subseteq X. A point x∈Xx \in X is a limit point of AA if every neighborhood of xx meets AA at a point other than xx: x∈A′  ⟺  ∀ Ux, (Ux∖{x})∩A≠∅.x \in A' \iff \forall\, U_x,\ \big(U_x \setminus \{x\}\big) \cap A \neq \varnothing. The set of all limit points is denoted A′A'. The only difference from being a closure point is that the shared point may not be xx itself — so a limit point need not belong to AA.

Examples (standard topology on R\mathbb{R}).

Theorem 2.8 (closure via limit points). A‾=A∪A′\overline{A} = A \cup A'.

Corollary 2.9. AA is closed   ⟺  A′⊆A\iff A' \subseteq A (a set is closed exactly when it contains all of its limit points).

Sequence. A sequence in XX is a function f:Z+→Xf : \mathbb{Z}^+ \to X, written (xn)(x_n) where xn=f(n)x_n = f(n). Order matters (parentheses, not braces).

Convergence (general topology). (xn)(x_n) converges to xx if lim⁡n→∞xn=x  ⟺  ∀ Ux ∃ N∈Z+ such that xn∈Ux ∀ n≥N.\lim_{n\to\infty} x_n = x \iff \forall\, U_x\ \exists\, N \in \mathbb{Z}^+\ \text{such that}\ x_n \in U_x\ \forall\, n \ge N. Intuition: for any neighborhood of xx, the sequence is eventually trapped inside it. On standard R\mathbb{R} this specializes (open balls) to: ∀ ε>0 ∃ N\forall\, \varepsilon > 0\ \exists\, N with ∣xn−x∣<ε|x_n - x| < \varepsilon for all n≥Nn \ge N — exactly the calculus definition.

Limit points via sequences (standard Rn\mathbb{R}^n). If xx is a limit point of AA, then there exists a sequence in AA converging to xx (take xk∈B(x,1/k)∩Ax_k \in B(x, 1/k) \cap A).

Hausdorff spaces. XX is Hausdorff if any two distinct points have disjoint neighborhoods. In a Hausdorff space a convergent sequence has a unique limit. (The trivial/chaotic topology is not Hausdorff: there every sequence converges to every point.) All standard topologies on Rn\mathbb{R}^n are Hausdorff.

Boundary. The boundary of AA is ∂A=A‾∖A∘,\partial A = \overline{A} \setminus A^\circ, the closure minus the interior — the points wedged between "in AA" and "not in AA." It is always a closed set (and in nice R2\mathbb{R}^2 sets, a closed loop), regardless of which boundary points belong to AA.

Worked Example

Claim. In standard R\mathbb{R}, lim⁡n→∞1n=0\displaystyle \lim_{n\to\infty} \frac{1}{n} = 0.

Goal: for every ε>0\varepsilon > 0, find N∈Z+N \in \mathbb{Z}^+ so that ∣1n−0∣<ε\left|\tfrac{1}{n} - 0\right| < \varepsilon for all n≥Nn \ge N. Fix ε>0\varepsilon > 0. By the Archimedean property there is a positive integer NN with N>1εN > \tfrac{1}{\varepsilon}, equivalently 1N<ε\tfrac{1}{N} < \varepsilon. For any n≥Nn \ge N, a larger denominator gives a smaller value, so 1n≤1N<ε\tfrac{1}{n} \le \tfrac{1}{N} < \varepsilon:

∣1n−0∣=1n(1n>0)≤1N(n≥N)<ε(N>1ε ⇒ 1N<ε).\begin{aligned} \left|\tfrac{1}{n} - 0\right| &= \tfrac{1}{n} && (\tfrac{1}{n} > 0) \\ &\le \tfrac{1}{N} && (n \ge N) \\ &< \varepsilon && \left(N > \tfrac{1}{\varepsilon}\ \Rightarrow\ \tfrac{1}{N} < \varepsilon\right). \end{aligned}

Thus ∣1n−0∣<ε\left|\tfrac{1}{n} - 0\right| < \varepsilon for all n≥Nn \ge N, so the sequence converges to 00. ■\blacksquare

The same engine handles other limits, e.g. lim⁡n→∞n22n2+n=12\displaystyle \lim_{n\to\infty}\frac{n^2}{2n^2 + n} = \frac{1}{2}: the +n+n is negligible against n2n^2, so the ratio is driven arbitrarily close to (but never equal to) 12\tfrac12.

Summary