11.2 Connectedness, Compactness, Countability and Separation
The properties that tell spaces apart: connectedness and components, compactness and its interplay with the Hausdorff property, the countability axioms and separability, the separation ladder $T_0$ to $T_4$ with its standard counterexamples, and Urysohn's lemma with its uses.
Section 11.1 produced a large supply of spaces: the line with its standard, Sorgenfrey, cofinite and cocountable topologies, products and boxes, gluings of a square. To tell them apart — and to know which theorems of analysis survive in them — we need topological properties, properties defined through open sets alone and therefore preserved by homeomorphisms. The intermediate value theorem is really a statement about connectedness, and the extreme value theorem a statement about compactness; both turn out to hold for continuous maps on any space with the property. The countability axioms say how many open sets are needed to describe a space, and decide whether sequences are enough. The separation axioms say how well open sets can tell points and closed sets apart; at the top of the ladder, Urysohn's lemma manufactures continuous real functions out of pure topology, the first step towards deciding which spaces come from a metric.
GATE MA asks almost exclusively "which of the following is true" and "how many" questions here: is a space connected, compact, separable, first or second countable, Hausdorff, regular, normal, metrizable; how many components does a set have; can two spaces be homeomorphic; what does Urysohn's lemma give. The theorems are short and the counterexamples decisive, so this section states every result with its hypotheses, proves the ones that carry the weight, and collects a table of spaces that separate the properties. Metric-space versions of connectedness and compactness are in §4.1; here they are developed for general topological spaces.
Connectedness
For a space $X$ the following are equivalent:
- $X$ is connected;
- the only clopen subsets of $X$ are $\varnothing$ and $X$;
- every continuous map from $X$ to the discrete space $\{0,1\}$ is constant.
Moreover, if $U,V$ is a separation of $X$ and $Y\subseteq X$ is connected, then $Y\subseteq U$ or $Y\subseteq V$.
Proof. (1)⇒(2): a proper nonempty clopen set $C$ gives the separation $C$, $X\setminus C$. (2)⇒(3): if $f:X\to\{0,1\}$ is continuous, $f^{-1}(0)$ is clopen (the preimage of a clopen set), hence $\varnothing$ or $X$. (3)⇒(1): a separation $U,V$ gives the continuous non-constant map $\mathbf{1}_V$. For the last statement, $Y\cap U$ and $Y\cap V$ are disjoint open subsets of $Y$ covering it; if both were nonempty they would separate $Y$. ∎
- A subset of $\mathbb{R}$ (standard topology) is connected if and only if it is an interval.
- Continuous images of connected spaces are connected. (Hence the intermediate value theorem: a continuous real function on a connected space takes every value between any two of its values.)
- If connected sets $A_\alpha$ all contain a common point (or merely each meets a fixed connected set $A_0$), their union is connected.
- If $A$ is connected and $A\subseteq B\subseteq\overline{A}$, then $B$ is connected; in particular $\overline{A}$ is connected.
- A finite product of connected spaces is connected. $\mathbb{R}^\omega$ with the product topology is connected; with the box topology it is not.
- Path-connected spaces (§4.1) are connected; the converse fails (the topologist's sine curve), but open connected subsets of $\mathbb{R}^n$ are path-connected.
Proof. (1) If $Y$ is not an interval, there are $a\lt c\lt b$ with $a,b\in Y$, $c\notin Y$, and $Y\cap(-\infty,c)$, $Y\cap(c,\infty)$ separate $Y$. Conversely let $I$ be an interval and suppose $I=U\cup V$ is a separation with $a\in U$, $b\in V$, say $a\lt b$. Then $[a,b]\subseteq I$; let $s=\sup\big(U\cap[a,b]\big)$. If $s\in U$, then $s\lt b$ and, $U$ being open in $I$, a little interval $[s,s+\delta)$ lies in $U\cap[a,b]$, contradicting the supremum. If $s\in V$, then $s\gt a$ and some $(s-\delta,s]$ lies in $V$, so $s-\delta$ would already be an upper bound of $U\cap[a,b]$. Both are impossible.
(2) If $f:X\to Y$ is continuous and $f(X)=U\cup V$ were a separation of $f(X)$, then $f^{-1}(U),f^{-1}(V)$ would separate $X$. For the intermediate value theorem, $f(X)\subseteq\mathbb{R}$ is connected, hence an interval by (1).
(3) Let $Y=\bigcup A_\alpha$ and $g:Y\to\{0,1\}$ continuous. On each connected $A_\alpha$ the map $g$ is constant (Theorem 11.15), and the constants agree because the sets share a point (or because each $A_\alpha$ meets $A_0$, on which $g$ is constant). So $g$ is constant.
(4) If $g:B\to\{0,1\}$ is continuous, it is constant on $A$, say $g=0$ there. Then $g^{-1}(0)$ is closed in $B$ and contains $A$, so it contains $\overline{A}\cap B=B$.
(5) For $X\times Y$: fix $(a,b)$. For each $x$, the "cross" $(\{x\}\times Y)\cup(X\times\{b\})$ is connected by (3) (two connected sets through $(x,b)$, each homeomorphic to $Y$ or $X$), and all crosses contain $(a,b)$; their union is $X\times Y$. Induction gives finite products. In $\mathbb{R}^\omega$ with the product topology, the sequences that are eventually $0$ form a union of the connected sets $\mathbb{R}^n\times\{0\}\times\{0\}\times\cdots$, all containing $0$, and this union is dense (Exercise IV.5 of §11.1); apply (4). In the box topology the bounded sequences $A$ and the unbounded sequences $B$ separate $\mathbb{R}^\omega$: for any $x$, the box $\prod_n(x_n-1,x_n+1)$ consists of sequences that are bounded exactly when $x$ is, so both $A$ and $B$ are open.
(6) If $X$ is path-connected and $U,V$ separate it, a path from a point of $U$ to a point of $V$ would be a connected subset $\gamma([0,1])$ meeting both — impossible by Theorem 11.15. The other two facts are proved in §4.1. ∎
Connectedness depends on the topology. $\mathbb{R}_\ell$ is disconnected — $(-\infty,0)$ and $[0,\infty)$ separate it — and in fact totally disconnected: a subset with two points $a\lt b$ is split by $[a,b)$ and its complement. A discrete space with two or more points is disconnected. On the other hand, in the indiscrete topology, in $\mathcal{T}_{\mathrm{cof}}$ and $\mathcal{T}_{\mathrm{coc}}$ on infinite (uncountable) sets, in the right-ray topology and in the Sierpiński space, any two nonempty open sets meet, so no separation exists and these spaces are connected.
- The components of $X$ are connected, closed, pairwise disjoint and cover $X$; every connected subset lies in exactly one component. Each path component lies inside a component.
- If $X$ has finitely many components, each of them is also open (so clopen), and $X$ has exactly $2^k$ clopen subsets, where $k$ is the number of components.
- A homeomorphism $h:X\to Y$ maps components onto components, and $X\setminus\{p\}\cong Y\setminus\{h(p)\}$. Hence the number of components, and the number of components of $X\setminus\{p\}$ for each $p$, are topological invariants.
Proof. (1) The component $C_x$ is a union of connected sets containing $x$, hence connected (Theorem 11.16(3)); $\overline{C_x}$ is connected (11.16(4)) and contains $x$, so $\overline{C_x}\subseteq C_x$, i.e. $C_x$ is closed. If $C_x\cap C_y\ne\varnothing$, then $C_x\cup C_y$ is connected and contains $x$, so $C_y\subseteq C_x$, and symmetrically. A connected set containing $x$ lies in $C_x$ by definition. A path component is the union of the connected path images through $x$. (2) The complement of a component is the union of the finitely many others, a closed set, so each component is open. A clopen set $D$ is a union of components (a component meeting $D$ lies in $D$ by Theorem 11.15), and any union of components is clopen; there are $2^k$ choices. (3) Homeomorphisms preserve connectedness in both directions, and $h$ restricts to a homeomorphism $X\setminus\{p\}\to Y\setminus\{h(p)\}$. ∎
Compactness
- A closed subset of a compact space is compact.
- In a Hausdorff space (Definition 11.18: distinct points have disjoint open neighbourhoods), a compact set $K$ and a point $x\notin K$ have disjoint neighbourhoods; hence compact subsets of Hausdorff spaces are closed.
- Continuous images of compact spaces are compact. A continuous $f:X\to\mathbb{R}$ on a nonempty compact space is bounded and attains its maximum and minimum.
- A continuous map from a compact space to a Hausdorff space is closed; a continuous bijection from a compact space onto a Hausdorff space is a homeomorphism.
- A finite product of compact spaces is compact (tube lemma). Tychonoff's theorem: any product of compact spaces is compact in the product topology.
- Heine–Borel: a subset of $\mathbb{R}^n$ is compact if and only if it is closed and bounded.
- $X$ is compact if and only if every family of closed sets with the finite intersection property (every finite subfamily has nonempty intersection) has nonempty intersection.
Proof. (1) If $C$ is closed and $\{U_\alpha\}$ covers $C$, add $X\setminus C$ to get an open cover of $X$; a finite subcover, with $X\setminus C$ removed, covers $C$.
(2) For each $k\in K$ choose disjoint open $U_k\ni x$ and $V_k\ni k$. Finitely many $V_{k_1},\dots,V_{k_m}$ cover $K$; then $U=\bigcap_iU_{k_i}$ and $V=\bigcup_iV_{k_i}$ are disjoint open sets with $x\in U$ and $K\subseteq V$. So every point outside $K$ has a neighbourhood missing $K$: $K$ is closed.
(3) If $\{V_\alpha\}$ covers $f(X)$, then $\{f^{-1}(V_\alpha)\}$ covers $X$; a finite subcover gives finitely many $V_\alpha$ covering $f(X)$. A compact subset of $\mathbb{R}$ is closed and bounded (by (6), or directly: the cover by $(-n,n)$ shows boundedness and (2) closedness), so it contains its supremum and infimum.
(4) A closed $C\subseteq X$ is compact by (1), $f(C)$ is compact by (3), and closed by (2). A continuous closed bijection has a continuous inverse.
(5) (Tube lemma, sketch.) Let $X$, $Y$ be compact and $\mathcal{W}$ an open cover of $X\times Y$ by basic boxes. For fixed $x$, the slice $\{x\}\times Y\cong Y$ is covered by finitely many boxes $U_i\times V_i$; then $U_x=\bigcap_iU_i$ is open, contains $x$, and the "tube" $U_x\times Y$ is covered by these finitely many boxes. Finitely many $U_x$ cover $X$, so finitely many boxes cover $X\times Y$. Tychonoff's theorem for infinite products needs the axiom of choice and is only stated here.
(6) A closed bounded set lies in a cube $[-M,M]^n$, which is compact by (5) once $[-M,M]$ is (proved in §4.1), and it is closed in the cube, so (1) applies. Conversely a compact set is closed by (2) and bounded since the balls $B(0,n)$ cover it. (7) Take complements: an open cover without a finite subcover is the same as a family of closed sets with the finite intersection property and empty intersection. ∎
Compactness, too, depends on the topology. In $\mathcal{T}_{\mathrm{cof}}$ every subset is compact: one member $X\setminus F$ of a cover misses only the finitely many points of $F$, each of which needs one more member. In $\mathcal{T}_{\mathrm{coc}}$ on $\mathbb{R}$ the compact sets are exactly the finite sets: if $\{x_1,x_2,\dots\}\subseteq K$ are distinct, the sets $U_n=\mathbb{R}\setminus\{x_k:k\ne n\}$ are open (cocountable), cover $\mathbb{R}$, and each contains only one $x_k$. In a discrete space compact means finite, and an indiscrete space is compact. In $\mathbb{R}_\ell$ the interval $[0,1]$ is not compact (Example 3), and one can show that every compact subset of $\mathbb{R}_\ell$ is countable.
Countability Axioms and Separability
A local basis at $x$ is a family $\mathcal{B}_x$ of open sets containing $x$ such that every open set containing $x$ contains a member of $\mathcal{B}_x$. A space is
- first countable if every point has a countable local basis;
- second countable if the topology has a countable basis;
- separable if it has a countable dense subset;
- Lindelöf if every open cover has a countable subcover.
- Second countable $\Rightarrow$ first countable, separable and Lindelöf.
- Every metric space is first countable. For a metric space: second countable $\iff$ separable $\iff$ Lindelöf.
- In a first countable space, $x\in\overline{A}$ if and only if some sequence in $A$ converges to $x$; and a map out of a first countable space is continuous if and only if it preserves limits of sequences.
- Subspaces of first (second) countable spaces are first (second) countable; countable products of first (second) countable spaces are first (second) countable. Open subspaces of separable spaces are separable, and closed subspaces of Lindelöf spaces are Lindelöf — but separability does not pass to closed subspaces, and the Lindelöf property does not pass to products.
Proof. (1) Let $\mathcal{B}=\{B_1,B_2,\dots\}$ be a countable basis. The $B_n$ containing $x$ form a countable local basis. Choosing a point $d_n\in B_n$ (for $B_n\ne\varnothing$) gives a countable set meeting every nonempty open set, i.e. a dense set. Given an open cover, for each $n$ such that $B_n$ lies in some member of the cover choose one such member $U_n$; these countably many $U_n$ cover $X$, because each $x$ lies in some cover member $U$ and then in some $B_n\subseteq U$.
(2) $\{B(x,\frac1n):n\in\mathbb{N}\}$ is a local basis. If $D$ is countable and dense, the balls $B(d,\frac1n)$, $d\in D$, $n\in\mathbb{N}$, form a countable basis: given $x\in U$ with $B(x,r)\subseteq U$, pick $n$ with $\frac2n\lt r$ and $d\in D\cap B(x,\frac1n)$; then $x\in B(d,\frac1n)\subseteq B(x,\frac2n)\subseteq U$. Lindelöf $\Rightarrow$ separable: for each $n$ the cover by all balls $B(x,\frac1n)$ has a countable subcover; the countably many centres, over all $n$, form a dense set. With (1) the three properties are equivalent.
(3) Let $B_1\supseteq B_2\supseteq\cdots$ be a local basis at $x$ (replace $B_n$ by $B_1\cap\dots\cap B_n$). If $x\in\overline{A}$, choose $a_n\in B_n\cap A$; every neighbourhood of $x$ contains some $B_N$ and hence all $a_n$ with $n\ge N$, so $a_n\to x$. The converse holds in every space (§11.1, closure and sequences). For maps: if $f$ preserves sequential limits and $x\in\overline{A}$, take $a_n\to x$ in $A$; then $f(a_n)\to f(x)$, so $f(x)\in\overline{f(A)}$, and the closure form of continuity (§11.1: $f(\overline{A})\subseteq\overline{f(A)}$ for all $A$) gives continuity.
(4) (Sketch.) Traces of a countable (local) basis on a subspace form a countable (local) basis. For a countable product, finite intersections of the countably many sets $\pi_n^{-1}(B)$, $B$ in the countable bases, form a countable basis. A countable dense $D\subseteq X$ meets every nonempty open subset of an open $U$, so $D\cap U$ is dense in $U$. For a closed $C$ in a Lindelöf space, add $X\setminus C$ to a cover of $C$. The counterexamples come from the Sorgenfrey line and plane (Theorem 11.20 and Exercise III.3). ∎
- $\mathbb{R}_\ell$ is first countable, separable and Lindelöf, but not second countable. Consequently $\mathbb{R}_\ell$ is not metrizable.
- The Sorgenfrey plane $\mathbb{R}_\ell\times\mathbb{R}_\ell$ is separable and first countable but not Lindelöf; its anti-diagonal $L=\{(x,-x)\}$ is a closed subspace that is not separable.
Proof. (1) $\{[x,x+\frac1n)\}$ is a countable local basis, and $\mathbb{Q}$ is dense (§11.1, Example 3). Suppose $\mathcal{B}$ is any basis. For each $x$ choose $B_x\in\mathcal{B}$ with $x\in B_x\subseteq[x,x+1)$; then $x=\min B_x$, so $x\ne y$ gives $B_x\ne B_y$, and $\mathcal{B}$ is uncountable. Lindelöf (sketch): given a cover, we may assume it consists of basic sets $[a_\alpha,b_\alpha)$. The union $W$ of the open intervals $(a_\alpha,b_\alpha)$ is standard-open, and since $\mathbb{R}$ is second countable, countably many of these intervals already cover $W$. Each point $x\notin W$ is a left endpoint $a_{\alpha(x)}$ of some cover member, and for $x\lt y$ outside $W$ the intervals $(x,b_{\alpha(x)})$ and $(y,b_{\alpha(y)})$ are disjoint (else $y\in(x,b_{\alpha(x)})\subseteq W$); disjoint open intervals each contain a different rational, so there are only countably many such $x$, and one cover member for each finishes the job. Finally, a separable metrizable space is second countable (Theorem 11.19(2)), so $\mathbb{R}_\ell$ is not metrizable.
(2) $\mathbb{Q}^2$ is dense and the products of local bases are local bases. By §11.1, Example 7(e), $L$ is closed and discrete; an uncountable discrete space is not separable (every point must be in the dense set). The cover consisting of $\mathbb{R}^2\setminus L$ (open) and the sets $[x,x+1)\times[-x,-x+1)$, each meeting $L$ in one point, has no countable subcover. ∎
Separation Axioms
A space $X$ is
- $T_0$ if for any two distinct points, some open set contains one of them but not the other;
- $T_1$ if for any two distinct points $x,y$ there is an open set containing $x$ but not $y$;
- $T_2$, or Hausdorff, if any two distinct points have disjoint open neighbourhoods.
- $T_2\Rightarrow T_1\Rightarrow T_0$. $X$ is $T_1$ if and only if every singleton is closed; a finite $T_1$ space is discrete.
- In a Hausdorff space every sequence has at most one limit.
- Metric spaces and order topologies are Hausdorff. Subspaces and products of $T_i$ spaces are $T_i$ ($i=0,1,2$); a finer topology than a $T_i$ topology is $T_i$.
- $X$ is Hausdorff if and only if the diagonal $\{(x,x)\}$ is closed in $X\times X$. If $f,g:Y\to X$ are continuous and $X$ is Hausdorff, $\{f=g\}$ is closed.
Proof. (1) The implications are immediate from the definitions. If $X$ is $T_1$ and $x\in X$, each $y\ne x$ has an open neighbourhood missing $x$; their union $X\setminus\{x\}$ is open. Conversely, if singletons are closed, $X\setminus\{y\}$ is an open set containing $x$ but not $y$. In a finite $T_1$ space every set is a finite union of closed singletons, hence closed; so every set is also open.
(2) If $x_n\to x$ and $x_n\to y\ne x$, take disjoint $U\ni x$, $V\ni y$: eventually $x_n\in U\cap V=\varnothing$. (3) Metric spaces: balls of radius $\frac12d(x,y)$ (§11.1, metric topologies). In an order topology with $x\lt y$: if some $z$ lies between them, $(-\infty,z)$ and $(z,+\infty)$ work; otherwise $(-\infty,y)$ and $(x,+\infty)$ are disjoint. For subspaces intersect the separating sets with the subspace; for products, distinct points differ in some coordinate $\alpha$, and the preimages under $\pi_\alpha$ of separating sets in $X_\alpha$ work. A finer topology has more open sets available.
(4) $(x,y)\notin\Delta$ has a box $U\times V$ missing $\Delta$ exactly when $U\cap V=\varnothing$. Then $\{f=g\}$ is the preimage of $\Delta$ under the continuous map $y\mapsto(f(y),g(y))$. ∎
Let $X$ be a $T_1$ space. $X$ is regular ($T_3$) if for every closed $C$ and every $x\notin C$ there are disjoint open sets $U\ni x$ and $V\supseteq C$; it is normal ($T_4$) if any two disjoint closed sets have disjoint open neighbourhoods.
(Some books do not include $T_1$ in these definitions; then a two-point indiscrete space is "normal". In a GATE problem, read the definition the question uses. With $T_1$ included, as here and in the CSIR course, the ladder below holds.)
- $T_4\Rightarrow T_3\Rightarrow T_2\Rightarrow T_1\Rightarrow T_0$.
- $X$ ($T_1$) is regular iff for every open $U\ni x$ there is an open $V$ with $x\in V\subseteq\overline{V}\subseteq U$; normal iff for every closed $A$ inside an open $U$ there is an open $V$ with $A\subseteq V\subseteq\overline{V}\subseteq U$.
- Every metric space is normal. Every compact Hausdorff space is normal. Every regular Lindelöf space (in particular every regular second countable space) is normal.
- Subspaces and products of regular spaces are regular. Closed subspaces of normal spaces are normal, but arbitrary subspaces and products (even of two copies of $\mathbb{R}_\ell$) need not be.
Proof. (1) In a $T_1$ space points are closed, so $T_4\Rightarrow T_3$ (take $A=\{x\}$) and $T_3\Rightarrow T_2$ (separate $x$ from the closed set $\{y\}$). (2) For regularity apply the definition to $x$ and $C=X\setminus U$: disjoint $V\ni x$ and $W\supseteq C$ give $\overline{V}\subseteq X\setminus W\subseteq U$. Conversely, given $x\notin C$, take $U=X\setminus C$ and use $V$ and $X\setminus\overline{V}$. Normality is the same with $A$ in place of $x$.
(3) In a metric space let $A,B$ be disjoint closed sets and put $U=\{x:d(x,A)\lt d(x,B)\}$, $V=\{x:d(x,A)\gt d(x,B)\}$. These are open (the distance to a set is $1$-Lipschitz) and disjoint; $A\subseteq U$ because for $a\in A$, $d(a,A)=0\lt d(a,B)$ (positive, as $B$ is closed and $a\notin B$); similarly $B\subseteq V$. For compact Hausdorff $X$: closed sets are compact, and Theorem 11.18(2) separates a compact set from a point; repeating that argument with a compact set $A$ in place of the point $x$ (cover $B$ by finitely many neighbourhoods disjoint from neighbourhoods of $A$) separates $A$ from $B$. The Lindelöf statement is only quoted here.
(4) (Sketch.) Regularity passes to subspaces by intersecting, and to products using (2) with basic boxes and $\overline{\prod V_\alpha}=\prod\overline{V_\alpha}$. A closed subset of a closed subspace is closed in $X$, so normality passes to closed subspaces. For the failure in products see Theorem 11.23(4). ∎
- The Sierpiński space (and the right-ray topology on $\mathbb{R}$) is $T_0$ but not $T_1$.
- An infinite set with $\mathcal{T}_{\mathrm{cof}}$, and $\mathbb{R}$ with $\mathcal{T}_{\mathrm{coc}}$, are $T_1$ but not Hausdorff.
- $\mathbb{R}_K$ is Hausdorff but not regular.
- $\mathbb{R}_\ell$ is normal, but the Sorgenfrey plane $\mathbb{R}_\ell\times\mathbb{R}_\ell$ is regular and not normal.
Proof. (1) In the Sierpiński space $\{1\}$ is open, so $T_0$ holds, but every open set containing $0$ contains $1$. In the right-ray topology, for $x\lt y$ the ray $(x,\infty)$ contains $y$ and not $x$, but every open set containing $x$ contains $y$. (2) Singletons are finite (countable), hence closed: $T_1$. Two nonempty open sets always meet (§11.1, metrizability), so the space is not Hausdorff.
(3) $\mathbb{R}_K$ is finer than $\mathbb{R}$, hence Hausdorff, and $K$ is closed in it (§11.1, the $K$-topology) with $0\notin K$. Suppose $U\ni0$ and $V\supseteq K$ are open and disjoint. $U$ contains a basic set around $0$, which must be of the form $(a,b)\setminus K$ (an interval around $0$ meets $K\subseteq V$). Choose $n$ with $\frac1n\lt b$. Since $\frac1n\in V$, some basic set around $\frac1n$ lies inside $V$; it must be an interval $(c,d)$ (the other kind omits the points of $K$), and we may take $0\lt c\lt\frac1n$. Any $z\in(c,\frac1n)$ with $z\notin K$ lies in $(a,b)\setminus K\subseteq U$ and in $(c,d)\subseteq V$ — a contradiction.
(4) $\mathbb{R}_\ell$ is $T_1$ (finer than $\mathbb{R}$). Let $A,B$ be disjoint closed sets. For $a\in A$ choose $[a,x_a)$ missing $B$, and for $b\in B$ choose $[b,x_b)$ missing $A$. Put $U=\bigcup_a[a,x_a)$ and $V=\bigcup_b[b,x_b)$. If $[a,x_a)\cap[b,x_b)\ne\varnothing$ with, say, $a\lt b$, then $b\in[a,x_a)$, which misses $B$ — impossible; $b\lt a$ is symmetric. So $U,V$ are disjoint and $\mathbb{R}_\ell$ is normal. Hence it is regular, and so is the product (Theorem 11.22(4)).
The Sorgenfrey plane is not normal (proof via a counting argument known as Jones's lemma). Let $D=\mathbb{Q}^2$, which is dense, and $L$ the anti-diagonal, closed and discrete. Every subset $A\subseteq L$ is closed, and so is $L\setminus A$. If the plane were normal, choose disjoint open $U_A\supseteq A$ and $V_A\supseteq L\setminus A$ for each $A$. The map $A\mapsto U_A\cap D$ is injective: if $A\ne B$, say $p\in A\setminus B$, then $p\in U_A\cap V_B$, a nonempty open set, which contains some $d\in D$; then $d\in U_A$ but $d\notin U_B$ (because $U_B\cap V_B=\varnothing$). So the power set of $L$, which has the cardinality $2^{\mathfrak{c}}$, injects into the power set of the countable set $D$, which has cardinality $\mathfrak{c}$ — contradicting Cantor's theorem. ∎
Urysohn's Lemma and Its Uses
Proof. Metric case. The denominator never vanishes: $d(x,A)=0=d(x,B)$ would put $x$ in $\overline{A}\cap\overline{B}=A\cap B=\varnothing$. Both distances are continuous, so $f$ is continuous, with values in $[0,1]$, $f=0$ on $A$ and $f=1$ on $B$.
General case (careful sketch). Let $P$ be the set of dyadic rationals $\frac k{2^m}$ in $[0,1]$. Put $U_1=X\setminus B$, an open set containing $A$, and by Theorem 11.22(2) choose an open $U_0$ with $A\subseteq U_0\subseteq\overline{U_0}\subseteq U_1$. Inductively, if $U_p$ is defined for all $p$ with denominator $2^m$ so that $p\lt q$ implies $\overline{U_p}\subseteq U_q$, then for a new point $r$ between consecutive old points $p\lt q$ apply Theorem 11.22(2) to $\overline{U_p}\subseteq U_q$ to get $\overline{U_p}\subseteq U_r\subseteq\overline{U_r}\subseteq U_q$. This gives a nested family of open sets indexed by $P$; put $U_p=X$ for $p\gt1$. Define $f(x)=\inf\{p:x\in U_p\}$. Then $f=0$ on $A\subseteq U_0$, and $f=1$ on $B$ since $B$ misses every $U_p$ with $p\le1$. Continuity follows from two observations: $x\in\overline{U_p}\Rightarrow f(x)\le p$, and $x\notin U_p\Rightarrow f(x)\ge p$. Given $f(x)\in(c,d)$, choose dyadic $p,q$ with $c\lt p\lt f(x)\lt q\lt d$; the open set $U_q\setminus\overline{U_p}$ contains $x$ and is mapped into $[p,q]\subseteq(c,d)$.
Converse. $f^{-1}([0,\frac12))$ and $f^{-1}((\frac12,1])$ are disjoint open sets containing $A$ and $B$. ∎
- Tietze extension theorem. If $X$ is normal, $C\subseteq X$ is closed and $g:C\to[a,b]$ (or $g:C\to\mathbb{R}$) is continuous, then $g$ extends to a continuous function on all of $X$ with values in $[a,b]$ (respectively $\mathbb{R}$).
- Urysohn metrization theorem. Every regular second countable space is metrizable. (It embeds in $[0,1]^\omega$ with the product metric of §11.1.)
Hence: a compact Hausdorff space is metrizable if and only if it is second countable; a second countable space is metrizable if and only if it is regular.
The last sentence combines the theorem with Theorems 11.19(2) and 11.22(3): a metric space is regular, and a compact metric space is separable, hence second countable. Typical uses in GATE problems: $\mathbb{R}_\ell$ is normal but not metrizable (separable without a countable basis); the Sorgenfrey plane is not metrizable (not even normal); $[0,1]^\omega$ with the product topology is a compact metrizable space; and a function continuous on a closed subset of $\mathbb{R}^n$ extends continuously to $\mathbb{R}^n$ (Tietze; $\mathbb{R}^n$ is normal).
- Connected? Look for a clopen proper subset (a separation), or a continuous onto map to a disconnected set such as $\{0,1\}$ or $\mathbb{R}\setminus\{0\}$ ($\det$ on matrices, the sign of a coordinate). To prove connectedness, write the set as a union of connected pieces meeting each other, or as a continuous image of something connected. If any two nonempty open sets meet, the space is connected.
- Compact? In $\mathbb{R}^n$: closed and bounded. Elsewhere: produce an open cover without a finite subcover (nested increasing sets, or an infinite closed discrete subset), or recognise a closed subset of a compact space, a continuous image of a compact space, or a cofinite topology.
- Countability. Metric: first countable, and second countable $\iff$ separable. Not separable: an uncountable family of disjoint nonempty open sets. Not second countable: an uncountable family of points each needing its own basic set ($\mathbb{R}_\ell$).
- Separation. Check $T_1$ by asking whether points are closed; Hausdorff fails as soon as two nonempty open sets must meet; metric, compact Hausdorff and $\mathbb{R}_\ell$ are normal; $\mathbb{R}_K$ and the Sorgenfrey plane are the standard failures of $T_3$ and $T_4$.
- Homeomorphic? Compare invariants: connectedness, number of components, cut points, compactness, countability, separation axioms, and the cardinality of the space or of its topology.
- Connectedness depends on the topology: $\mathbb{R}$ is connected, $\mathbb{R}_\ell$ is totally disconnected, the cofinite and cocountable lines are connected. $\mathbb{R}^\omega$ is connected in the product topology but not in the box topology.
- Totally disconnected $\ne$ discrete: $\mathbb{Q}$, $\{0\}\cup\{\frac1n\}$, $\mathbb{R}_\ell$. With infinitely many components, components need not be open.
- Connected $\not\Rightarrow$ path-connected (topologist's sine curve, §4.1); the intersection of two connected sets need not be connected.
- Closed and bounded is a theorem about $\mathbb{R}^n$: $[0,1]$ is not compact in $\mathbb{R}_\ell$; $K\cup\{0\}$ is not compact in $\mathbb{R}_K$; the unit ball of $\ell^2$ is not compact (§4.1).
- Compact sets need not be closed without the Hausdorff property: every subset of a cofinite space is compact. A finer topology can destroy compactness ($[0,1]$ in $\mathbb{R}$ versus $\mathbb{R}_\ell$), and a coarser one can destroy the Hausdorff property.
- A continuous bijection is a homeomorphism only from compact to Hausdorff: $[0,2\pi)\to S^1$ fails because $[0,2\pi)$ is not compact; $\mathrm{id}:[0,1]\to([0,1],\text{indiscrete})$ fails because the target is not Hausdorff.
- Separable is not hereditary: the anti-diagonal of the Sorgenfrey plane. Separable $\not\Rightarrow$ second countable ($\mathbb{R}_\ell$, the cofinite line) — but for metric spaces it does.
- Products can lose properties: $\mathbb{R}_\ell$ is Lindelöf and normal; $\mathbb{R}_\ell\times\mathbb{R}_\ell$ is neither.
- Unique limits $\not\Rightarrow$ Hausdorff: the cocountable line. $T_1\not\Rightarrow T_2$: the cofinite line, the line with two origins.
- Hausdorff $\not\Rightarrow$ regular: $\mathbb{R}_K$; regular $\not\Rightarrow$ normal: the Sorgenfrey plane; normal $\not\Rightarrow$ metrizable: $\mathbb{R}_\ell$.
- Conventions: without "$T_1$" in the definition, an indiscrete space would count as regular and normal; always check which definition a problem uses.
- Urysohn's lemma needs normality (disjoint closed sets), and the Tietze theorem needs a closed domain: $\sin\frac1x$ on $(0,1]$ has no continuous extension to $[0,1]$.
- CSIR Unit 2 · §9.3 Separation Axioms — characterisations of $T_1$, the diagonal criterion, subspaces and products of $T_0$–$T_2$ spaces, the line with two origins, metric spaces as normal spaces, the $K$-topology in detail, Jones's lemma, regular Lindelöf spaces, Urysohn's lemma with the full dyadic proof and its converse, completely regular spaces, Tietze and Urysohn metrization.
- CSIR Unit 2 · §9.4 Connectedness and Compactness — three forms of connectedness, the connected subsets of $\mathbb{R}$, products and closures of connected sets, components, path components and cut points, the topologist's sine curve, the tube lemma, compact Hausdorff spaces, Tychonoff's theorem, local compactness and the one-point compactification.
- Connectedness, compactness and the countability properties of metric spaces are in §4.1; the basic definitions of first and second countability are also discussed in CSIR Unit 2 · §9.1.
Two Labs
Exercises
- 1 Which are connected? (a) $\mathbb{R}\setminus\mathbb{Q}$; (b) $\{(x,y):xy\ge0\}$; (c) $\{(x,y):xy\gt0\}$; (d) the points of $\mathbb{R}^2$ with at least one rational coordinate; (e) the graph of $\sin\frac1x$, $0\lt x\le1$.
- 2 Show that $\mathbb{R}$ with $\mathcal{T}_{\mathrm{coc}}$ is connected, and that so is any space in which every two nonempty open sets meet.
- 3 Prove that $\mathbb{R}_\ell$ is totally disconnected, and find all continuous maps from $\mathbb{R}$ (standard) to $\mathbb{R}_\ell$.
- 4 Prove that $\mathbb{R}$ and $\mathbb{R}^2$ are not homeomorphic, and that $[0,1]$ and $[0,1]^2$ are not.
- 5 How many components has $\{(x,y):(x^2+y^2-1)(x^2+y^2-4)\,y\ne0\}$?
- 6 Let $f:S^1\to\mathbb{R}$ be continuous. Prove that $f(z)=f(-z)$ for some $z\in S^1$.
- 7 Sort the letters K, L, M, N, Y (drawn with straight strokes) into homeomorphism classes.
- 8 In $\mathbb{R}^\omega$ with the box topology, show that the bounded sequences form a clopen set, and deduce that the component of $0$ is not all of $\mathbb{R}^\omega$.
- 1 Which are compact? (a) $\{(x,y):x^4+y^4\le1\}$; (b) $[0,1]\cap\mathbb{Q}$; (c) the unit sphere of $\mathbb{R}^3$; (d) $\{(x,y):x\ge0,\ 0\le y\le e^{-x}\}$.
- 2 Prove that every subset of a cofinite space is compact.
- 3 Prove that a subset of $\mathbb{R}$ with $\mathcal{T}_{\mathrm{coc}}$ is compact if and only if it is finite.
- 4 Give an open cover of $[0,1]$ in $\mathbb{R}_\ell$, different from the one in Example 3, with no finite subcover. Is $\{0\}\cup\{1+\frac1n\}$ compact in $\mathbb{R}_\ell$?
- 5 Prove that a continuous map from a compact space to a Hausdorff space is closed. Show by an example that "Hausdorff" cannot be dropped.
- 6 Let $f:[0,1]\to\mathbb{R}^2$ be continuous and injective. Show that $f$ is an embedding. Is every continuous injective $g:[0,1)\to\mathbb{R}^2$ an embedding?
- 7 Show that $O(n)$ and $SO(n)$ are compact and $GL_n(\mathbb{R})$ is not.
- 8 Let $X$ be compact and $f:X\to X$ continuous with $f(x)\ne x$ for all $x$, where $X$ is a metric space. Show that $\inf_xd(x,f(x))\gt0$.
- 1 Show that $\mathbb{R}$ with $\mathcal{T}_{\mathrm{cof}}$ is separable, compact and Lindelöf but not first countable.
- 2 Show that $\mathbb{R}$ with the discrete topology is first countable but not separable, and that $\ell^\infty$ is not separable.
- 3 Show that the Sorgenfrey plane is separable, but its closed subspace $L=\{(x,-x)\}$ is not.
- 4 Show directly that $\mathbb{R}^n$ is second countable and that every open cover of an open set $U\subseteq\mathbb{R}^n$ has a countable subcover.
- 5 In a first countable space, prove that $x\in\overline{A}$ iff some sequence in $A$ converges to $x$. Show by an example that the "if" direction alone holds in general.
- 6 Is $C[0,1]$ with the sup metric separable? Is it second countable?
- 7 Prove that every second countable space has at most $\mathfrak{c}$ open sets, and deduce that $\mathbb{R}$ with the discrete topology is not second countable.
- 1 Show that the topology $\{\varnothing,\{a\},\{a,b\},X\}$ on $\{a,b,c\}$ is $T_0$ but not $T_1$, and find all $T_1$ topologies on a finite set.
- 2 Which separation axioms does the right-ray topology on $\mathbb{R}$ satisfy? Is it compact? connected? separable?
- 3 Show that a subspace of a Hausdorff space is Hausdorff, and that a quotient of a Hausdorff space need not be (§11.1).
- 4 Show that in a Hausdorff space two disjoint compact sets have disjoint open neighbourhoods.
- 5 In $\mathbb{R}$ let $A=(-\infty,0]$ and $B=[2,\infty)$. Write the metric Urysohn function, compute $f(0.5)$, and identify the open sets $\{f\lt\frac12\}$ and $\{f\gt\frac12\}$.
- 6 In $\mathbb{R}^2$ let $A=\{y\le0\}$ and $B=\{(x,y):x\gt0,\ xy\ge1\}$. Show that $A$ and $B$ are disjoint closed sets at distance $0$, and estimate the Urysohn function $f$ of Theorem 11.24 at $(1,\frac12)$.
- 7 Prove that a closed subspace of a normal space is normal.
- 8 Define $g$ on $C=\{0\}\cup\{\frac1n:n\in\mathbb{N}\}$ by $g(0)=0$, $g(\frac1n)=\frac{(-1)^n}{n}$. Show that $g$ extends to a continuous function on $\mathbb{R}$ and describe one extension. What goes wrong for $h(0)=0$, $h(\frac1n)=(-1)^n$?
E1. GATE MA · MCQ For which one of the following topologies on $\mathbb{R}$ is $\mathbb{R}$ connected? (A) the topology generated by the intervals $[a,b)$; (B) the discrete topology; (C) the cofinite topology; (D) the topology generated by all intervals $[a,b)$ and $(a,b]$.
E2. GATE MA · MCQ Let $\mathcal{T}=\{\varnothing,\mathbb{R}\}\cup\{(a,\infty):a\in\mathbb{R}\}$. Then $(\mathbb{R},\mathcal{T})$ is (A) $T_1$ but not Hausdorff; (B) $T_0$ but not $T_1$; (C) Hausdorff; (D) not $T_0$.
E3. GATE MA · MSQ Which of the following are compact? (One or more options may be correct.) (A) $[0,1]$ as a subspace of $\mathbb{R}_\ell$; (B) $[0,1]$ as a subspace of $\mathbb{R}$ with the cofinite topology; (C) $\mathbb{R}$ with the cocountable topology; (D) $\{0\}\cup\{\frac1n:n\in\mathbb{N}\}$ as a subspace of $\mathbb{R}_\ell$.
E4. GATE MA · MSQ Which of the following statements are true? (One or more options may be correct.) (A) Every second countable space is separable. (B) Every separable metric space is second countable. (C) Every separable topological space is second countable. (D) Every subspace of a separable space is separable.
E5. GATE MA · NAT Let $A=\{(x,y):x^2+y^2\le1\}$ and $B=\{(x,y):x\ge4\}$ in $\mathbb{R}^2$ with the Euclidean metric, and $f(p)=\dfrac{d(p,A)}{d(p,A)+d(p,B)}$. The value $f(2,2)$, correct to two decimal places, is ______.
E6. GATE MA · NAT Let $X=[0,1]\cup[2,3]\cup\{5\}\cup(6,7)$ with the subspace topology from $\mathbb{R}$. The number of subsets of $X$ that are both open and closed in $X$ is ______.
Suggested plan — 3 lectures × 60 min.
- Connectedness and compactness. Definition 11.14, Theorem 11.15, Theorem 11.16 (prove (1) and (2) in full: the intervals and the intermediate value theorem), components (Theorem 11.17) and Examples 1–2 — the letters are a good class activity. Then Definition 11.16 and Theorem 11.18 (prove (1)–(4)); Example 3 with Interactive 2 projected, case by case; Example 4.
- Countability. Definition 11.17 and Theorem 11.19 (prove (1) and (2)); Theorem 11.20 with the proof that $\mathbb{R}_\ell$ has no countable basis; Example 5 as a table on the board. GATE corner E4.
- Separation and Urysohn. Definitions 11.18–11.19, Theorems 11.21–11.23 (prove the metric case of normality and the normality of $\mathbb{R}_\ell$; sketch $\mathbb{R}_K$ and Jones's lemma); Example 6. Urysohn's lemma: the metric formula first, then the dyadic construction with Interactive 1 projected (play the hyperbola case); Example 7; statements of Tietze and Urysohn metrization. E1, E2, E3, E5, E6 under a timer.
Discussion questions. (1) Which properties in this section are preserved by continuous surjections (connectedness, compactness, separability), which by homeomorphisms only, and which pass to subspaces? Make a table. (2) Why does a finer topology make compactness harder and the Hausdorff property easier? What does Example 4 say about the standard topology of $[0,1]$? (3) In Interactive 1 the two sets of the hyperbola case are at distance $0$. Why does the metric Urysohn function still work, and where does the proof of normality of metric spaces use that $A$ and $B$ are closed?
Misconceptions to watch for. "Closed and bounded implies compact" outside $\mathbb{R}^n$; "compact sets are closed" without Hausdorff; "unique limits means Hausdorff"; confusing separable with second countable outside metric spaces; forgetting $T_1$ in the definitions of $T_3$, $T_4$; thinking a product of normal spaces is normal; treating totally disconnected as discrete; using Tietze on a non-closed set.
Suggested homework. I(1, 3, 5, 7), II(1, 3, 4, 6), III(1, 3, 5, 7), IV(1, 2, 5, 6, 8); GATE corner E3, E4, E6.
Research question. A data set is a finite cloud of points. How many "pieces" does it have — and how can the answer be made independent of an arbitrary choice of scale?
Why it matters. Topological data analysis is used to find clusters of patients with similar gene expression, groups of galaxies, communities in social networks and coverage gaps in wireless sensor networks. Its simplest tool counts connected components at every scale at once: replace each data point by a closed disc of radius $\frac\varepsilon2$ and count the components of the union as $\varepsilon$ grows. The record of the scales at which components merge (a "barcode") is stable under small errors in the data, which is what makes it useful.
What to do.
- Show that two discs of radius $\frac\varepsilon2$ meet iff their centres are at distance at most $\varepsilon$, and deduce that the components of the union correspond to the connected components of the graph joining points at distance $\le\varepsilon$.
- Generate (with a fixed random seed) three clusters of $30$ points each around $(0,0)$, $(4,0)$ and $(2,3)$ with spread $0.5$. Compute the number of components $N(\varepsilon)$ for $\varepsilon$ from $0$ to $5$ with a union–find (disjoint-set) structure, and plot it.
- Find the merge scales: they are the edge lengths of a minimum spanning tree of the points (explain why, using Kruskal's algorithm). Identify the long plateau where $N(\varepsilon)=3$.
- Move each point by a random amount at most $\delta=0.05$ and recompute the merge scales. Show that each changes by at most $2\delta$.
Tools: Python with numpy and scipy (scipy.spatial.distance and scipy.sparse.csgraph.minimum_spanning_tree), or a spreadsheet for a small cloud.
Hand in: the argument for step 1; the plot of $N(\varepsilon)$; the list of merge scales before and after the perturbation; one paragraph on why the length of the plateau measures how convincing the three clusters are.
Going further. Read about persistent homology, which tracks not only components but loops and voids as the scale grows, and about its use in mapping the coverage of sensor networks.
🧪 Self-check quiz
Q1. Which of these spaces is connected?
Q2. A compact subset of a Hausdorff space is always
Q3. The Sorgenfrey line $\mathbb{R}_\ell$ is
Q4. Which space is $T_1$ but not Hausdorff?
Q5. Urysohn's lemma produces a continuous $f:X\to[0,1]$ with $f=0$ on $A$, $f=1$ on $B$, for every pair of disjoint closed sets $A,B$, provided $X$ is