Skip to content

Connectedness

Definition. A topological space XX is disconnected if there exist nonempty disjoint open sets U,VU, V with X=U∪VX = U \cup V. Such a pair {U,V}\{U, V\} is called a separation of XX.

XX is connected if it is not disconnected.

Equivalently, XX is connected if and only if the only clopen subsets of XX are ∅\emptyset and XX.

Proof of the equivalence. If X=U∪VX = U \cup V is a separation, then U=X∖VU = X \setminus V is both open and closed, and neither ∅\emptyset nor XX. Conversely, if C⊆XC \subseteq X is clopen with C∉{∅,X}C \notin \{\emptyset, X\}, then U=CU = C and V=X∖CV = X \setminus C are nonempty disjoint open sets with U∪V=XU \cup V = X; a separation. ■\blacksquare

Equivalently again, XX is connected if and only if every continuous map h:X→{0,1}h : X \to \{0, 1\} into the two-point space with the discrete topology is constant: preimages of {0}\{0\} and {1}\{1\} are clopen in XX, so a non-constant hh is a nontrivial clopen set, and conversely. This formulation - connectedness as the statement that there are no non-constant maps into the discrete two-point space; is the one that makes the proofs below short.

Example 6.1. [0,1][0, 1] is connected. [0,1)∪(2,3][0, 1) \cup (2, 3] is disconnected (with the subspace topology from R\mathbb{R}).

Example 6.2. Q\mathbb{Q} with the subspace topology from R\mathbb{R} is totally disconnected: the only connected subsets are singletons.

Proof. Let A⊆QA \subseteq \mathbb{Q} be connected. As a subset of R\mathbb{R} (same subspace topology), AA is an interval by Theorem 6.1. If AA contained two points a<ba < b, it would contain every cc with a<c<ba < c < b; including irrational cc, contradicting A⊆QA \subseteq \mathbb{Q}. So AA is a singleton. ■\blacksquare

Theorem 6.1. A subset of R\mathbb{R} (with the standard topology) is connected if and only if it is an interval.

(Here an interval is any set I⊆RI \subseteq \mathbb{R} with the property: if a,b∈Ia, b \in I and a<c<ba < c < b, then c∈Ic \in I.)

Proof. (⇐\Leftarrow) Intervals are connected. Let II be an interval and suppose I=U∪VI = U \cup V with U,VU, V nonempty, disjoint, and open in II. Pick a∈Ua \in U and b∈Vb \in V; swap the names if necessary so that a<ba < b. Since [a,b]⊆I[a, b] \subseteq I (interval property), define

c=sup⁡{x∈[a,b]:[a,x]⊆U}.c = \sup \{ x \in [a, b] : [a, x] \subseteq U \}.

The set is nonempty: UU is open in II and contains aa, so [a,a+δ)⊆U[a, a + \delta) \subseteq U for some δ>0\delta > 0. Hence a<c≤ba < c \leq b, and [a,c)⊆U[a, c) \subseteq U by definition of the supremum. Since c∈[a,b]⊆Ic \in [a, b] \subseteq I, the point cc lies in UU or in VV.

If c∈Uc \in U: openness of UU in II gives ε>0\varepsilon > 0 with (c−ε,c+ε)∩I⊆U(c - \varepsilon, c + \varepsilon) \cap I \subseteq U. Then [a,c+ε/2]⊆U[a, c + \varepsilon/2] \subseteq U (the segment [a,c)[a, c) from the supremum, the segment (c−ε,c+ε/2](c - \varepsilon, c + \varepsilon/2] from openness, both inside II), contradicting cc as an upper bound for the set.

So c∈Vc \in V. Openness of VV in II gives ε>0\varepsilon > 0 with (c−ε,c+ε)∩I⊆V(c - \varepsilon, c + \varepsilon) \cap I \subseteq V. Choose xx with max⁡(a,c−ε)<x<c\max(a, c - \varepsilon) < x < c; possible since c>ac > a. Then x∈[a,c)⊆Ux \in [a, c) \subseteq U and x∈(c−ε,c)⊆Vx \in (c - \varepsilon, c) \subseteq V, contradicting U∩V=∅U \cap V = \emptyset.

(⇒\Rightarrow) Connected sets are intervals. Contrapositive: if II is not an interval, there are a<c<ba < c < b with a,b∈Ia, b \in I and c∉Ic \notin I. Then

U=I∩(−∞,c),V=I∩(c,∞)U = I \cap (-\infty, c), \qquad V = I \cap (c, \infty)

are open in II, nonempty (a∈Ua \in U, b∈Vb \in V), disjoint, and cover II (as c∉Ic \notin I). So II is disconnected. ■\blacksquare

This theorem is why the intermediate value theorem exists and why Q\mathbb{Q} is dust: connected subsets of Q\mathbb{Q}, being connected subsets of R\mathbb{R}, are intervals; but an interval containing two distinct rationals contains the irrationals between them, which Q\mathbb{Q} does not. So the only connected subsets of Q\mathbb{Q} are singletons (Example 6.2).

Definition. A space XX is path-connected if for any two points x,y∈Xx, y \in X, there exists a continuous function γ:[0,1]→X\gamma : [0, 1] \to X with γ(0)=x\gamma(0) = x and γ(1)=y\gamma(1) = y.

Proposition 6.1. Every path-connected space is connected. The converse is false.

Proof. Let XX be path-connected and fix a point x0∈Xx_0 \in X. For each x∈Xx \in X choose a path γx:[0,1]→X\gamma_x : [0, 1] \to X from x0x_0 to xx. Its image is connected (Proposition 6.4 below, applied to the connected interval [0,1][0, 1]), and all the images contain x0x_0. So

X=⋃x∈Xγx([0,1])X = \bigcup_{x \in X} \gamma_x([0, 1])

is a union of connected sets with a common point; connected by the pencil lemma (Proposition 6.1.1 below). ■\blacksquare

Proposition 6.1.1 (Pencil lemma). If {Cα}\{C_\alpha\} is a family of connected subsets of XX with ⋂αCα≠∅\bigcap_\alpha C_\alpha \neq \emptyset, then ⋃αCα\bigcup_\alpha C_\alpha is connected.

Proof. Let C=⋃αCαC = \bigcup_\alpha C_\alpha and pick p∈⋂αCαp \in \bigcap_\alpha C_\alpha. Let h:C→{0,1}h : C \to \{0, 1\} be continuous into the discrete two-point space. For each α\alpha, the sets h−1(0)∩Cαh^{-1}(0) \cap C_\alpha and h−1(1)∩Cαh^{-1}(1) \cap C_\alpha are disjoint clopen subsets of CαC_\alpha covering it; connectedness of CαC_\alpha forces one of them empty, so hh is constant on each CαC_\alpha. The constants all agree; each equals h(p)h(p). So hh is constant on CC, and CC is connected by the two-point characterisation. ■\blacksquare

Example 6.3 (Topologist’s sine curve). Let

S={(x,sin⁡(1/x)):0<x≤1}∪{(0,y):−1≤y≤1}⊆R2.S = \{(x, \sin(1/x)) : 0 < x \leq 1\} \cup \{(0, y) : -1 \leq y \leq 1\} \subseteq \mathbb{R}^2.

SS is connected but not path-connected.

Proof that SS is connected. Write AA for the graph part and VV for the vertical segment. The set AA is the image of the interval (0,1](0, 1]; connected by Theorem 6.1; under the continuous map x↦(x,sin⁡(1/x))x \mapsto (x, \sin(1/x)), so AA is connected (Proposition 6.4). We claim A‾=S\overline{A} = S, after which Proposition 6.7 (closures of connected sets are connected) finishes the proof.

A‾⊆S\overline{A} \subseteq S: if (xn,sin⁡(1/xn))→(x0,y0)(x_n, \sin(1/x_n)) \to (x_0, y_0) with xn>0x_n > 0, then either x0>0x_0 > 0; in which case continuity of sin⁡(1/x)\sin(1/x) at x0x_0 forces y0=sin⁡(1/x0)y_0 = \sin(1/x_0), so the limit is in AA; or x0=0x_0 = 0, in which case ∣sin⁡(1/xn)∣≤1|\sin(1/x_n)| \leq 1 forces y0∈[−1,1]y_0 \in [-1, 1], so the limit is in VV.

S⊆A‾S \subseteq \overline{A}: take (0,y)∈V(0, y) \in V and any ball BB around it of radius ε\varepsilon and xx-reach δ\delta. Since 1/t1/t runs over a full period of sin⁡\sin as tt runs over any interval (0,δ)(0, \delta); choose tt with 1/t=arcsin⁡(y)+2πk1/t = \arcsin(y) + 2\pi k for a sufficiently large integer kk; there is x∈(0,δ)x \in (0, \delta) with sin⁡(1/x)\sin(1/x) within ε\varepsilon of yy. So B∩A≠∅B \cap A \neq \emptyset, and (0,y)∈A‾(0, y) \in \overline{A}.

Proof that SS is not path-connected. Suppose γ:[0,1]→S\gamma : [0, 1] \to S is a continuous path with γ(0)∈V\gamma(0) \in V and γ(1)∈A\gamma(1) \in A; write πx,πy\pi_x, \pi_y for the coordinate projections, which are continuous. The set

C=γ−1(V)=(πx∘γ)−1({0})C = \gamma^{-1}(V) = (\pi_x \circ \gamma)^{-1}(\{0\})

is closed (as VV is closed and γ\gamma continuous) and nonempty (0∈C0 \in C). Let b=sup⁡C∈Cb = \sup C \in C; then γ(b)∈V\gamma(b) \in V, say γ(b)=(0,yb)\gamma(b) = (0, y_b), while γ(t)∈A\gamma(t) \in A for all t>bt > b.

Take δ=1/3\delta = 1/3 and consider the rectangle U=(−η,η)×(yb−δ,yb+δ)⊆R2U = (-\eta, \eta) \times (y_b - \delta, y_b + \delta) \subseteq \mathbb{R}^2, a neighbourhood of (0,yb)(0, y_b). By continuity of γ\gamma there is η′>0\eta' > 0 with γ((b,b+η′))⊆U∩S\gamma((b, b + \eta')) \subseteq U \cap S. Now πx∘γ\pi_x \circ \gamma is continuous, and the image of the connected interval (b,b+η′)(b, b + \eta') is a connected subset of R\mathbb{R}; an interval, by Theorem 6.1; containing positive values (points of (b,b+η′)(b, b+\eta') map into AA, whose xx-coordinates are positive) and with infimum 00 (as t→b+t \to b^+, πx(γ(t))→πx(γ(b))=0\pi_x(\gamma(t)) \to \pi_x(\gamma(b)) = 0). Hence the image contains all of (0,x1](0, x_1] for some x1>0x_1 > 0. So for every x∗∈(0,x1]x^* \in (0, x_1] there is t∗∈(b,b+η′)t^* \in (b, b + \eta') with (πx∘γ)(t∗)=x∗(\pi_x \circ \gamma)(t^*) = x^*.

Choose x∗∈(0,x1)x^* \in (0, x_1) with sin⁡(1/x∗)∉(yb−δ,yb+δ)\sin(1/x^*) \notin (y_b - \delta, y_b + \delta); possible because 1/x1/x runs over a full period of sin⁡\sin on (1/x1,∞)(1/x_1, \infty), so sin⁡\sin attains values outside any interval of length 2δ=2/3<22\delta = 2/3 < 2. For the corresponding t∗t^*: γ(t∗)=(x∗,sin⁡(1/x∗))\gamma(t^*) = (x^*, \sin(1/x^*)) has πy(γ(t∗))∉(yb−δ,yb+δ)\pi_y(\gamma(t^*)) \notin (y_b - \delta, y_b + \delta), so γ(t∗)∉U\gamma(t^*) \notin U; contradicting γ((b,b+η′))⊆U\gamma((b, b+\eta')) \subseteq U. So no such path exists, and SS is not path-connected. ■\blacksquare

The moral: connectedness passes to closures (Proposition 6.7), but path-connectedness does not - no path can traverse the infinitely many oscillations and still be continuous at the vertical segment.

Example 6.4. Rn\mathbb{R}^n is path-connected for all n≥1n \geq 1: any two points are joined by the straight-line path γ(t)=(1−t)p+tq\gamma(t) = (1-t)p + tq, continuous as each coordinate is polynomial in tt. The same paths work in any convex subset of Rn\mathbb{R}^n.

Definition. A connected component of XX is a maximal connected subset of XX. The connected components of XX form a partition of XX.

Proposition 6.2. Connected components are closed. They need not be open; in a locally connected space they are always open.

Proof. Let CC be a component of XX; a maximal connected subset. By Proposition 6.7, C‾\overline{C} is connected; it contains CC, so maximality of CC forces C‾=C\overline{C} = C. So components are closed; in every topological space, no separation axioms required.

For openness: if XX is locally connected, let x∈Cx \in C and let UU be any open neighbourhood of xx. Local connectedness gives a connected open neighbourhood VV of xx with V⊆UV \subseteq U. Now C∪VC \cup V is connected (pencil lemma: CC and VV share the point xx) and contains CC, so by maximality C∪V⊆CC \cup V \subseteq C, i.e. V⊆CV \subseteq C. Thus every neighbourhood UU of xx contains a neighbourhood VV of xx lying in CC, which makes CC open. ■\blacksquare

Without local connectedness, components can fail to be open: the components of Q\mathbb{Q} are singletons (Example 6.2), and no singleton is open in Q\mathbb{Q}.

Definition. A path component of XX is a maximal path-connected subset. Path components also partition XX, and each path component is contained in a connected component.

Definition. XX is locally connected if for every x∈Xx \in X and every open neighbourhood UU of xx, there exists a connected open neighbourhood VV of xx with V⊆UV \subseteq U.

Proposition 6.3. Every open subset of Rn\mathbb{R}^n is locally connected.

Proof. Let U⊆RnU \subseteq \mathbb{R}^n be open, x∈Ux \in U, and let NN be any neighbourhood of xx in UU. Openness gives ε>0\varepsilon > 0 with B(x,ε)⊆UB(x, \varepsilon) \subseteq U; shrinking along the neighbourhood NN, there is δ>0\delta > 0 with B(x,δ)⊆N∩UB(x, \delta) \subseteq N \cap U. The open ball B(x,δ)B(x, \delta) is convex; the straight-line segment between two of its points stays inside it, by the triangle inequality ∣(1−t)p+tq−x∣≤(1−t)∣p−x∣+t∣q−x∣<δ|(1-t)p + tq - x| \leq (1-t)|p - x| + t|q - x| < \delta; hence path-connected (linear paths), hence connected. So B(x,δ)B(x, \delta) is the required connected open neighbourhood of xx inside NN. ■\blacksquare

Example 6.5. The topologist’s sine curve is connected but not locally connected.

Proof that SS is not locally connected. Let p=(0,y)∈Vp = (0, y) \in V and consider the basic neighbourhood N=B(p,ε)∩SN = B(p, \varepsilon) \cap S with ε<1\varepsilon < 1. The set of xx-coordinates of graph points in the ball,

E={x∈(0,1]:x2+(sin⁡(1/x)−y)2<ε2},E = \{ x \in (0, 1] : x^2 + (\sin(1/x) - y)^2 < \varepsilon^2 \},

meets the condition periodically: as x→0+x \to 0^+, the value 1/x1/x sweeps through infinitely many full periods of sin⁡\sin, so EE has infinitely many connected components accumulating at 00. Choose a boundary point x1∈(0,1)x_1 \in (0, 1) between two components of EE, with EE on both sides: then no point of NN has xx-coordinate x1x_1 (graph points at x1x_1 sit on the boundary of the ball, and vertical-segment points have x=0x = 0). Hence

N=(N∩{πx<x1})∪(N∩{πx>x1})N = \bigl(N \cap \{\pi_x < x_1\}\bigr) \cup \bigl(N \cap \{\pi_x > x_1\}\bigr)

splits NN into two nonempty, disjoint, open subsets; the first contains (0,y)∈V(0, y) \in V, the second contains the component of EE above x1x_1. So NN is disconnected. Since every neighbourhood of pp in SS contains such an NN, no neighbourhood of pp is connected: SS is not locally connected at any point of VV. ■\blacksquare

Proposition 6.4. The continuous image of a connected space is connected. Therefore connectedness is a topological invariant.

Proof. If f:X→Yf : X \to Y is continuous and XX is connected, suppose f(X)=U∪Vf(X) = U \cup V with U,VU, V open and disjoint in f(X)f(X). Then f−1(U)f^{-1}(U) and f−1(V)f^{-1}(V) are open, disjoint, and cover XX, contradicting connectedness. ■\blacksquare

Proposition 6.5 (Intermediate Value Theorem). If f:X→Rf : X \to \mathbb{R} is continuous and XX is connected, then ff attains every value between any two of its values. This generalises the classical IVT from R\mathbb{R} to any connected space.

Proof. f(X)f(X) is connected (Proposition 6.4) and is a subset of R\mathbb{R}, so by Theorem 6.1 it is an interval. An interval containing two values contains every value between them. The classical IVT is the case X=[a,b]X = [a, b], connected by Theorem 6.1. ■\blacksquare

Proposition 6.6 (Products). The product of connected spaces is connected (with the product topology), for finite and even arbitrary index sets.

Proof. Two factors. Fix (x0,y0)∈X×Y(x_0, y_0) \in X \times Y. For each y∈Yy \in Y consider

Ty=({x0}×Y)∪(X×{y}).T_y = (\{x_0\} \times Y) \cup (X \times \{y\}).

The two pieces are homeomorphic to YY and XX respectively, hence connected, and they share the point (x0,y)(x_0, y). By the pencil lemma, TyT_y is connected. Every TyT_y contains the common set {x0}×Y∋(x0,y0)\{x_0\} \times Y \ni (x_0, y_0), and X×Y=⋃yTyX \times Y = \bigcup_y T_y since (x,y)∈X×{y}⊆Ty(x, y) \in X \times \{y\} \subseteq T_y. The pencil lemma once more: X×YX \times Y is connected.

Arbitrary products. Let P=∏i∈IXiP = \prod_{i \in I} X_i with each XiX_i connected, and fix a=(ai)∈Pa = (a_i) \in P. Let DD be the set of points agreeing with aa in all but finitely many coordinates, and for a finite F⊆IF \subseteq I let

DF={x∈P:xi=ai for i∉F}≅∏i∈FXi,D_F = \{ x \in P : x_i = a_i \text{ for } i \notin F \} \cong \prod_{i \in F} X_i,

which is connected by repeated application of the two-factor case. Each DFD_F contains aa, so D=⋃FDFD = \bigcup_F D_F is connected by the pencil lemma. And DD is dense in PP: a basic open set ∏iUi\prod_i U_i (with Ui=XiU_i = X_i for all ii outside some finite FF) meets DFD_F; pick any point of ∏i∈FUi\prod_{i \in F} U_i in the restricted coordinates and aa elsewhere. So D⊆P⊆D‾D \subseteq P \subseteq \overline{D}, and PP is connected by Proposition 6.7. ■\blacksquare

Note where the two factors are used: the pencil lemma welds the slices to a common spine, and the closure argument promotes finite support to arbitrary support. Neither step needs any separation axioms.

Proposition 6.7 (Closure). If A⊆XA \subseteq X is connected, then any set BB with A⊆B⊆A‾A \subseteq B \subseteq \overline{A} is connected. In particular, the closure of a connected set is connected.

Proof. Let h:B→{0,1}h : B \to \{0, 1\} be continuous into the discrete two-point space; by the two-point characterisation it suffices to show hh is constant. The restriction h∣Ah|_A is constant (say 00), since AA is connected. The set h−1(0)h^{-1}(0) is closed in BB (preimage of the closed - indeed clopen; subset {0}\{0\} of the discrete space) and contains AA, so it contains A‾∩B=B\overline{A} \cap B = B: every point of BB is a limit point of AA, and hh is 00 on all of AA, so by continuity hh is 00 on all of BB. Hence hh is constant, and BB is connected. ■\blacksquare

6.7 Total Disconnectedness and the Cantor Set

Section titled “6.7 Total Disconnectedness and the Cantor Set”

Definition. A space is totally disconnected if its only connected subsets are singletons. Examples include Q\mathbb{Q} (with the subspace topology from R\mathbb{R}) and the Cantor set CC.

Proposition 6.8. The Cantor set CC is totally disconnected, compact, uncountable, and perfect (every point is a limit point). It is homeomorphic to {0,1}N\{0, 1\}^{\mathbb{N}} with the product topology.

Proof. Write C=⋂n=0∞CnC = \bigcap_{n=0}^{\infty} C_n, where C0=[0,1]C_0 = [0, 1] and Cn+1C_{n+1} deletes the open middle third of each interval of CnC_n. Equivalently, CC is the set of reals in [0,1][0, 1] admitting a base-3 (ternary) expansion using only the digits 00 and 22: surviving stage nn means the first nn ternary digits can be chosen in {0,2}\{0, 2\}, and the nested intersection is exactly those xx with such an expansion (the ambiguous case, a trailing digit 11, corresponds to an endpoint that also has an all-{0,2}\{0,2\} expansion, e.g. 1/3=0.13=0.0222…31/3 = 0.1_3 = 0.0222\ldots_3).

Compact. Each CnC_n is a finite union of closed intervals, hence closed, and CC is an intersection of closed sets, hence closed and bounded in R\mathbb{R}; compact by Heine—Borel.

Perfect. The endpoints of the deleted middle thirds belong to every CnC_n; an endpoint of a gap at stage nn is an endpoint of a retained interval at every later stage; so they lie in CC. Now let x∈Cx \in C and ε>0\varepsilon > 0. Choose nn with 3−n<ε3^{-n} < \varepsilon; then xx lies in some stage-nn interval JJ of length 3−n3^{-n}. The two endpoints of JJ are in CC, and at least one of them differs from xx, giving a point of CC within ε\varepsilon of xx. So every point of CC is a limit point of CC.

Uncountable. Define φ:{0,1}N→C\varphi : \{0, 1\}^{\mathbb{N}} \to C by φ((an))=∑n=1∞2an3−n\varphi((a_n)) = \sum_{n=1}^{\infty} 2 a_n 3^{-n}. This is the ternary digit map, and it is a bijection: injective, because if the sequences first differ at index kk, then

∣φ(a)−φ(b)∣≥2⋅3−k−∑n>k2⋅3−n=2⋅3−k−3−k=3−k>0,|\varphi(a) - \varphi(b)| \geq 2 \cdot 3^{-k} - \sum_{n > k} 2 \cdot 3^{-n} = 2 \cdot 3^{-k} - 3^{-k} = 3^{-k} > 0,

so no two sequences map to the same point (the usual expansion ambiguity never arises for digits restricted to {0,2}\{0, 2\}); surjective, because every point of CC has a ternary expansion with digits only 00 and 22, and that expansion is the required preimage. So ∣C∣=∣{0,1}N∣=∣P(N)∣|C| = |\{0, 1\}^{\mathbb{N}}| = |\mathcal{P}(\mathbb{N})|, uncountable by Cantor’s theorem.

Homeomorphism. The map φ\varphi is continuous: if two sequences agree on the first NN digits, their images differ by at most ∑n>N2⋅3−n=3−N\sum_{n > N} 2 \cdot 3^{-n} = 3^{-N}, which is exactly convergence in the product topology (a basic open set restricts finitely many coordinates). A continuous bijection from the compact space {0,1}N\{0,1\}^{\mathbb{N}} (a product of finite discrete spaces) to the Hausdorff space C⊆RC \subseteq \mathbb{R} is a homeomorphism; the criterion proved in the compactness page. So C≅{0,1}NC \cong \{0, 1\}^{\mathbb{N}}.

Total disconnectedness. Let A⊆CA \subseteq C be connected. As a subset of R\mathbb{R}, AA is an interval (Theorem 6.1). But CC contains no interval of positive length: the total length of CnC_n is (2/3)n(2/3)^n, which tends to 00, while A⊆C⊆CnA \subseteq C \subseteq C_n for every nn. An interval of positive length cannot fit inside sets of arbitrarily small total length, so AA is a single point. ■\blacksquare

The digit map is the structural insight: the Cantor set is exactly the countable product of two-point spaces, with the middle-thirds construction supplying the embedding into [0,1][0,1].

6.8 Worked Example: Proving Disconnectedness

Section titled “6.8 Worked Example: Proving Disconnectedness”

Problem. Show that GL⁡(2,R)\operatorname{GL}(2, \mathbb{R}) (the set of 2×22 \times 2 invertible real matrices with the subspace topology from R4\mathbb{R}^4) is disconnected.

Solution

The determinant map det⁡:GL⁡(2,R)→R∖{0}\det : \operatorname{GL}(2, \mathbb{R}) \to \mathbb{R}\setminus\{0\} is continuous (it is a polynomial in the matrix entries). The image is R∖{0}\mathbb{R}\setminus\{0\}, which is disconnected (separated by 00). Since the continuous image of a connected space must be connected, GL⁡(2,R)\operatorname{GL}(2, \mathbb{R}) cannot be connected.

In fact, GL⁡(2,R)\operatorname{GL}(2, \mathbb{R}) has exactly two connected components: matrices with positive determinant and matrices with negative determinant.

■\blacksquare

6.9 Worked Example: Path-Connectedness of Rn∖{0}\mathbb{R}^n \setminus \{0\}

Section titled “6.9 Worked Example: Path-Connectedness of Rn∖{0}\mathbb{R}^n \setminus \{0\}Rn∖{0}”

Problem. For which nn is Rn∖{0}\mathbb{R}^n \setminus \{0\} path-connected?

Solution

For n≥2n \geq 2, any two points p,q≠0p, q \neq 0 can be joined by a path avoiding the origin. For example, if pp and qq are not antipodal, use the straight line segment; if they are antipodal (q=−pq = -p), take a path through a third point. Thus Rn∖{0}\mathbb{R}^n \setminus \{0\} is path-connected for n≥2n \geq 2.

For n=1n = 1, R∖{0}=(−∞,0)∪(0,∞)\mathbb{R} \setminus \{0\} = (-\infty, 0) \cup (0, \infty) is disconnected (and hence not path-connected).

■\blacksquare

6.10 Intuition: What Does Connectedness Mean?

Section titled “6.10 Intuition: What Does Connectedness Mean?”

Connectedness asks whether a space comes in separate pieces. A connected space cannot be split into two non-empty open sets that are disjoint, which means there is no “gap” that separates it. Intuitively, you can travel between any two points without leaving the space. The intermediate value theorem is the most familiar consequence: a continuous function on a connected domain must take every value between its endpoints.

Path-connectedness is a stronger and more intuitive version: a path-connected space has a continuous path between every pair of points. Every path-connected space is connected, but the converse fails for exotic spaces like the topologist’s sine curve, which is connected but has no continuous path crossing the oscillating part. In Rn\mathbb{R}^n, connected subsets are exactly the intervals and their higher-dimensional analogues. Connectedness is the topological tool for proving that certain configurations are impossible: if you can show a space is disconnected, you have proven that it splits into distinct parts that cannot be continuously deformed into each other.

Mistake 1: Assuming that connected implies path-connected. Every path-connected space is connected, but the converse is false. The topologist’s sine curve is a classic example of a connected space that is not path-connected. Do not assume that a space is path-connected just because it is connected.

Mistake 2: Confusing connectedness with convexity. Convexity is a geometric property that implies path-connectedness (and hence connectedness) in Rn\mathbb{R}^n, but connectedness does not imply convexity. For example, the unit circle S1S^1 is connected but not convex. A set can be connected without being convex.

Mistake 3: Forgetting that the continuous image of a connected set is connected. If f:X→Yf: X \to Y is continuous and XX is connected, then f(X)f(X) is connected. This property is often used to prove that certain sets are connected, but it is not an “if and only if” statement. The preimage of a connected set under a continuous map need not be connected.

Mistake 4: Assuming that a disconnected set can be written as a union of disjoint open sets in the ambient space. A set AA is disconnected if it can be written as A=(U∩A)∪(V∩A)A = (U \cap A) \cup (V \cap A) where UU and VV are disjoint open sets in the ambient space. The sets UU and VV are open in the ambient space, not necessarily in AA. Do not confuse the subspace topology with the ambient topology.

Mistake 5: Confusing total disconnectedness with disconnectedness. A space is totally disconnected if its only connected subsets are singletons. A space can be disconnected without being totally disconnected. For example, [0,1]∪[2,3][0, 1] \cup [2, 3] is disconnected but not totally disconnected, since [0,1][0, 1] and [2,3][2, 3] are connected subsets.