Connected Sets

This scene is about connected sets. Move the cut point and check whether the set splits into two open pieces that never touch. A set that cannot be split is connected. An interval never splits: wherever you cut, the cut point itself belongs to neither side. Two far-apart pieces, or the rationals with their dense gaps, split cleanly when the cut lands in a gap. For example, look at one 🧩 jigsaw puzzle. While the pieces interlock, it is a single board. Lying in two completely separate piles, it is already split apart.

X=U∪V, U∩V=∅, U,V open in X, U,V≠∅  ⟹  X disconnectedX = U \cup V,\ U \cap V = \emptyset,\ U, V \text{ open in } X,\ U, V \neq \emptyset \implies X \text{ disconnected}
Q∩[0,1]={q∈Q:q<2/2}∪{q∈Q:q>2/2}\mathbb{Q} \cap [0,1] = \{q \in \mathbb{Q} : q < \sqrt{2}/2\} \cup \{q \in \mathbb{Q} : q > \sqrt{2}/2\}

Examples & Exercises

Prove completely that the interval [0,1][0,1] is connected. That is, show that if U,VU, V are disjoint sets, both open in [0,1][0,1], with [0,1]=U∪V[0,1] = U \cup V, then at least one of U,VU, V is empty.

[0,1]=U∪V, U∩V=∅, U,V open in [0,1]  ⟹  U=∅ or V=∅[0,1] = U \cup V,\ U \cap V = \emptyset,\ U, V \text{ open in } [0,1] \implies U = \emptyset \text{ or } V = \emptyset
t=0.5t = 0.5: the cut point 0.50.5 itself belongs to neither side, so the pair does not cover — every cut of [0,1][0,1] fails this way.
Show proof
  1. Suppose [0,1]=U∪V[0,1] = U \cup V is a separation, with U,VU, V both nonempty. Renaming if needed, assume 0∈U0 \in U (00 lies in one of U,VU, V, and the symmetry x↦1−xx \mapsto 1 - x makes the two cases identical). Put S={x∈[0,1]:[0,x]⊆U}S = \{x \in [0,1] : [0,x] \subseteq U\} and c=sup⁡Sc = \sup S. Since 0∈S0 \in S, S≠∅S \ne \emptyset; it is bounded above by 11, so c∈[0,1]c \in [0,1] exists by completeness.
  2. First, c∈Uc \in U. Since VV is open in [0,1][0,1], its complement U=[0,1]∖VU = [0,1] \setminus V is closed in [0,1][0,1] — that is, UU equals its own closure in the subspace: U‾∩[0,1]=U\overline{U} \cap [0,1] = U. If c=0c = 0 then c=0∈Uc = 0 \in U. If c>0c > 0, then for every ε>0\varepsilon > 0 the definition of supremum gives x∈Sx \in S with c−ε<x≤cc - \varepsilon < x \le c. That xx satisfies x∈[0,x]⊆Ux \in [0,x] \subseteq U, so the neighborhood (c−ε,c+ε)(c-\varepsilon, c+\varepsilon) meets UU, and hence c∈U‾c \in \overline{U}. Therefore c∈U‾∩[0,1]=Uc \in \overline{U} \cap [0,1] = U.
  3. Next, [0,c]⊆U[0,c] \subseteq U: if y<cy < c then the definition of supremum gives x∈Sx \in S with y<x≤cy < x \le c, so y∈[0,x]⊆Uy \in [0,x] \subseteq U; and the endpoint cc itself lies in UU by the previous step. Now c=1c = 1. Suppose instead c<1c < 1. Since c∈Uc \in U and UU is open in [0,1][0,1], there is δ>0\delta > 0 with (c−δ,c+δ)∩[0,1]⊆U(c-\delta, c+\delta) \cap [0,1] \subseteq U. Put δ′=min⁡(δ/2,(1−c)/2)>0\delta' = \min(\delta/2, (1-c)/2) > 0 (positive since c<1c < 1). Then [0,c+δ′]=[0,c]∪(c,c+δ′]⊆U[0, c+\delta'] = [0,c] \cup (c, c+\delta'] \subseteq U. So c+δ′∈Sc + \delta' \in S, yet c+δ′>c=sup⁡Sc + \delta' > c = \sup S — a contradiction. Hence c=1c = 1.
  4. Hence [0,1]=[0,c]⊆U[0,1] = [0,c] \subseteq U. But a separation requires V≠∅V \ne \emptyset, so some v∈Vv \in V exists. Then v∈[0,1]⊆Uv \in [0,1] \subseteq U and v∈Vv \in V, giving U∩V≠∅U \cap V \ne \emptyset — contradicting disjointness. Therefore no separation of [0,1][0,1] exists, and [0,1][0,1] is connected. This is why every cut in the scene fails: a cut point t∈[0,1]t \in [0,1] can belong to neither open side, and this proof shows the same phenomenon defeats every open pair, not just interval cuts.

Prove completely that if XX is path-connected then XX is connected. Path-connected means that for any a,b∈Xa, b \in X there is a continuous γ:[0,1]→X\gamma : [0,1] \to X with γ(0)=a\gamma(0) = a and γ(1)=b\gamma(1) = b. You may use Problem 1 (connectedness of [0,1][0,1]).

γ:[0,1]→X continuous,A=γ−1(U),B=γ−1(V)\gamma : [0,1] \to X \text{ continuous},\quad A = \gamma^{-1}(U),\quad B = \gamma^{-1}(V)
[0,1]=A∪B,A∩B=∅,A,B open in [0,1][0,1] = A \cup B,\quad A \cap B = \emptyset,\quad A, B \text{ open in } [0,1]
X=[0,1]∪[2,3]X = [0,1] \cup [2,3], t=1.5t = 1.5: a space where a separation succeeds — no path can cross between the pieces, for its preimage would separate [0,1][0,1].
Show proof
  1. Let XX be path-connected, and suppose for a contradiction that X=U∪VX = U \cup V is a separation: U,VU, V open in XX, disjoint, and both nonempty. Pick a∈Ua \in U and b∈Vb \in V. By path-connectedness there is a continuous γ:[0,1]→X\gamma : [0,1] \to X with γ(0)=a\gamma(0) = a and γ(1)=b\gamma(1) = b.
  2. Put A=γ−1(U)A = \gamma^{-1}(U) and B=γ−1(V)B = \gamma^{-1}(V). Since γ\gamma is continuous and U,VU, V are open in XX, both AA and BB are open in [0,1][0,1]: if t0∈At_0 \in A then γ(t0)∈U\gamma(t_0) \in U; openness of UU gives ε>0\varepsilon > 0 with UU containing the ε\varepsilon-neighborhood of γ(t0)\gamma(t_0); continuity then sends a whole δ\delta-neighborhood of t0t_0 into that ε\varepsilon-neighborhood, so the δ\delta-neighborhood lies in AA. The same argument works for BB.
  3. Check the four separation conditions in [0,1][0,1]. Cover: for every t∈[0,1]t \in [0,1], γ(t)∈X=U∪V\gamma(t) \in X = U \cup V, so t∈A∪Bt \in A \cup B; that is, [0,1]=A∪B[0,1] = A \cup B. Disjoint: if t∈A∩Bt \in A \cap B then γ(t)∈U∩V=∅\gamma(t) \in U \cap V = \emptyset, impossible; so A∩B=∅A \cap B = \emptyset. Nonempty: 0∈A0 \in A (since γ(0)=a∈U\gamma(0) = a \in U) and 1∈B1 \in B (since γ(1)=b∈V\gamma(1) = b \in V). Thus [0,1]=A∪B[0,1] = A \cup B is a separation of [0,1][0,1].
  4. This contradicts Problem 1 — [0,1][0,1] is connected, so it admits no separation. Hence no separation X=U∪VX = U \cup V of XX can exist, and XX is connected. The converse is false: the closure of the topologist's sine curve is connected but not path-connected.

Prove completely that every subset A⊆QA \subseteq \mathbb{Q} with at least two points is disconnected. Conclude that Q\mathbb{Q} is totally disconnected: its only connected subsets are singletons and ∅\emptyset.

α=p+q−p2,2=q−pα−p∉Q\alpha = p + \frac{q - p}{\sqrt{2}}, \qquad \sqrt{2} = \frac{q - p}{\alpha - p} \notin \mathbb{Q}
A=(A∩(−∞,α))∪(A∩(α,∞))A = \bigl(A \cap (-\infty, \alpha)\bigr) \cup \bigl(A \cap (\alpha, \infty)\bigr)
The cut c=2/2≈0.7071c = \sqrt{2}/2 \approx 0.7071: being irrational, it hits no rational point, so it covers and splits Q∩[0,1]\mathbb{Q} \cap [0,1].
Show proof
  1. Let A⊆QA \subseteq \mathbb{Q} have at least two points, and pick p,q∈Ap, q \in A with p<qp < q. Put α=p+q−p2\alpha = p + \frac{q - p}{\sqrt{2}}. Since 0<12<10 < \frac{1}{\sqrt{2}} < 1 (that is, 1<2<21 < \sqrt{2} < 2), α\alpha is a strict convex combination of pp and qq: p<α<qp < \alpha < q.
  2. The number α\alpha is irrational. Suppose α∈Q\alpha \in \mathbb{Q}. Then α−p∈Q\alpha - p \in \mathbb{Q}, and α−p=q−p2>0\alpha - p = \frac{q - p}{\sqrt{2}} > 0, so it is nonzero; rearranging gives 2=q−pα−p\sqrt{2} = \frac{q - p}{\alpha - p}. Since q−p∈Qq - p \in \mathbb{Q} is nonzero, this ratio is rational — contradicting the irrationality of 2\sqrt{2}. Hence α∉Q\alpha \notin \mathbb{Q}; in particular, α\alpha equals no element of AA.
  3. Put U=A∩(−∞,α)U = A \cap (-\infty, \alpha) and V=A∩(α,∞)V = A \cap (\alpha, \infty). Open: (−∞,α)(-\infty, \alpha) and (α,∞)(\alpha, \infty) are open rays in R\mathbb{R}, so UU and VV are open in the subspace AA. Disjoint: x<αx < \alpha and x>αx > \alpha are incompatible, so U∩V=∅U \cap V = \emptyset. Cover: if a∈Aa \in A then a∈Qa \in \mathbb{Q}, so a≠αa \ne \alpha, and therefore a<αa < \alpha (so a∈Ua \in U) or a>αa > \alpha (so a∈Va \in V). That is, A=U∪VA = U \cup V. Nonempty: p∈Up \in U (as p<αp < \alpha) and q∈Vq \in V (as α<q\alpha < q).
  4. Thus A=U∪VA = U \cup V is a separation of AA, so AA is disconnected. Since AA was an arbitrary subset of Q\mathbb{Q} with at least two points, every connected subset of Q\mathbb{Q} has at most one point — Q\mathbb{Q} is totally disconnected. The scene's cut 2/2\sqrt{2}/2 is exactly this α\alpha for p=0p = 0, q=1q = 1: α=12=22\alpha = \frac{1}{\sqrt{2}} = \frac{\sqrt{2}}{2}. Between any two rationals an irrational slips in and splits the set.

5. Topology