MATH 5345H --- Week 7: The quotient topology and identification spaces

← Back to Lecture Notes

0.1  (§22) The quotient topology

  • Rmk. Injections gave us subspaces and embeddings. Surjections will give quotients — the dual construction, with finest in place of coarsest.

  • Def. An equivalence relation on X: reflexive, symmetric, transitive. Classes [x]={yxy}; they are nonempty, cover X, are mutually disjoint. X/={[x]}, canonical surjection π(x)=[x].

  • Lem. Any surjection f:XY arises this way: set xyf(x)=f(y); then h([x])=f(x) is a bijection X/Y with f=hπ.

  •  Def. Quotient topology on Y from a surjection f:XY: UY openf1(U) open in X. A surjection carrying this topology is a quotient map.

    Proof. It is a topology because f1 commutes with and . Quotient maps are continuous by construction.

  • Lem. The quotient topology is the finest topology on Y making f continuous.

  • Lem. f surjective is a quotient map (AY closed f1(A) closed).

  • Lem. A bijective quotient map is a homeomorphism, and conversely.

  • Def. f is an open map if f(U) is open for all open U; a closed map if f(A) is closed for all closed A.

  • Lem. f open f(B) open for every basis element B.

  •  Lem. A surjective open map is a quotient map. So is a surjective closed map.

    Proof.U=f(f1(U)) by surjectivity; then apply openness (resp. closedness).

  •  Ex. f:[0,1]S1, f(t)=(cos2πt,sin2πt): continuous, surjective, closed (compactness), hence a quotient map. Not open. With 01: [0,1]/S1.

  •  Ex. Torus. g=f×f:[0,1]2S1×S1 is a quotient map; identify (s,0)(s,1) and (0,t)(1,t). Classes: the four corners; the paired edge points; the interior singletons. So T2[0,1]2/.

  • Ex. {n,z,p} by sign. Quotient topology {,{n},{p},{n,p},Y}: not Hausdorff; z is the only closed point.

  • Ex. π1:2 is open (hence a quotient map) but not closed: the hyperbola C={xy=1} is closed, π1(C)={0} is not.

  • Caution. Ex. Restricting a quotient map can destroy the property. With A=C{(0,0)}, the map π1|A:A is a continuous surjection but not a quotient map: {0} is not open in , yet its preimage {(0,0)} is open in A.

  • Thm. f:XY a quotient map, BY, A=f1(B) saturated, g=f|A:AB. Then g is a quotient map if either (1) A is open, or A is closed; or (2) f is an open map, or a closed map.

    Proof. Two identities do the work: g1(V)=f1(V) for VB, and g(AU)=Bf(U). Both use A=f1(B).

  • Rmk. Composites of quotient maps are quotient maps. Caution. Products of quotient maps need not be; some local compactness hypothesis is required. Quotients of Hausdorff spaces need not be Hausdorff.

  •  Thm. (Universal property.) f:XY a quotient map, h:XZ constant on the fibres of f. Then h factors uniquely as h=gf; moreover g is continuous h is, and g is a quotient map h is.

  • Cor. h:XZ a continuous surjection, xyh(x)=h(y). Then h induces a continuous bijection g:X/Z; g is a homeomorphism h is a quotient map; and if Z is Hausdorff so is X/.

1  Connectedness and Compactness

1.1  (§23) Connected spaces

  • Def. For disjoint spaces C,D: the disjoint union CD carries the topology consisting of those W with CW open in C and DW open in D; the finest making both inclusions continuous.

  • Rmk. CD is also called the sum, or coproduct, of C and D; it has a universal property dual to that of the product C×D. Every X is trivially CD with C= or D=; if XCD non-trivially, X is disconnected.

  •  Def. A separation of X: a pair U,V of disjoint nonempty open sets with UV=X. X is connected if no separation exists.

  • Rmk. Being connected is a topological property. The empty space needs care: some authors declare not connected, much as 1 has no proper factors yet is not counted as a prime.

  •  Lem. X is connected the only clopen subsets are and X.

    Proof. In a separation V=XU, so “U,V both open” =U clopen”, and “both nonempty” =U,X”. Cleanest working form of the definition.

  • Lem. U,V a separation XUV; conversely X=CD with both nonempty gives a separation.

  • Lem. Equivalent formulation: a separation is a pair of disjoint nonempty sets A,B with AB=X, neither containing a limit point of the other (A¯B==AB¯).

  • Ex. One-point spaces are connected. Sierpiński {a,b} is connected ({a} open not closed, {b} closed not open). [1,0)(0,1] is disconnected.

  • Rmk. Coming in §24: is connected, as is every interval [a,b], [a,b), (a,b], (a,b) for ab.

  • Ex. Every X with 2 points is disconnected: pick irrational a between p<q in X and cut at a.

  •  Lem. U,V a separation of X, AX connected AU or AV.

    Proof.AU is clopen in A. The single most-used lemma of the section.

  •  Thm. A union of connected subspaces with a point in common is connected.

  •  Thm. ABA¯, A connected B connected. (Adding limit points cannot disconnect.)

  •  Thm. The continuous image of a connected space is connected.

1.2  (§24) Connected subspaces of the real line

  • Def. C is convex if a<b in C [a,b]C.

  • Ex. The convex subsets of : ; the intervals (a,b),[a,b),(a,b],[a,b]; the rays; and .

  •  Thm. Every convex C is connected. In particular is connected.

    Proof. Reduce to [a,b]. Given [a,b]=UV with aU,bV, let c=supU. Then cU¯=U, so cb; but U open gives points >c in U — contradicting the supremum. Least upper bound property is doing all the work.

  •  Thm. (Intermediate value theorem.) f:X continuous, X connected, r between f(a) and f(b) c with f(c)=r.

    Proof. Otherwise f1(,r) and f1(r,) separate X.

  • Def. A path from x to y: a map f:[a,b]X, f(a)=x, f(b)=y. X is path connected if any two points are joined by a path.

  •  Lem. Path connected connected.

    Proof. A separation of X would pull back to a separation of [a,b].

  • Lem. The continuous image of a path connected space is path connected.

  • Def. C in a real vector space is convex if (1t)x+tyC for x,yC, t[0,1].

  • Ex. Convex subsets of n are path connected — e.g. the unit ball Bn={x1}.

  • Ex. n{0} is path connected for n2 (not for n=1). Hence Sn1={x=1} is path connected for n2, being the image of xx/x. (S0={±1} is not.)

  •  Ex. Topologist’s sine curve. S={(x,sin(1/x))0<x1},   S¯=SV with V={0}×[1,1]. S is connected (continuous image of (0,1]), hence so is S¯. But S¯ is not path connected.

    Proof. A path from V into S: reparametrise so f(0)V, f(t)S for t>0. Write f=(x,y). Construct tn0 with y(tn)=(1)n via the IVT applied to x. Then y(tn) does not converge — contradicting continuity.