TheoremComplete

Quotient Map Characterization

Quotient maps are the central notion for constructing new spaces from old ones by identification. This theorem provides multiple equivalent conditions for recognizing quotient maps, which is essential since verifying the definition directly can be difficult.


Main Characterization

Theorem3.8Characterization of Quotient Maps

Let p:XYp: X \to Y be a surjective continuous map. The following are equivalent:

  1. pp is a quotient map (i.e., UYU \subseteq Y is open     \iff p1(U)p^{-1}(U) is open in XX).
  2. CYC \subseteq Y is closed     \iff p1(C)p^{-1}(C) is closed in XX.
  3. pp maps saturated open sets to open sets.
  4. pp maps saturated closed sets to closed sets.

Recall that SXS \subseteq X is saturated with respect to pp if S=p1(p(S))S = p^{-1}(p(S)).

Proof

(1 \Leftrightarrow 2): Taking complements: UU is open in YY iff YUY \setminus U is closed in YY, and p1(U)p^{-1}(U) is open in XX iff p1(YU)=Xp1(U)p^{-1}(Y \setminus U) = X \setminus p^{-1}(U) is closed in XX. So condition (1) for open sets is equivalent to condition (2) for closed sets.

(1 \Leftrightarrow 3): Suppose (1) holds. If UXU \subseteq X is saturated and open, then U=p1(p(U))U = p^{-1}(p(U)), so p1(p(U))p^{-1}(p(U)) is open in XX. By (1), p(U)p(U) is open in YY.

Conversely, suppose (3) holds. Since pp is continuous, the forward direction of (1) is clear. For the converse, suppose p1(V)p^{-1}(V) is open in XX for some VYV \subseteq Y. The set p1(V)p^{-1}(V) is saturated (preimages are always saturated), so by (3), p(p1(V))=Vp(p^{-1}(V)) = V (using surjectivity) is open in YY.

(2 \Leftrightarrow 4): Identical argument with "open" replaced by "closed."


Sufficient Conditions

Theorem3.9Sufficient Conditions for Quotient Maps

Each of the following is sufficient for a continuous surjection p:XYp: X \to Y to be a quotient map:

  1. pp is an open map.
  2. pp is a closed map.
  3. pp admits a continuous right inverse (i.e., a continuous section s:YXs: Y \to X with ps=idYp \circ s = \operatorname{id}_Y).
Proof

(1): If pp is open, then for any saturated open UXU \subseteq X, p(U)p(U) is open. This is condition (3) above.

(2): If pp is closed, then for any saturated closed CXC \subseteq X, p(C)p(C) is closed. This is condition (4) above.

(3): Suppose s:YXs: Y \to X is a continuous section. For any VYV \subseteq Y: if p1(V)p^{-1}(V) is open in XX, then V=p(s(V))p(p1(V))V = p(s(V)) \subseteq p(p^{-1}(V)). But also s1(p1(V))=(ps)1(V)=Vs^{-1}(p^{-1}(V)) = (p \circ s)^{-1}(V) = V, and since ss is continuous, VV is open if p1(V)p^{-1}(V) is open. (More precisely: V=s1(p1(V))V = s^{-1}(p^{-1}(V)) is open since ss is continuous.)


Quotients and Composition

Theorem3.10Composition of Quotient Maps

If p:XYp: X \to Y and q:YZq: Y \to Z are quotient maps, then qp:XZq \circ p: X \to Z is a quotient map.

Proof

First, qpq \circ p is surjective (composition of surjections) and continuous (composition of continuous maps). Let WZW \subseteq Z and suppose (qp)1(W)=p1(q1(W))(q \circ p)^{-1}(W) = p^{-1}(q^{-1}(W)) is open in XX. Since pp is a quotient map, q1(W)q^{-1}(W) is open in YY. Since qq is a quotient map, WW is open in ZZ.


Quotient Maps and the Universal Property

Theorem3.11Passing to the Quotient

Let p:XYp: X \to Y be a quotient map and f:XZf: X \to Z a continuous map that is constant on the fibers of pp (i.e., p(x1)=p(x2)    f(x1)=f(x2)p(x_1) = p(x_2) \implies f(x_1) = f(x_2)). Then there exists a unique continuous map fˉ:YZ\bar{f}: Y \to Z with f=fˉpf = \bar{f} \circ p.

XfZpfˉY\begin{array}{ccc} X & \xrightarrow{f} & Z \\ \downarrow^{p} & \nearrow_{\bar{f}} & \\ Y & & \end{array}

Moreover:

  • fˉ\bar{f} is a quotient map if ff is a quotient map.
  • fˉ\bar{f} is injective if and only if ff is constant on fibers and distinct fibers have distinct images.
  • fˉ\bar{f} is a homeomorphism if and only if ff is a quotient map and fˉ\bar{f} is a bijection.
Proof

Existence: Define fˉ(y)=f(x)\bar{f}(y) = f(x) for any xp1(y)x \in p^{-1}(y). This is well-defined since ff is constant on fibers.

Uniqueness: If fˉp=f=gˉp\bar{f} \circ p = f = \bar{g} \circ p, then fˉ(y)=fˉ(p(x))=f(x)=gˉ(p(x))=gˉ(y)\bar{f}(y) = \bar{f}(p(x)) = f(x) = \bar{g}(p(x)) = \bar{g}(y) for any xp1(y)x \in p^{-1}(y).

Continuity: For WW open in ZZ, p1(fˉ1(W))=(fˉp)1(W)=f1(W)p^{-1}(\bar{f}^{-1}(W)) = (\bar{f} \circ p)^{-1}(W) = f^{-1}(W), which is open since ff is continuous. Since pp is a quotient map, fˉ1(W)\bar{f}^{-1}(W) is open.

ExampleInduced Map on Projective Space

The map f:SnRPnf: S^n \to \mathbb{R}P^n given by f(x)=[x]f(x) = [x] (equivalence class under the antipodal relation xxx \sim -x) is a quotient map. If g:SnYg: S^n \to Y is continuous with g(x)=g(x)g(x) = g(-x) for all xx, then gg factors uniquely through RPn\mathbb{R}P^n: there exists a unique continuous gˉ:RPnY\bar{g}: \mathbb{R}P^n \to Y with g=gˉfg = \bar{g} \circ f.