A Brief Note on Quotient Maps
And how they relate to quasi-homeomorphisms
(PDF version of this post.)
In the previous post I introduced quasi-homeomorphisms; these are related to another important topological concept, quotient maps. Quasi-homeoΒmorphisms are in fact just a special kind of quotient map, one that discriminates as finely as is possible for any continuous function on a given domain.
Definition. A quotient map is a surjective (onto) function π : π β π between topological spaces such that any set π β π is open if and only if πβ»ΒΉ[π] β π is open.
This definition ensures that the topology on π is precisely the one βinducedβ by π from π, with no extra or missing open sets. This definition implies that quotient maps are continuous: a subset π of π is open only if πβ»ΒΉ[π] is open in π. Since a quotient map π must be continuous, it is required that π(π₯) = π(π¦) whenever π₯ βΌ π¦ (π₯ and π¦ are topologically indistinguishable).
Quotient maps and equivalence relations
Quotient maps are related to equivalence relations. Recall that an equivalence relation β is a sort of poor-manβs equality, one that throws away some of the information in its arguments but otherwise behaves like equality in that it is reflexive (π₯ β π₯), commutative (π₯ β π¦ iff π¦ β π₯), and transitive (π₯ β π¦ and π¦ β π§ implies π₯ β π§). Examples weβve seen include equality ( = ), logical equivalence of propositional formulas ( β‘ ), and topological indistinguishability ( βΌ ).
The set of values equivalent to π₯ is the equivalence class of π₯, written
or just [π₯].1 The entire collection of equivalence classes for an equivalence relation β on a set π is known as the quotient space that β induces on π, written π/β. (Itβs called the quotient space because it divides upβpartitionsβthe space π into equivalence class.) The quotient topology on the quotient space defines a set π of equivalence classes to be open iff their union is an open set in π.
For a given equivalence relation β on π, if we define π : π β (π/β) such that π(π₯) = [π₯], then π is a quotient map:
π is surjective: every πΈ β (π/β) is an equivalence class, and every equivalence class πΈ has at least one member π₯ β π, so πΈ = π(π₯).
π β (π/β) is open iff the union of its elements is open in π. But the union of the elements of π is the set of all π₯ β π whose equivalence class is in π; that is, the set of all π₯ β π such that π(π₯) β π, and this is just the definition of πβ»ΒΉ[π]. So π is open iff πβ»ΒΉ[π] is open.
Conversely, any quotient map π : π β π induces an equivalence relation on π: define π₯ β π¦ iff π(π₯) = π(π¦). Then π is homeomorphic to (π/β), that is, there is a bijection π : (π/β) β π that pairs up corresponding elements of (π/β) and π, such that both π and its inverse πβ»ΒΉ are continuous.
For a geometric example of a quotient map, consider the set πΒΉ β βΒ², defined to be the unit circle:
Then the function π : β β πΒΉ defined by
is a quotient map, assuming the standard topologies on β and βΒ², and assuming for πΒΉ the subset topology induced by βΒ². If we think of π‘ as time, and π(π‘) as the position of a dot on the plane at time π‘, then the dot moves counterclockwise in a circle at a constant speed of one unit of distance per unit of time (see figure below). Since π(π ) = π(π‘) iff π and π‘ differ by a multiple of 2π, the equivalence relation associated with π is then π β π‘ iff π β‘ π‘ (mod 2π).
Note that if we had written π : β β βΒ² then π would not be a quotient map, because it would not be surjective. The codomain π (set of potential values) of a function π : π β π must be considered part of its identity, not just its range (set of actual values attained).
Although we have defined a quotient map π : π β π in terms of the topology on π, one often goes in the other direction: start with a surjective function π and then define the topology on π to be the one that makes π a quotient map:
That is exactly how the quotient topology on a quotient space was defined, starting with the function π.
Quotient maps and quasi-homeomorphisms
Recall the definition we gave for a quasi-homeomorphism: a function π : π β π between topological spaces having the following properties:
π is surjective (onto).
π(π₯) = π(π¦) iff π₯ βΌ π¦ (π₯ and π¦ are topologically indistinguishable).
π preserves open sets both ways:
If π β π is open then π[π] β π is open.
If π β π is open then πβ»ΒΉ[π] β π is open.
Property 2 can be simplified to βπ(π₯) = π(π¦) only if π₯ βΌ π¦,β as the βifβ direction is implied by Property 3b (continuity). But once we have the concept of a quotient map, we can further simplify the definition:
Proposition. A quasi-homeomorphism π is a quotient map such that π(π₯) = π(π¦) only when π₯ βΌ π¦.
Proof. ( β ) Suppose π is a quasi-homeomorphism. Then it is surjective (Property 1). If π β π is open then πβ»ΒΉ[π] β π is open (Property 3b). If πβ»ΒΉ[π] is open then, by Property 3a, π[πβ»ΒΉ[π]] = π is open. So π is a quotient map. And, of course, Property 2 guarantees that π(π₯) = π(π¦) only when π₯ βΌ π¦.
(β) Suppose that π is a quotient map such that π(π₯) = π(π¦) only if π₯ βΌ π¦. Since π is surjective, Property 1 is satisfied. Since π is continuous, Property 3b holds; furthermore, continuity implies that π(π₯) = π(π¦) if π₯ βΌ π¦, and so both the βifβ and βonly ifβ directions of Property 2 hold.
It remains only to show that Property 3a holds. Suppose that π β π is open. Since π is a quotient map, π[π] is open iff πβ»ΒΉ[π[π]] is open. If we can show that π = πβ»ΒΉ[π[π]], then Property 3a does indeed hold. But π β πβ»ΒΉ[π[π]] only if there is some π₯ β π and π¦ β π with π(π₯) = π(π¦). But π(π₯) = π(π¦) implies that π₯ and π¦ are topologically indistinguishable, and π is open, so they either both belong to π or neither belongs to π. β
The codomain of a quasi-homeomorphism is πβ
In the previous article I claimed that if π : π β π is a quasi-homeomorphism, then π must be a πβ space: all distinct points of π are topologically distinguishable. I did not give a proof, thinking it was a simple one-liner, but upon looking more closely I realized that the proof is more involved than that. Therefore Iβll give the proof now.
Theorem. If π : π β π is a quasi-homeomorphism then π is a πβ space.
Proof. Suppose that π‘β, π‘β β π and π‘β β π‘β. We will show that π‘β and π‘β are topologically distinct.
Since π is surjective there exist π β, π β β π with π(π β) = π‘β and π(π β) = π‘β. As π‘β β π‘β it follows that π(π β) β π(π β), so by Property 2 the points π β and π β are topologically distinguishable. Therefore there is an open set containing one but not the other; without loss of generality, assume that there is an open set π β π with π β β π and π β β π. By Property 3a, π[π] is an open set in π, and clearly π‘β = πβ(π β) β π[π].
For any π β² β π, if π(π β²) = π‘β then π(π β²) = π(π β), hence π β² βΌ π β, hence π β² β π. Therefore π‘β β π[π], and we see that π[π] is an open set containing π‘β but not π‘β, hence π‘β and π‘β are topologically distinguishable. β
One last note
One thing I may not have made clear enough is the relation between homeomorphisms and quasi-homeomorphisms. If every pair of points in space π is topologically distinguishableβπ is a πβ spaceβthen a quasi-homeomorphism from π to π is exactly the same thing as a homeomorphism from π to π. And if π does contain any pair of points that are topologically indistinguishable, then it is impossible to have a homeomorphism from π to π, because any continuous function π : π β π must map topologically indistinguishable points in π to the same point in π, and hence can not be one-to-one in this case.
In summary, for any topological space π, either
π is a πβ space, so homeomorphisms π β π and quasi-homeomorphisms π β π are the same thing; or
π is not a πβ space, so there exists no homeomorphism π β π.
Yes, that conflicts with our notation for the set of satisfying truth assignments for a propositional formulaβalthough [π] (the equivalence class) and [π] (the set of satisfying truth assignments) are closely related.)



Brilliant framing of quasi-homeomorphisms through the quotient map lens. The way this reframes Property 3a by showing U = fβ»ΒΉ[f[U]] when dealing with topologically indistingusihable points really clicked for me. I once got stuck tryingto prove something similiar in a measure theory context, and seeing the connection to equivalence relations would've saved me hours. The unit cirlce example nails why we care about codomain versus range too.