Separating curves, absorption, and the two cells of a crosscut #
Definition 2.4 calls a Jordan curve separating when its complement has exactly two regions, one bounded and one unbounded, each with the curve as its boundary. Everything Part I proves about a polygonal curve and Part II proves about a general one is packaged in that single predicate, and the two lemmas below are the ones needed on both sides of the divide: they are proved from the definition alone.
The interface: inside and outside are total functions #
The blueprint writes Int(C) and Ext(C) for the two regions and its consumers — the two
crosscut theorems — name them in their statements. So the two regions must be functions of
C, not sets produced by an existential inside the definition of separation. They are defined
here for an arbitrary set: inside C is the union of the bounded components of the complement
and outside C the union of the unbounded ones; together they partition Cᶜ.
Separation is then the assertion that each of the two halves is one region with boundary C
(Definition 2.4); that one is bounded and the other unbounded is then automatic, and so is
"exactly two regions", since a connected clopen piece of Cᶜ is a component of it.
IsRegionOf C Ω says Ω is one of the two, and IsRegionPair C Ω Ω' that Ω, Ω' are the two
in one order or the other. That is what lets Lemma 2.6 be stated once instead of twice: the
blueprint's Ω is sometimes Int(C) and sometimes Ext(C), and every step of its proof is
symmetric in that choice.
What the crosscut lemma actually needs #
Lemma 2.6 is stated for a crosscut P of a separating curve C with arcs A₁, A₂, but its
proof never uses that P and the Aᵢ are arcs, nor the endpoints p, q, nor even that
C = A₁ ∪ A₂. What it uses is: Aᵢ ⊆ C, P ∩ C ⊆ Aᵢ (the crosscut meets the curve only in
the shared endpoints, which lie on both arcs), Jᵢ = Aᵢ ∪ P separating, and A₁ ≠ A₂. The
lemmas below are therefore stated at that level of generality, with arcs_ne supplying
A₁ ≠ A₂ from the arc picture and isJordanCurve_union supplying the sentence "these are
Jordan curves" from IsJordanCurve.of_two_arcs.
One further hypothesis of the blueprint is dropped: Ω is described there as the region
containing P ∖ {p, q}, but what actually pins the roles of Ω, Ω†, Wᵢ and Vᵢ is the
inclusion Ω† ⊆ Wᵢ, which is a hypothesis here. A consumer supplies it by
IsSeparating.exists_isRegionPair_subset, which is the formal content of the blueprint's
"being connected it lies in one region of ℝ² ∖ Jᵢ, so Wᵢ and Vᵢ are well defined".
Blueprint #
inside,outside,IsSeparating— Definition 2.4 (separating Jordan curve).IsSeparating.absorption— Lemma 2.5 (absorption).crosscut_cells— Lemma 2.6 (crosscut cells), all three parts, with the per-part lemmascell_subset_region_diff,cell_isComponent,closure_cell_inter_curvestated separately because the bundled form carries the hypotheses of both indices at once.
The inside and the outside of a set #
Both are defined for an arbitrary set, without any separation hypothesis: inside C collects
the points whose component in the complement is bounded, outside C the rest of the
complement.
The union of the bounded components of the complement — the blueprint's Int(C) once C
is known to be separating.
Equations
- Schoenflies.inside C = {x : Schoenflies.Plane | x ∉ C ∧ Bornology.IsBounded (connectedComponentIn Cᶜ x)}
Instances For
The union of the unbounded components of the complement — the blueprint's Ext(C) once C
is known to be separating.
Equations
- Schoenflies.outside C = {x : Schoenflies.Plane | x ∉ C ∧ ¬Bornology.IsBounded (connectedComponentIn Cᶜ x)}
Instances For
Boundedness of the component is constant along a component, so the inside is a union of whole components.
Definition 2.4: a separating Jordan curve #
Definition 2.4. A Jordan curve is separating when its complement has exactly two regions,
one bounded and one unbounded, each with boundary C.
Since inside C and outside C always partition the complement into the bounded-component
part and the unbounded-component part, "exactly two regions, one bounded and one unbounded"
says precisely that each part is a single region; and the boundedness of one and unboundedness
of the other are then consequences rather than hypotheses (IsSeparating.isBounded_inside,
IsSeparating.not_isBounded_outside).
- isJordanCurve : IsJordanCurve C
The set is a Jordan curve.
- isConnected_inside : IsConnected (inside C)
The bounded part of the complement is a single region.
- isConnected_outside : IsConnected (outside C)
The unbounded part of the complement is a single region.
The inside has the curve as its boundary.
The outside has the curve as its boundary.
Instances For
The inside is bounded: it is connected and lies inside the complement, hence inside the
component of any of its points, which is bounded by definition of inside.
The outside is unbounded: it contains an unbounded component.
The inside really is one region of the complement, that is, a connected component of it: it is a union of components which is itself connected. Together with its twin below this is the "exactly two regions" clause of Definition 2.4.
The outside is a connected component of the complement.
The two regions, named #
IsRegionOf C Ω is "Ω is a region of ℝ² ∖ C" and IsRegionPair C Ω Ω' is "Ω and Ω'
are the two regions". Both are stated without a separation hypothesis; the facts that make
them useful carry one.
Ω is a region of the complement of C.
Equations
- Schoenflies.IsRegionOf C Ω = (Ω = Schoenflies.inside C ∨ Ω = Schoenflies.outside C)
Instances For
Ω and Ω' are the two regions of the complement of C, in one order or the other.
Equations
- Schoenflies.IsRegionPair C Ω Ω' = (Ω = Schoenflies.inside C ∧ Ω' = Schoenflies.outside C ∨ Ω = Schoenflies.outside C ∧ Ω' = Schoenflies.inside C)
Instances For
Every component of the complement is one of the two regions: there are no others. This is the last clause of "exactly two regions".
A region is a connected component of the complement.
Every point of the curve is a limit of points of the region: this is the half of
frontier Ω = C that absorption uses.
The closure of a region is the region together with the curve.
A connected set off the curve and off one region lies in the other. This is the only place where "exactly two regions" is used, and it is used in exactly that form.
A nonempty connected set disjoint from the curve lies in one of the two regions. This is
the well-definedness clause of Lemma 2.6: "being connected it lies in one region of
ℝ² ∖ Jᵢ, so Wᵢ and Vᵢ are well defined".
Lemma 2.5: absorption #
Lemma 2.5 (absorption), set-theoretic core. If every point of C is a limit of points of
Ω, and Ω sits inside an open V whose frontier lies in J, then C misses V only where
J does.
Lemma 2.5 (absorption). If a region Ω of the complement of a separating curve C is
contained in a region V of the complement of a separating curve J, then all of C off J
lies in V.
Lemma 2.6: the two cells of a crosscut #
The setting: C is a separating curve, Ω and Ω' its two regions, J a separating curve
with P ⊆ J ⊆ C ∪ P, and W, V the two regions of ℝ² ∖ J, with Ω' ⊆ W. The blueprint's
J is Aᵢ ∪ P for one of the two arcs Aᵢ; the proof only needs the two inclusions.
The far region Ω' absorbs all of C off J, so the near cell V misses C entirely:
the part of C off J is in W, and the part on J is off V because V misses J.
Lemma 2.6, first half of (a): the cell V is contained in Ω ∖ P. It misses C, it is
connected, and it misses Ω' because Ω' ⊆ W; so it lies in Ω. It misses P because it
misses J ⊇ P.
Lemma 2.6 (a): the cell V is a connected component of Ω ∖ P. Its frontier is J, which
lies in C ∪ P and so misses the open set Ω ∖ P; Lemma 1.7 does the rest.
Lemma 2.6 (c): the closure of the cell meets C exactly in J ∩ C. The closure is
V ∪ J, and V misses C.
The arc-level side conditions #
Two small facts turn the blueprint's arc picture into the hypotheses the lemmas above take.
The two curves of Lemma 2.6 are Jordan curves: an arc from p to q and a second arc from
p to q meeting it only at those two points glue to a Jordan curve.
Lemma 2.6, bundled #
Lemma 2.6 (crosscut cells). Let C be a separating curve with regions Ω and Ω',
let A₁, A₂ ⊆ C be the two arcs cut out by a crosscut P which meets C only in points of
both arcs, and suppose the two curves Jᵢ = Aᵢ ∪ P are separating. Let Wᵢ be the region of
ℝ² ∖ Jᵢ containing Ω' and Vᵢ the other one. Then
- (a) each
Vᵢis a connected component ofΩ ∖ P; - (b)
V₁ ≠ V₂; - (c)
closure Vᵢ ∩ C = Aᵢ.
The blueprint proves (b) from the boundaries; here it falls straight out of (c), since
V₁ = V₂ would make A₁ = A₂.