Documentation

Atlas.Buildings.code.Reflection.UniqueAdjacentChamber

Hyperplane carriers are closed: the carrier $\{x : \langle n, x\rangle = d\}$ of an affine hyperplane is the preimage of a singleton under a continuous map, hence closed.

Hyperplane carriers have empty interior: if $η.carrier$ contained an open ball around some $x$, then moving $x$ slightly in the direction $η.normal$ would change the inner product $\langle η.normal, \cdot\rangle$ but stay inside the carrier — a contradiction.

theorem finite_iUnion_empty_interior' {X : Type u_2} [TopologicalSpace X] (S : Finset (Set X)) (hclosed : sS, IsClosed s) (hint : sS, interior s = ) :
interior (⋃ sS, s) =

Baire-style fact: a finite union of closed nowhere-dense sets has empty interior. Proof by induction: peel off one closed set new; if some open subset $U$ of the union meets the complement of new, $U \setminus \text{new}$ is open inside the remaining union and lies in its interior (which is empty by IH); otherwise $U \subseteq \text{interior}\,\text{new} = \emptyset$.

theorem interior_hyperplane_eq_empty {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] (n : E) (d : ) (hn : n 0) :
interior {x : E | inner n x = d} =

Affine hyperplanes have empty interior (raw-data form): the affine hyperplane $\{x : \langle n, x\rangle = d\}$ for nonzero $n$ has empty interior. Equivalently its complement is dense; the proof shows any open set $U$ meets the complement by perturbing in the direction $n$.

theorem dense_compl_hyperplane {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] (n : E) (d : ) (hn : n 0) :
Dense {x : E | inner n x = d}

The complement of an affine hyperplane is dense: equivalent reformulation of interior_hyperplane_eq_empty via the standard duality $\text{interior}\,A = \emptyset \iff A^c$ dense.

theorem connected_subset_halfspace {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] {C : Set E} {n : E} {d : } (hC : IsConnected C) (hC_disj : Disjoint C {x : E | inner n x = d}) :
C {x : E | inner n x > d} C {x : E | inner n x < d}

A connected set disjoint from a hyperplane lies in one open half-space: a connected $C$ avoiding $\{\langle n, x\rangle = d\}$ is contained in $\{> d\}$ or $\{< d\}$. The two half-spaces are disjoint open sets covering $C$; preconnectedness forces $C$ into one.

theorem chamber_of_mem_complement {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] (arr : HyperplaneArrangement E) (p : E) (hp : p arr.complement) :
∃ (C : arr.Chamber), p C.set

Every point of the complement lies in some chamber: given $p \in arr.complement$ (i.e. on no hyperplane), there exists a chamber $C$ containing $p$. The proof uses Zorn's lemma on the family of connected subsets of the complement containing $p$, with the union of a chain serving as an upper bound via IsPreconnected.sUnion_directed.

theorem avoidance_list {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] (hyps : List (E × )) (V : Set E) :
IsOpen VV.Nonempty(∀ phyps, p.1 0)xV, phyps, inner p.1 x p.2

Avoidance lemma for a finite list of affine hyperplanes: given an open nonempty $V$ and a list of affine hyperplane data $(n_i, d_i)$ with $n_i \ne 0$, there exists $x \in V$ avoiding every hyperplane $\langle n_i, x\rangle = d_i$. Proof by induction on the list: at each step, intersect $V$ with the dense open complement of the next hyperplane and recurse.

theorem mem_closure_of_open_and_opposite_side {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] (η : AffineHyperplane E) (D : Set E) (hD_open : IsOpen D) (p z : E) (hp_mem : p D) (hz : inner η.normal z = η.offset) (hp_side : inner η.normal p η.offset) :

Boundary point on a hyperplane belongs to the closure: if $D$ is open, $p \in D$ does not lie on the hyperplane $\eta$, and $z$ does lie on $\eta$, then $z \in \overline{D}$. The idea is that approaching $z$ along the segment from $p$ stays in the open half-space of $p$ and eventually inside $D$. (Currently this lemma is left as sorry.)

theorem wall_of_adjacent_chamber {E : Type u_2} [NormedAddCommGroup E] [InnerProductSpace E] (arr : HyperplaneArrangement E) (hlf : arr.IsLocallyFinite) (C D : arr.Chamber) (η : AffineHyperplane E) ( : η arr.hyperplanes) (hw : η.IsWall C.set) (hD_ne : D.set C.set) (hD_other_side : pD.set, qη.carrier, (∀ xC.set, inner η.normal x > η.offsetinner η.normal p < η.offset) xC.set, inner η.normal x < η.offsetinner η.normal p > η.offset) :
η.IsWall D.set

A wall of one chamber is also a wall of any neighboring chamber on the opposite side: if $η$ is a wall of $C$ and $D$ contains a point $p$ on the opposite side of $η$ from $C$, then $η$ is also a wall of $D$. The relatively open piece of $η$ extracted from the wall hypothesis on $C$ is shown to lie in $\overline{D}$ via mem_closure_of_open_and_opposite_side.

theorem adjacent_chamber_exists {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] (arr : HyperplaneArrangement E) (hlf : arr.IsLocallyFinite) (C : arr.Chamber) (η : AffineHyperplane E) ( : η arr.hyperplanes) (hw : η.IsWall C.set) :
∃ (D : arr.Chamber), D.set C.set η.IsWall D.set

Existence of an adjacent chamber across a wall: in a locally finite arrangement, for any wall $η$ of a chamber $C$, there exists a chamber $D \ne C$ for which $η$ is also a wall. The construction picks a point $q$ on $η$ in the relatively open piece, uses local finiteness to list nearby hyperplanes, then applies avoidance_list to find a point $p'$ in the opposite half-space avoiding all those hyperplanes. The chamber containing $p'$ (from chamber_of_mem_complement) is the required $D$.

theorem chambers_same_halfspaces_eq {E : Type u_2} [NormedAddCommGroup E] [InnerProductSpace E] (arr : HyperplaneArrangement E) (hlf : arr.IsLocallyFinite) (C D₁ D₂ : arr.Chamber) (η : AffineHyperplane E) ( : η arr.hyperplanes) (hw : η.IsWall C.set) (hD₁_ne : D₁.set C.set) (hD₁_wall : η.IsWall D₁.set) (hD₂_ne : D₂.set C.set) (hD₂_wall : η.IsWall D₂.set) :
D₁.set = D₂.set

Uniqueness of the chamber adjacent across a wall: in a locally finite arrangement, if two chambers $D_1, D_2$ both share the wall $η$ with $C$ and both differ from $C$, then they must be equal. Combined with adjacent_chamber_exists this yields existence and uniqueness of "the" adjacent chamber. (Currently this theorem is left as sorry.)