Chapter 05 — Small Cancellation Theory

Proofs omitted. Hover over highlighted terms to see their Lean types.

5.1 — Diagrams

Definition 5.1.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-1-1: the chapter owner for the underlying oriented map is TwoComplex.

open Set local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) #check TwoComplex

Definition 5.1.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-1-2: a path in a 1-complex from a to b is the canonical quiver path Quiver.Path a b, with empty path Quiver.Path.nil.

open Quiver.Path universe u v -- Layer triage: -- `source-facing`: Definition 5-1-2 introduces paths, closed paths, reduced paths, and simple -- paths in a fixed `1`-complex. -- `core/canonical`: `OneComplex` from Definition 3-2-1 is the owner abstraction for oriented -- edges, while `Quiver.Path`, `Quiver.Path.Loop`, `Quiver.Path.IsReduced`, and -- `Quiver.Path.IsSimple` are the canonical owner declarations for the path notions in this item. -- `bridge/view`: `OneComplex.Path.edges` recovers the textbook sequence of oriented edges from the -- canonical total-edge list `Quiver.Path.edgeList`. -- Domain sampling: -- 1. `Quiver.Path` is mathlib's canonical endpoint-aware finite path type and already includes the -- empty path. -- 2. `Quiver.Path.Loop` is the established owner for closed paths with a remembered basepoint. -- 3. `Quiver.Path.IsReduced` and `Quiver.Path.IsSimple` already define the reduced/simple path -- predicates upstream in Chapter 3. -- 4. `Quiver.Path.edgeList` is the canonical ordered total-edge list, from which the textbook -- oriented-edge sequence is derived. -- Primitive vs. derived: -- this item contributes no new owner object beyond the existing path API on `Quiver.Path`; the -- only additional source-facing bridge needed here is the oriented-edge list of a path. namespace OneComplex #check Quiver.Path

Definition 5.1.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-1-3: a diagram over a group F is an oriented map together with a label on each oriented edge such that the label of the oppositely oriented edge is the inverse of the original label.

universe u v variable {F : Type u} [Group F] structure GroupDiagram (F : Type u) [Group F] where /-- The underlying oriented map. -/ source : TwoComplex.{v} /-- The label assigned to each oriented edge of the underlying map. -/ label : source.skeleton.Edge → F /-- Reversing the orientation of an edge inverts its label. -/ label_inv (e : source.skeleton.Edge) : label e⁻¹ = (label e)⁻¹

Definition 5.1.4

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-1-4: the labels of a region D are the group elements obtained by applying pathLabel to loops representing the boundary cycle ∂D.

universe u v open Quiver.Path open OneComplex open scoped TwoComplex variable {F : Type u} [Group F] namespace GroupDiagram def regionLabels (M : GroupDiagram F) (D : M.source.Face) : Set F := { g | ∃ p : Loop M.source.skeleton, cyclicPath p = (∂ D) ∧ M.pathLabel p.2 = g }

Definition 5.1.5

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-1-5: a diagram for the finite sequence (c₁, ..., cₙ) in the free group with chosen basis basis is a group-labelled singular disc whose oriented edges have nontrivial labels, whose chosen outer boundary loop reads the reduced product c₁ ⋯ cₙ, and whose geometric regions admit an orientation whose boundary word is cyclically reduced and whose label product is conjugate to one of the listed relators.

universe u v w open CategoryTheory Quiver.Path OneComplex OneComplex.Hom variable {X : Type u} {F : Type v} [Group F] structure FreeGroupDiagram (basis : FreeGroupBasis X F) (relators : List F) extends GroupDiagram F where /-- A chosen outer boundary loop of the diagram, viewed in the full subcomplex of the underlying map. -/ outerBoundary : let S := source.fullSubcomplex Loop S.skeleton.toOneComplex /-- The whole labelled map is a genuine singular disc with boundary given by the chosen outer boundary loop. -/ singularDisc : let S := source.fullSubcomplex S.IsSingularDisc (cyclicPath outerBoundary) /-- Every oriented edge has nontrivial label. -/ label_ne_one (e : source.skeleton.Edge) : label e ≠ 1 /-- The outer boundary word is reduced without cancellation. -/ outerBoundary_reduced : let S := source.fullSubcomplex let p := mapLoop S.skeleton.inclusion outerBoundary FreeGroup.IsReduced (toGroupDiagram.pathLabelWord basis p.2) /-- The outer boundary label product is exactly `c₁ ⋯ cₙ`. -/ outerBoundary_product : let S := source.fullSubcomplex let p := mapLoop S.skeleton.inclusion outerBoundary toGroupDiagram.pathLabel p.2 = relators.prod /-- Every geometric region admits an oriented representative together with a based boundary path whose boundary word is cyclically reduced and whose label product is conjugate to one of the listed relators. -/ regionBoundary_condition (D : TwoComplex.GeometricFace source) : ∃ E : source.Face, ⟦E⟧ = D ∧ ∃ v : source.skeleton, ∃ q : source.BoundaryPath E v, FreeGroup.IsCyclicallyReduced (toGroupDiagram.pathLabelWord basis q.1) ∧ ∃ i : Fin relators.length, IsConj (toGroupDiagram.pathLabel q.1) (relators.get i)

Theorem 5.1.6

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-1-6: every finite ordered sequence of nontrivial elements of a free group with chosen basis basis admits a chapter-5 free-group diagram whose geometric faces are in ordered bijective correspondence with the full relator sequence.

universe u v variable {X : Type u} {F : Type v} [Group F] namespace FreeGroupDiagram -- Proof sketch: write each nontrivial relator `cᵢ` as a conjugate of a cyclically reduced word, -- build the corresponding one-face lollipop diagram for each factor, and attach those diagrams in -- the given order around a common basepoint. This gives one geometric face for each index `i`. -- Whenever the outer boundary word has adjacent inverse letters, fold or delete the cancellable -- boundary pair; this preserves the cyclically reduced face labels and the ordered face data while -- strictly shortening the boundary. Iterating terminates with a reduced outer boundary word. theorem exists_orderedFaceRealization (basis : FreeGroupBasis X F) (relators : List F) (hrelators : ∀ i : Fin relators.length, relators.get i ≠ 1) : ∃ D : FreeGroupDiagram basis relators, ∃ e : Fin relators.length ≃ D.source.GeometricFace, ∀ i : Fin relators.length, ∃ E : D.source.Face, ⟦E⟧ = e i ∧ ∃ v : D.source.skeleton, ∃ q : D.source.BoundaryPath E v, FreeGroup.IsCyclicallyReduced (D.toGroupDiagram.pathLabelWord basis q.1) ∧ IsConj (D.toGroupDiagram.pathLabel q.1) (relators.get i)

Lemma 5.1.7

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-1-7: if S is a singular disc in a group-labelled 2-complex M, then the label of its chosen outer boundary loop is an ordered product of conjugates of labels chosen from the regions of S.

universe u v open Quiver.Path open Group open OneComplex.Hom variable {F : Type u} [Group F] theorem boundaryCycleWord_eq_list_prod_conjugates_of_regionLabels (M : GroupDiagram F) (S : M.source.Subcomplex) (p : Loop S.skeleton.toOneComplex) (hS : S.IsSingularDisc (cyclicPath p)) : ∃ cs : List F, ∃ e : Fin cs.length ≃ S.complex.GeometricFace, (∀ i, ∃ D : S.complex.Face, ⟦D⟧ = e i ∧ cs[i] ∈ conjugatesOfSet (M.regionLabels D.1)) ∧ let q := mapLoop S.skeleton.inclusion p M.pathLabel q.2 = cs.prod

Definition 5.1.8

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-1-8: for a symmetrized subset R of F, an R-diagram is a group diagram whose label on every boundary cycle of every region belongs to R.

universe u open Quiver.Path variable {F : Type u} [Group F] namespace GroupDiagram def IsRDiagram (M : GroupDiagram F) (R : Set F) : Prop := ∀ D : M.source.Face, M.regionLabels D ⊆ R

5.2 — The Small Cancellation Hypotheses

Definition 5.2.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-2-1: the relator set R satisfies the small-cancellation condition C'(λ) with respect to basis if every nonempty piece piece occurring in a symmetrized relator from R* has length strictly less than λ times the length of that symmetrized relator.

universe u v open GroupPresentation variable {X : Type u} {F : Type v} [Group F] namespace FreeGroupBasis def condition_c_prime (basis : FreeGroupBasis X F) (R : Set F) (lambda : ℝ) : Prop := ∀ {q : CyclicWord X} (_ : q ∈ symmetrizedRelatorFamily (basis.repr '' R)) {piece : List (X × Bool)} (_ : q.HasPart piece) (_ : basis.is_piece R piece), (piece.length : ℝ) < lambda * q.length

Definition 5.2.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-2-2: the relator set R satisfies the small-cancellation condition C(p) with respect to basis if no symmetrized relator from R* is the concatenation of fewer than p piece-words.

universe u v open GroupPresentation variable {X : Type u} {F : Type v} [Group F] namespace FreeGroupBasis def condition_c (basis : FreeGroupBasis X F) (R : Set F) (p : ℕ) : Prop := ∀ {q : CyclicWord X} (_ : q ∈ symmetrizedRelatorFamily (basis.repr '' R)) {pieces : List (List (X × Bool))}, (∀ piece ∈ pieces, basis.is_piece R piece) → pieces.length < p → q.1 ≠ (pieces.flatten : Cycle (X × Bool))

Definition 5.2.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-2-3: the relator set R satisfies T(q) with respect to basis if every cyclic relator cycle from the symmetrized relator set generated by R of length h with 3 ≤ h < q and no successive inverse pair has at least one adjacent cyclic product whose reduced basis words concatenate without cancellation.

universe u v variable {X : Type u} {F : Type v} [Group F] open GroupPresentation namespace FreeGroupBasis def condition_t (basis : FreeGroupBasis X F) (R : Set F) (q : ℕ) : Prop := ∀ rs : Cycle F, 3 ≤ rs.length → rs.length < q → (∀ r ∈ rs, r ∈ basis.symmetrizedRelatorSet R) → rs.Chain (fun r s ↦ r * s ≠ 1) → basis.hasReducedCyclicProduct rs

Definition 5.2.4

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-2-4: a sequence of conjugates of elements of R is a minimal R-sequence when its product cannot also be expressed as the product of a shorter sequence of conjugates of elements of R.

universe u open Group Subgroup variable {F : Type u} [Group F] def IsMinimalRelatorSequence (R : Set F) (cs : List F) : Prop := IsRelatorConjugateSequence R cs ∧ ∀ ds : List F, ds.length < cs.length → IsRelatorConjugateSequence R ds → ds.prod ≠ cs.prod -- Proof sketch: each factor belongs to `conjugatesOfSet R`, hence to `normalClosure R`; the -- subgroup is closed under multiplication, so the list product also lies in `normalClosure R`.

Definition 5.2.5

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-2-5: a diagram over F is reduced when, whenever two face boundary cycles can be written as e δ₁ and δ₂ e⁻¹ for a common oriented edge e, the complementary labels φ(δ₁) and φ(δ₂) are not inverse to one another.

universe u open Quiver.Path variable {F : Type u} [Group F] namespace GroupDiagram def IsReduced (M : GroupDiagram F) : Prop := ∀ {a b : M.source.skeleton} (D₁ D₂ : M.source.Face) (e : a ⟶ b) (q₁ : M.source.BoundaryPath D₁ a) (q₂ : M.source.BoundaryPath D₂ a) (δ₁ : Quiver.Path b a) (δ₂ : Quiver.Path a b), q₁.1 = e.toPath.comp δ₁ → q₂.1 = δ₂.comp (Quiver.reverse e).toPath → M.pathLabel δ₂ ≠ (M.pathLabel δ₁)⁻¹

Lemma 5.2.6

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-2-6: if M is a diagram for a minimal R-sequence cs, with exactly one geometric region for each term of cs, then the underlying group diagram of M is reduced.

universe u v variable {X : Type u} {F : Type v} [Group F] theorem freeGroupDiagram_isReduced_of_isMinimalRelatorSequence (basis : FreeGroupBasis X F) {R : Set F} {cs : List F} (hcs : IsMinimalRelatorSequence R cs) (M : FreeGroupDiagram basis cs) (hfaces : Nat.card (TwoComplex.GeometricFace M.source) = cs.length) : M.IsReduced

Definition 5.2.7

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-2-7 (1): a boundary vertex of a planar map is a vertex whose image lies on the boundary of the map.

open Set local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding open OneComplex variable {C : TwoComplex} def IsBoundaryVertex (embedding : TwoManifoldEmbedding C 𝔼²) (v : C.skeleton) : Prop := embedding.vertexMap v ∈ embedding.boundary

Definition 5.2.8

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-2-8 (1): the degree of a vertex is the number of oriented edges having that vertex as initial vertex, so a loop based at the vertex contributes twice through its two orientations.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace OneComplex def vertexDegree (C : OneComplex) [Finite C.Edge] (v : C) : ℕ := Nat.card { e : C.Edge // C.initial e = v }

Lemma 5.2.9

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-2-9 (1): if R satisfies C(k), then every region of a reduced R-diagram whose boundary contains no boundary edge of the ambient planar map has degree at least k.

universe u v open FreeGroupBasis OneComplex local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) variable {X : Type u} {F : Type v} [Group F] variable (basis : FreeGroupBasis X F) {R : Set F} {M : GroupDiagram F} variable (embedding : TwoComplex.TwoManifoldEmbedding M.source 𝔼²) variable [embedding.IsPlanarMap] -- Proof sketch: every edge in the boundary of `D` is interior, so reducedness forces the label on -- each oriented boundary edge to be a piece. The boundary label of `D` is therefore a product of -- `M.source.regionDegree D` pieces, and `C(k)` forbids such a factorization with fewer than `k` -- pieces. theorem regionDegree_ge_of_condition_c (hRDiagram : M.IsRDiagram R) (hReduced : M.IsReduced) {k : ℕ} (hck : C(k)[basis, R]) {D : TwoComplex.GeometricFace M.source} (hD : M.source.boundaryGeometricEdges D ⊆ { e : OneComplex.GeometricEdge M.source.skeleton | embedding.IsInteriorEdge e }) : k ≤ M.source.regionDegree D

5.3 — The Basic Formulas

Definition 5.3.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-3-1 (1): the positive integral solutions of 1 / p + 1 / q = 1 / 2 are exactly (3, 6), (4, 4), and (6, 3).

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) theorem positive_integral_reciprocal_sum_eq_half_iff {p q : ℕ} (hp : 0 < p) (hq : 0 < q) : (1 : ℚ) / p + 1 / q = 1 / 2 ↔ (p, q) = (3, 6) ∨ (p, q) = (4, 4) ∨ (p, q) = (6, 3)

Theorem 5.3.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-3-2: for a finite planar map, the two basic curvature formulas are the identities obtained by expanding the canonical Euler term V - E + F for positive integers p and q satisfying 1 / p + 1 / q = 1 / 2.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding open OneComplex variable {C : TwoComplex} variable (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] theorem basic_formulas_of_planar_map (p q : ℕ) (hp : 0 < p) (hq : 0 < q) (hReciprocal : (1 : ℚ) / p + 1 / q = (1 : ℚ) / 2) : (p : ℚ) * (embedding.vertexCount - embedding.edgeCount + embedding.regionCount) = embedding.boundaryVertexDefectSum p + embedding.interiorVertexDefectSum p + ((p : ℚ) / q) * embedding.regionDegreeDefectSum q - ((p : ℚ) / q) * embedding.boundaryEdgeCount ∧ (p : ℚ) * (embedding.vertexCount - embedding.edgeCount + embedding.regionCount) = embedding.boundaryVertexAdjustedDefectSum p q + embedding.interiorVertexDefectSum p + ((p : ℚ) / q) * embedding.regionDegreeDefectSum q + ((p : ℚ) / q) * (embedding.boundaryVertexCount - embedding.boundaryEdgeCount)

Lemma 5.3.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-3-3: if every interior vertex of a planar map has degree at least 2, then the number of interior vertices is at most the number of interior edges.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding open OneComplex variable {C : TwoComplex} variable (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] -- Proof sketch: sum the degrees of the interior vertices. The hypothesis gives a lower bound `2` -- on each summand, while each interior geometric edge contributes at most two incident interior -- endpoints, yielding twice the interior-vertex count `≤` twice the interior-edge count and hence -- the desired inequality. theorem interiorVertexCount_le_interiorEdgeCount_of_interiorVertexDegree_ge_two (hdegree : embedding.HasInteriorVertexDegreeAtLeast 2) : Nat.card { v : C.skeleton // embedding.IsInteriorVertex v } ≤ Nat.card { e : GeometricEdge C.skeleton // embedding.IsInteriorEdge e }

Corollary 5.3.4

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Corollary 5-3-4: the curvature sum ∑_M^• [p / q + 2 - d(v)] of a connected simply connected [p, q] map with more than one vertex is at least p.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} variable (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] -- Proof sketch: apply the basic formula `(3.2)` to the connected simply connected `[p, q]` map, -- so the Euler term is `1`; then use the `[p, q]` lower bounds to show the interior-vertex and -- region-defect contributions are nonpositive, and use the “more than one vertex” hypothesis -- together with the deletion-of-isolated-vertices argument of the text to control the remaining -- boundary term. theorem curvature_formula_lower_bound_of_simplyConnected_squareBracketPQMap (hconnected : Quiver.IsStronglyConnected (Quiver.Symmetrify C.skeleton)) [IsSimplyConnected C] (p q : ℕ) (hp : 0 < p) (hq : 0 < q) (hPQ : embedding Is[p, q]) (hreciprocal : (1 : ℚ) / p + 1 / q = 1 / 2) (hmoreThanOneVertex : (1 : ℚ) < embedding.vertexCount) : (p : ℚ) ≤ embedding.boundaryVertexAdjustedDefectSum p q

Corollary 5.3.5

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Corollary 5-3-5: the curvature sum ∑_M^• [p / q + 2 - i(D)] over the boundary regions of a connected simply connected (q, p) map with more than one region is at least p.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} variable (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] theorem curvature_formula_lower_bound_of_simplyConnected_roundBracketQPMap (hconnected : Quiver.IsStronglyConnected (Quiver.Symmetrify C.skeleton)) [IsSimplyConnected C] (p q : ℕ) (hp : 0 < p) (hq : 0 < q) (hQP : embedding Is(q, p)) (hreciprocal : (1 : ℚ) / p + 1 / q = 1 / 2) (hmoreThanOneRegion : (1 : ℚ) < embedding.regionCount) : (p : ℚ) ≤ σ'(embedding)[p, q]

5.4 — Dehn’s Algorithm and Greendlinger’s Lemma

Lemma 5.4.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-4-1: in a simply connected (q, p) map with (q, p) equal to (3, 6), (4, 4), or (6, 3), if every region whose boundary contains no boundary edge of the ambient map has degree at least p, then every oriented face, equivalently every region, has simple boundary.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} -- Proof sketch: argue by contradiction using a minimal submap cut off by a non-simple boundary -- loop in the boundary of `D`. That submap inherits simple connectedness, all of its vertices -- except possibly one boundary vertex have degree at least `q`, and every region in it has degree -- at least `p` by the no-boundary-edge hypothesis. Corollary `3.3` then gives a curvature -- inequality contradicting `(q, p) ∈ {(3, 6), (4, 4), (6, 3)}`. theorem face_hasSimpleBoundary_of_simplyConnected_roundBracketQPMap (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] [TwoComplex.IsSimplyConnected C] (q p : ℕ) (hQP : embedding Is(q, p)) (hPair : (q, p) = (3, 6) ∨ (q, p) = (4, 4) ∨ (q, p) = (6, 3)) (hRegionDegree : ∀ D : GeometricFace C, C.boundaryGeometricEdges D ⊆ {e | embedding.IsInteriorEdge e} → p ≤ C.regionDegree D) (D : C.Face) : C.HasSimpleBoundary D

Definition 5.4.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-4-2: an extremal disk of a planar map embedding is a submap K that is a topological disk and has a boundary cycle whose edges occur in order in some boundary cycle of the whole map.

open Quiver.Path local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} def IsExtremalDisk (embedding : TwoManifoldEmbedding C 𝔼²) (K : Subcomplex C) : Prop := ∃ boundaryK : CyclicPath K.skeleton.toOneComplex, ∃ boundaryM : CyclicPath C.skeleton, K.IsSingularDisc boundaryK ∧ embedding.IsBoundaryCycle boundaryM ∧ boundaryCycleOccursInOrder K boundaryK boundaryM

Lemma 5.4.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-4-3: if a connected simply connected planar map has no vertices of degree 1 and its boundary is not a simple closed path, then the map contains at least two distinct extremal disks.

open Quiver.Path local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} -- Proof sketch: induct on the number of regions. Choose a shortest closed subpath of a boundary -- cycle of `M`; the degree-one hypothesis makes this subpath simple, so it bounds an extremal -- singular-disc submap. Delete the vertices of degree one from the complementary submap and -- apply the induction hypothesis there; either the complement is itself extremal or it contains -- another extremal disk that remains extremal in the ambient map, yielding an injective -- `Fin 2`-indexed family of extremal disks. theorem exists_two_extremalDisks_of_not_boundary_simpleClosedPath (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] (hconnected : Quiver.IsStronglyConnected (Quiver.Symmetrify C.skeleton)) [TwoComplex.IsSimplyConnected C] (hnoDegreeOne : let _ : Finite C.skeleton.Edge := finite_orientedEdge embedding ∀ v : C.skeleton, C.skeleton.vertexDegree v ≠ 1) (hboundary : ¬ TwoManifoldEmbedding.HasSimpleBoundary embedding) : ∃ disks : Fin 2 → Subcomplex C, Function.Injective disks ∧ ∀ i : Fin 2, embedding.IsExtremalDisk (disks i)

Definition 5.4.4

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-4-4: the intersection ∂D ∩ ∂M is a consecutive part of the planar map when it is the union of a nonempty consecutive sequence of closed edges that occurs along a boundary cycle of the region D and along some boundary cycle of the whole map.

open Set Quiver.Path local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} def BoundaryIntersectionIsConsecutivePart (embedding : TwoManifoldEmbedding C 𝔼²) (D : GeometricFace C) : Prop := ∃ edges : List (OneComplex.GeometricEdge C.skeleton), edges ≠ [] ∧ frontier (embedding.region D) ∩ embedding.boundary = edgeSequenceSupport embedding edges ∧ GeometricFaceHasConsecutiveEdgeSequence C D edges ∧ BoundaryCycleHasConsecutiveEdgeSequence embedding edges

Theorem 5.4.5

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-4-5: in a connected simply connected (q, p) map with no vertices of degree 1 and more than one region, if every region whose boundary contains no boundary edge of the ambient map has degree at least p, then the starred curvature sum ∑_M^* [p / q + 2 - i(D)] is at least p.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} variable (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] -- Proof sketch: first handle the case where the map boundary is a simple closed path by -- induction on the number of regions. If every boundary region has consecutive boundary -- intersection, apply Corollary `5-3-5`; otherwise choose a region whose boundary intersection is -- not consecutive, split the map along that region into two smaller maps, apply the induction -- hypothesis to both pieces, and combine the resulting starred sums. When the boundary is not a -- simple closed path, use Lemma `5-4-3` to obtain two extremal disks and add their contributions -- to recover the same lower bound for the ambient starred sum. theorem consecutiveBoundaryRegion_curvature_formula_lower_bound_of_simplyConnected_roundBracketQPMap (hconnected : Quiver.IsStronglyConnected (Quiver.Symmetrify C.skeleton)) [TwoComplex.IsSimplyConnected C] (p q : ℕ) (hp : 0 < p) (hq : 0 < q) (hQP : embedding Is(q, p)) (hreciprocal : (1 : ℚ) / p + 1 / q = 1 / 2) (hnoDegreeOne : let _ : Finite C.skeleton.Edge := finite_orientedEdge embedding ∀ v : C.skeleton, C.skeleton.vertexDegree v ≠ 1) (hmoreThanOneRegion : (1 : ℚ) < embedding.regionCount) (hRegionDegree : ∀ D : GeometricFace C, C.boundaryGeometricEdges D ⊆ { e : OneComplex.GeometricEdge C.skeleton | embedding.IsInteriorEdge e } → p ≤ C.regionDegree D) : (p : ℚ) ≤ embedding.adjustedInteriorEdgeDefectSum embedding.BoundaryIntersectionIsConsecutivePart p q

Definition 5.4.6

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-4-6: a word s is a j-remnant with respect to R and basis if some relator r ∈ R is represented by the concatenated word s b₁ ⋯ bⱼ, where b₁, ..., bⱼ are pieces with respect to R.

universe u v variable {X : Type u} {F : Type v} [Group F] namespace FreeGroupBasis def is_j_remnant (basis : FreeGroupBasis X F) (R : Set F) (j : ℕ) (s : List (X × Bool)) : Prop := ∃ r ∈ R, ∃ pieces : Vector (List (X × Bool)) j, (∀ piece ∈ pieces.toList, basis.is_piece R piece) ∧ (basis.repr r).toWord = s ++ pieces.toList.flatten -- Proof sketch: unfold `is_j_remnant`; the theorem is exactly the existential witness data in the -- definition.

Theorem 5.4.7

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-4-7: if R satisfies C(p) and T(q) for one of the pairs (q, p) = (6, 3), (4, 4), or (3, 6), then every nontrivial element of the normal closure of R has a cyclically reduced conjugate w* whose cyclic word either lies in the symmetrized relator family R* or admits a decomposition u₁ s₁ ⋯ uₙ sₙ with each sₖ an i(sₖ)-remnant and ∑ [p / q + 2 - i(sₖ)] ≥ p; this second alternative is recorded through the canonical bridge predicate w*.HasGreendlingerRemnantConfiguration R p q rather than by repeating its witness data inline.

universe u variable {X : Type u} open FreeGroupBasis GroupPresentation theorem greendlinger_remnant_alternative_for_small_cancellation_pairs (R : Set (FreeGroup X)) {p q : ℕ} (hpair : (q, p) = (6, 3) ∨ (q, p) = (4, 4) ∨ (q, p) = (3, 6)) (hC : C(p)[basis, R]) (hT : T(q)[basis, R]) {w : FreeGroup X} (hw_ne : w ≠ 1) (hw_mem : w ∈ Subgroup.normalClosure R) : ∃ wStar : FreeGroup X, IsConj wStar w ∧ ∃ hwStar_cyclic : FreeGroup.IsCyclicallyReduced wStar.toWord, ((⟨wStar.toWord, hwStar_cyclic⟩ : CyclicWord X) ∈ symmetrizedRelatorFamily R ∨ wStar.HasGreendlingerRemnantConfiguration R p q)

Theorem 5.4.8

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-4-8: if R satisfies C'(1 / 6) and w is a nontrivial cyclically reduced word in the normal closure of R, then either the cyclic word of w lies in the symmetrized relator family R*, or some cyclically reduced conjugate of w satisfies one of Greendlinger's five sixth-group long-subword configurations. The existential conclusion is stated directly in terms of the canonical owners IsConj and IsCyclicallyReduced, rather than through an extra witness wrapper.

universe u open GroupPresentation variable {X : Type u} local notation "basis" => FreeGroupBasis.ofFreeGroup X theorem greendlinger_alternatives_for_sixth_groups (R : Set (FreeGroup X)) {w : FreeGroup X} (hR : C'((1 / 6 : ℝ))[basis, R]) (hw_ne : w ≠ 1) (hw_cyclic : FreeGroup.IsCyclicallyReduced w.toWord) (hw_mem : w ∈ Subgroup.normalClosure R) : (⟨w.toWord, hw_cyclic⟩ : CyclicWord X) ∈ symmetrizedRelatorFamily R ∨ ∃ wStar : FreeGroup X, IsConj wStar w ∧ FreeGroup.IsCyclicallyReduced wStar.toWord ∧ wStar.HasGreendlingerSixthConfiguration R

Theorem 5.4.9

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-4-9: if R satisfies C'(1 / 4) and T(4), and w is a nontrivial cyclically reduced word in the normal closure of R, then either the cyclic word of w lies in the symmetrized relator family R*, or that same cyclic word satisfies one of Greendlinger's two quarter-group configurations. The public conclusion stays on the intrinsic cyclic-word owner rather than introducing a second representative-level conjugacy witness.

universe u open GroupPresentation variable {X : Type u} local notation "basis" => FreeGroupBasis.ofFreeGroup X theorem greendlinger_alternatives_for_quarter_groups (R : Set (FreeGroup X)) {w : FreeGroup X} (hR_cprime : C'((1 / 4 : ℝ))[basis, R]) (hR_t : T(4)[basis, R]) (hw_ne : w ≠ 1) (hw_cyclic : FreeGroup.IsCyclicallyReduced w.toWord) (hw_mem : w ∈ Subgroup.normalClosure R) : (⟨w.toWord, hw_cyclic⟩ : CyclicWord X) ∈ symmetrizedRelatorFamily R ∨ CyclicWord.HasGreendlingerQuarterConfiguration (⟨w.toWord, hw_cyclic⟩ : CyclicWord X) R

Definition 5.4.10

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-4-10: a word w on the generators X is R-reduced with respect to basis if it is freely reduced and has no subword that is longer than half of a relator from the symmetrized family generated by R.

universe u v open GroupPresentation variable {X : Type u} {F : Type v} [Group F] namespace FreeGroupBasis def is_r_reduced (basis : FreeGroupBasis X F) (R : Set F) (w : List (X × Bool)) : Prop := FreeGroup.IsReduced w ∧ ∀ ⦃part : List (X × Bool)⦄, List.IsInfix part w → ¬ HasLongSymmetrizedRelatorPart (basis.repr '' R) part

5.5 — The Conjugacy Problem

Lemma 5.5.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-5-1: if M is an annular R-diagram, u is the label of an outer boundary cycle of M, and z⁻¹ is the label of an inner boundary cycle of M, then u and z represent conjugate elements of the presented group PresentedGroup R.

universe u open Quiver.Path local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) local notation "𝕊¹" => Metric.sphere (0 : 𝔼²) 1 variable {X : Type u} namespace GroupDiagram theorem outerBoundaryLabel_isConj_of_innerBoundaryInverseLabel {R : Set (FreeGroup X)} (M : AnnularRDiagram R) (outer inner : Loop M.source.skeleton) (houter : cyclicPath outer = M.outerBoundary) (hinner : cyclicPath inner = M.innerBoundary) (u z : FreeGroup X) (hu : M.pathLabel outer.2 = u) (hz : M.pathLabel inner.2 = z⁻¹) : IsConj (PresentedGroup.mk R u) (PresentedGroup.mk R z)

Lemma 5.5.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-5-2: if u and z are cyclically reduced words of the free group F = FreeGroup X, u and z are not conjugate in F although their images in G = F / N are conjugate, then there exists a reduced annular R-diagram with at least one region together with chosen outer and inner boundary loops whose boundary labels realize the cyclic words of u and z⁻¹, respectively.

universe u open Quiver.Path variable {X : Type u} local notation "basis" => FreeGroupBasis.ofFreeGroup X namespace GroupDiagram theorem exists_reduced_annularRDiagram_of_conjugate_in_relator_quotient (R : Set (FreeGroup X)) {u z : FreeGroup X} (hu_cyclic : FreeGroup.IsCyclicallyReduced u.toWord) (hz_cyclic : FreeGroup.IsCyclicallyReduced z.toWord) (h_not_conj : ¬ IsConj u z) (hconj_quotient : IsConj (PresentedGroup.mk R u) (PresentedGroup.mk R z)) : ∃ (M : AnnularRDiagram R) (outer inner : Loop M.source.skeleton) (houter : cyclicPath outer = M.outerBoundary) (hinner : cyclicPath inner = M.innerBoundary), Nonempty M.source.Face ∧ M.IsReduced ∧ ∃ houter_cyclic : FreeGroup.IsCyclicallyReduced (M.pathLabelWord basis outer.2), ∃ hinner_cyclic : FreeGroup.IsCyclicallyReduced (M.pathLabelWord basis inner.2), (⟨(M.pathLabelWord basis outer.2 : Cycle (X × Bool)), houter_cyclic⟩ : CyclicWord X) = ⟨u.toWord, hu_cyclic⟩ ∧ (⟨(M.pathLabelWord basis inner.2 : Cycle (X × Bool)), hinner_cyclic⟩ : CyclicWord X) = ⟨(z⁻¹).toWord, by simpa [FreeGroup.toWord_inv] using isCyclicallyReduced_invRev hz_cyclic⟩

Theorem 5.5.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-5-3: for a reduced annular R-diagram whose outer and inner boundary cycles satisfy the source boundary-arc hypotheses and whose regions do not meet both boundary components, the two small-cancellation cases C'(1 / 6) with (q, p) = (3, 6) and C'(1 / 4) together with T(4) and (q, p) = (4, 4) force the annular structure conclusions that no region boundary has all its edges interior to the ambient map, every region has i(D) = p / q + 2, and every interior vertex has degree q.

open Set Quiver.Path FreeGroupBasis GroupPresentation universe u v local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) variable {X : Type u} {F : Type v} [Group F] namespace TwoComplex namespace TwoManifoldEmbedding variable {M : GroupDiagram F} variable (basis : FreeGroupBasis X F) (R : Set F) variable (embedding : TwoManifoldEmbedding M.source 𝔼²) [embedding.IsPlanarMap] -- Proof sketch: apply the boundary-arc exclusion hypotheses to rule out regions whose boundary -- misses the ambient boundary or meets one boundary component in more than one arc, then pass to -- the simply connected discs obtained by cutting the annular map along boundary regions and apply -- the curvature estimate from Theorem `5-4-5` in the two allowed small-cancellation cases. The -- resulting nonpositive defect sums force every region to meet the boundary, give the uniform -- value `i(D) = p / q + 2`, and force every interior vertex to have degree exactly `q`. theorem suitable_annular_r_diagram_has_uniform_region_and_vertex_structure (σ τ : CyclicPath M.source.skeleton) (p q : ℕ) (hcase : ((q, p) = (3, 6) ∧ C'((1 / 6 : ℝ))[basis, R]) ∨ ((q, p) = (4, 4) ∧ C'((1 / 4 : ℝ))[basis, R] ∧ T(4)[basis, R])) (hRDiagram : M.IsRDiagram R) (hReduced : M.IsReduced) (hannular : embedding.HasAnnularBoundaryCycles σ τ) (houter : ∀ D : GeometricFace M.source, ¬ embedding.BoundaryCycleIntersectionLongerThanHalfRelator basis R σ D) (hinner : ∀ D : GeometricFace M.source, ¬ embedding.BoundaryCycleIntersectionLongerThanHalfRelator basis R τ D) (hseparate : ∀ D : GeometricFace M.source, ¬ RegionMeetsBothBoundaryCycles σ τ D) : (∀ D : GeometricFace M.source, ¬ M.source.boundaryGeometricEdges D ⊆ { e : OneComplex.GeometricEdge M.source.skeleton | embedding.IsInteriorEdge e }) ∧ (∀ D : GeometricFace M.source, (embedding.boundaryInteriorEdgeCount D : ℚ) = (p : ℚ) / q + 2) ∧ (let S := M.source.skeleton let _ : Finite S.Edge := finite_orientedEdge embedding ∀ v : S, embedding.IsInteriorVertex v → S.vertexDegree v = q)

Theorem 5.5.4

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-5-4 (1): if R satisfies either C'(1 / 6) or C'(1 / 4) together with T(4), and u, z are nontrivial cyclically R-reduced free-group elements that are not conjugate in the free group itself, then their images in the quotient G = F / N are conjugate if and only if there exists a conjugator h such that the outer small-cancellation hypothesis determines the relator-length bound 2q max(|u|, |z|) with q = 3 in the C'(1 / 6) case and q = 4 otherwise, the quotient equality u = h⁻¹ z h holds, and h has the source short factorization by one or two relator subwords. Because u and z are already cyclically R-reduced, the criterion is stated directly for u and z rather than introducing extra witnesses u* and z*.

universe u open FreeGroupBasis GroupPresentation variable {X : Type u} local notation "basis" => FreeGroupBasis.ofFreeGroup X namespace FreeGroup -- Proof sketch: for the forward implication, start from a reduced annular `R`-diagram for the -- quotient conjugacy class and apply the annular small-cancellation analysis to a region meeting -- both boundary components. Reading the corresponding boundary arc yields a conjugator that is one -- or two relator subwords, with relator lengths bounded by `2q max(|u|, |z|)` and with the -- `C'(1 / 8)` improvement forcing the one-subword case. The reverse implication is immediate from -- the displayed quotient equality `u = h⁻¹ z h`. theorem quotient_conjugacy_iff_exists_short_smallCancellation_conjugator (R : Set (FreeGroup X)) {u z : FreeGroup X} (hcase : C'((1 / 6 : ℝ))[basis, R] ∨ (C'((1 / 4 : ℝ))[basis, R] ∧ T(4)[basis, R])) (hu_ne : u ≠ 1) (hz_ne : z ≠ 1) (hu_cyclic : (basis).is_cyclically_r_reduced R u.toWord) (hz_cyclic : (basis).is_cyclically_r_reduced R z.toWord) (hnotconj : ¬ IsConj u z) : IsConj (PresentedGroup.mk R u) (PresentedGroup.mk R z) ↔ ∃ h : FreeGroup X, let qBound := if C'((1 / 6 : ℝ))[basis, R] then 3 else 4; let bound := 2 * qBound * max u.toWord.length z.toWord.length; let hasShortRelatorPart : List (X × Bool) → Prop := fun part ↦ ∃ q ∈ symmetrizedRelatorFamily R, q.HasPart part ∧ q.length < bound; PresentedGroup.mk R u = PresentedGroup.mk R (h⁻¹ * z * h) ∧ (((∃ part₁ : List (X × Bool), h = FreeGroup.mk part₁ ∧ hasShortRelatorPart part₁) ∨ ∃ part₁ part₂ : List (X × Bool), h = FreeGroup.mk part₁ * FreeGroup.mk part₂ ∧ hasShortRelatorPart part₁ ∧ hasShortRelatorPart part₂) ∧ (C'((1 / 8 : ℝ))[basis, R] → ∃ part₁ : List (X × Bool), h = FreeGroup.mk part₁ ∧ hasShortRelatorPart part₁))

Theorem 5.5.5

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-5-5: if R satisfies either C'(1 / 6) or C'(1 / 4) together with T(4), and M is a reduced annular R-diagram whose outer and inner boundary cycles satisfy the source boundary-arc exclusion hypotheses, then the existence of one region meeting both boundary components forces every region to meet both boundary cycles.

universe u v open Quiver.Path FreeGroupBasis local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) variable {X : Type u} {F : Type v} [Group F] namespace TwoComplex namespace TwoManifoldEmbedding variable {M : GroupDiagram F} variable (basis : FreeGroupBasis X F) (R : Set F) (embedding : TwoManifoldEmbedding M.source 𝔼²) variable [embedding.IsPlanarMap] (σ τ : CyclicPath M.source.skeleton) variable (hcase : C'((1 / 6 : ℝ))[basis, R] ∨ (C'((1 / 4 : ℝ))[basis, R] ∧ T(4)[basis, R])) variable (hRDiagram : M.IsRDiagram R) (hReduced : M.IsReduced) variable (hannular : embedding.HasAnnularBoundaryCycles σ τ) variable (houter : ∀ D : GeometricFace M.source, ¬ embedding.BoundaryCycleIntersectionLongerThanHalfRelator basis R σ D) (hinner : ∀ D : GeometricFace M.source, ¬ embedding.BoundaryCycleIntersectionLongerThanHalfRelator basis R τ D) (hexists : ∃ E : GeometricFace M.source, RegionMeetsBothBoundaryCycles σ τ E) -- Proof sketch: start with a region meeting both boundary components, cut the annular diagram -- along a boundary-to-boundary segment of that region, and apply the curvature estimate from -- Theorem `4-3` to the resulting disc. Greendlinger's lemma rules out the exceptional case, and -- the remaining degree computation propagates outward through adjacent regions; if islands remain, -- remove one and conclude by induction on the number of regions. theorem all_regions_meet_both_boundary_cycles_of_exists_crossing_region (D : GeometricFace M.source) : RegionMeetsBothBoundaryCycles σ τ D

5.6 — The Word Problem

Lemma 5.6.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-6-1: if a [p, q] map admits a finite component decomposition in which every component is either simply connected or annular in the induced restricted embedding, then the boundary edge-count and boundary vertex-count satisfy formulas (6.1) and (6.2).

universe u open Quiver.Path local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} theorem boundary_edge_and_vertex_counts_le_boundaryVertexDefectSum_of_squareBracketPQMap (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] (p q : ℕ) (hp : 0 < p) (hq : 0 < q) (hreciprocal : (1 : ℚ) / p + 1 / q = 1 / 2) (hPQ : embedding Is[p, q]) (hcomponents : embedding.HasSimplyConnectedOrAnnularComponents) : embedding.boundaryEdgeCount ≤ ((q : ℚ) / p) * embedding.boundaryVertexDefectSum p ∧ embedding.boundaryVertexCount ≤ ((q : ℚ) / p) * embedding.boundaryVertexDefectSum p

Definition 5.6.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

A geometric edge is incident with the boundary of the planar map when one of its endpoints is a boundary vertex.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} open OneComplex local notation "GeometricEdge" => OneComplex.GeometricEdge C.skeleton def IsBoundaryLayerEdge (embedding : TwoManifoldEmbedding C 𝔼²) : GeometricEdge → Prop := Quotient.lift (embedding.orientedEdgeIsBoundaryLayer) (embedding.orientedEdgeIsBoundaryLayer_eq_of_geometricEdgeSetoid)

Theorem 5.6.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-6-3: if R is a finite relator set in the finitely generated free group FreeGroup X and R satisfies C(6), or C(4) together with T(4), or C(3) together with T(6), then the quotient by the normal closure of R has solvable word problem. This is stated through the canonical presentation owner HasSolvableWordProblem R, and its hypotheses are stated directly in the chapter owner predicates C(p)[basis, R] and T(q)[basis, R]. The source assumption that R is symmetrized is absorbed by these owners: C(p) is stated on the symmetrized relator family, while T(q) is stated on the corresponding symmetrized relator set transported back to actual relators through the canonical basis basis.

universe u open FreeGroupBasis variable {X : Type u} namespace GroupPresentation -- Proof sketch: use the small-cancellation hypotheses to derive the Dehn-type reduction or -- maximum-principle criterion for null-homotopic loops in the Cayley complex of the presentation. -- Because `X` and `R` are finite, that criterion yields an effective procedure deciding whether a -- signed word lies in the normal closure of `R`, which is exactly `HasSolvableWordProblem R`. theorem hasSolvableWordProblem_of_finite_smallCancellation (R : Set (FreeGroup X)) [Finite X] [Primcodable X] (hR : R.Finite) (hcase : C(6)[basis, R] ∨ (C(4)[basis, R] ∧ T(4)[basis, R]) ∨ (C(3)[basis, R] ∧ T(6)[basis, R])) : HasSolvableWordProblem R

5.7 — The Conjugacy Problem

Lemma 5.7.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-7-1: if a [p, q] map admits a finite connected-component decomposition in which each component is either simply connected or annular in the induced restricted embedding, then the curvature sum ∑_M [p / q + 2 - d(v)], represented here by boundaryVertexAdjustedDefectSum p q, is nonnegative.

universe u open Quiver.Path local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} variable (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] -- Proof sketch: choose the component decomposition from -- `embedding.HasSimplyConnectedOrAnnularComponents`. Apply the simply connected -- boundary-curvature estimate to each disc component and the annular analogue to each annular -- component via the owner restricted embedding `embedding.restrictToSubcomplex (components i)`. -- Summing the componentwise inequalities gives a nonnegative total boundary curvature sum for the -- whole map. theorem boundaryVertexAdjustedDefectSum_nonnegative_of_hasSimplyConnectedOrAnnularComponents (p q : ℕ) (hp : 0 < p) (hq : 0 < q) (hreciprocal : (1 : ℚ) / p + 1 / q = 1 / 2) (hPQ : embedding Is[p, q]) (hcomponents : embedding.HasSimplyConnectedOrAnnularComponents) : 0 ≤ embedding.boundaryVertexAdjustedDefectSum p q

Lemma 5.7.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-7-2: deleting the boundary layers on both sides of a dual pair yields deleted maps that are again dual. Corollary_5_7_3 already provides the exact source-facing deleted-map witness with the canonical owner abstraction and without an existential wrapper, so this file keeps only a direct recall of that declaration instead of a parallel local copy.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C K : TwoComplex} variable (embedding : TwoManifoldEmbedding C 𝔼²) (dualEmbedding : TwoManifoldEmbedding K 𝔼²) variable (hdual : embedding.IsDualMap dualEmbedding) variable (S : Subcomplex C) (hS : S.IsBoundaryLayerComplement embedding) variable (SStar : Subcomplex K) (hSStar : SStar.IsBoundaryLayerComplement dualEmbedding) #check (boundaryLayerRemoval_preservesDuality embedding dualEmbedding hdual S hS SStar hSStar : (embedding.restrictToSubcomplex S).IsDualMap (dualEmbedding.restrictToSubcomplex SStar))

Corollary 5.7.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Corollary 5-7-3: if embedding is a map, dualEmbedding is a dual of it, and S, SStar are the carried complements of the boundary layers of embedding and dualEmbedding, then the induced deleted maps on S and SStar are themselves dual. The restricted ambient TwoComplex.Duality is the underlying field of this source-facing dual-map witness.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C K : TwoComplex} def boundaryLayerRemoval_preservesDuality (embedding : TwoManifoldEmbedding C 𝔼²) (dualEmbedding : TwoManifoldEmbedding K 𝔼²) (hdual : embedding.IsDualMap dualEmbedding) (S : Subcomplex C) (hS : S.IsBoundaryLayerComplement embedding) (SStar : Subcomplex K) (hSStar : SStar.IsBoundaryLayerComplement dualEmbedding) : (embedding.restrictToSubcomplex S).IsDualMap (dualEmbedding.restrictToSubcomplex SStar)

Theorem 5.7.4

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-7-4: if M is a (q, p) map whose connected components are each simply connected or annular, and M = M₀, M₁, ..., M_k is a sequence obtained by repeatedly deleting the boundary layer until the last map is equal to its own boundary layer, then the maximum of the source quantities β(Mᵢ) is bounded by (q / p) σ'(M), represented here by ((q : ℚ) / p) * σ'(embedding)[p, q] for the initial map and by maxBoundaryRegionCount for the boundary-layer deletion sequence.

local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} theorem maxBoundaryRegionCount_le_scaled_sigmaPrime_of_boundaryLayerDeletionSequence (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] (p q : ℕ) (hp : 0 < p) (hq : 0 < q) (hQP : embedding Is(q, p)) (hreciprocal : (1 : ℚ) / p + 1 / q = 1 / 2) (hcomponents : embedding.HasSimplyConnectedOrAnnularComponents) {k : ℕ} (sequence : BoundaryLayerDeletionSequence embedding k) : sequence.maxBoundaryRegionCount ≤ ((q : ℚ) / p) * σ'(embedding)[p, q]

Lemma 5.7.5

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-7-5: if A is an annular planar map, H is the subcomplex obtained from A by removing the boundary layer and its gaps, and A has no boundary linking pairs, then the induced restricted embedding of H.complex is again annular and has at least one region.

open Quiver.Path open OneComplex local notation "𝔼²" => EuclideanSpace ℝ (Fin 2) namespace TwoComplex namespace TwoManifoldEmbedding variable {C : TwoComplex} theorem boundaryLayerGapRemoval_has_annularRealizationWithRegion_of_no_boundaryLinkingPairs (embedding : TwoManifoldEmbedding C 𝔼²) [embedding.IsPlanarMap] (σ τ : CyclicPath C.skeleton) (hannular : embedding.HasAnnularBoundaryCycles σ τ) (H : Subcomplex C) (hH : H.IsBoundaryLayerGapRemoval embedding) (hnoLink : ∀ D₁ D₂ : GeometricFace C, ¬ embedding.IsBoundaryLinkingPair σ τ D₁ D₂) : ∃ σH τH : CyclicPath H.complex.skeleton, (embedding.restrictToSubcomplex H).HasAnnularBoundaryCycles σH τH ∧ Nonempty (GeometricFace H.complex)

Theorem 5.7.6

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-7-6: if R is a finite relator set in the finitely generated free group FreeGroup X and R satisfies C(6), or C(4) together with T(4), or C(3) together with T(6), then the quotient by the normal closure of R has solvable conjugacy problem. The source assumption that R is symmetrized is absorbed by the Chapter 5 owner predicates C(p)[basis, R] and T(q)[basis, R]: C(p) is stated directly on the symmetrized relator family, while T(q) is stated on the corresponding symmetrized relator set transported back to actual relators through the canonical basis basis.

universe u open FreeGroupBasis variable {X : Type u} namespace GroupPresentation theorem hasSolvableConjugacyProblem_of_finite_smallCancellation (R : Set (FreeGroup X)) [Finite X] [Primcodable X] (hR : R.Finite) (hcase : C(6)[basis, R] ∨ (C(4)[basis, R] ∧ T(4)[basis, R]) ∨ (C(3)[basis, R] ∧ T(6)[basis, R])) : HasSolvableConjugacyProblem R

5.8 — Applications to Knot Groups

Lemma 5.8.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-8-1: the canonical presentation attached to a standard alternating projection of a knot is finite and satisfies C(4) and T(4).

universe u v open FreeGroupBasis namespace Knot theorem presentation_of_standard_alternating_projection (K : Knot.{u, v}) {P : K.Projection} (hP_standard : K.IsStandardProjection P) (hP_alternating : K.IsAlternatingProjection P) : (K.presentationRelators P).Finite ∧ C(4)[ofFreeGroup (Fin (K.presentationRank P)), K.presentationRelators P] ∧ T(4)[ofFreeGroup (Fin (K.presentationRank P)), K.presentationRelators P]

Lemma 5.8.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-8-3 (1): every knot admits a projection that is standard.

universe u v namespace Knot theorem exists_standard_projection (K : Knot.{u, v}) : ∃ P : K.Projection, K.IsStandardProjection P

Lemma 5.8.4

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-8-4 (1): if R is a finite relator set in the finitely generated free group FreeGroup X and R satisfies C(4) and T(4), then the quotient by the normal closure of R has solvable word problem. The source phrase “T(4) for minimal sequences” is recorded here through the Chapter 5 owner predicate T(4)[basis, R].

universe u open FreeGroupBasis variable {X : Type u} namespace GroupPresentation variable (R : Set (FreeGroup X)) [Finite X] [Primcodable X] -- Proof sketch: specialize the Chapter `5` small-cancellation word-problem theorem to the -- `(C(4), T(4))` case. The source knot-group assumptions that `R` is symmetrized and all relators -- have length four are bookkeeping for this application, while the decision-problem conclusion is -- expressed directly through the canonical owner predicates `C(4)[basis, R]` and `T(4)[basis, R]`. theorem hasSolvableWordProblem_of_finite_C4_T4 (hR : R.Finite) (hC4 : C(4)[basis, R]) (hT4 : T(4)[basis, R]) : HasSolvableWordProblem R

Theorem 5.8.5

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-8-5: if K is an alternating knot, then the group of K has solvable conjugacy problem.

universe u v namespace Group open Knot theorem hasSolvableConjugacyProblem_of_alternatingKnot {K : Knot.{u, v}} (hK : K.IsAlternating) : HasSolvableConjugacyProblem K.knotGroup

5.9 — The Theory over Free Products

Lemma 5.9.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-9-1: if R satisfies C'(\lambda) and r ∈ R has a semi-reduced factorization r = b₁ ⋯ bⱼ c with each bᵢ a piece, then any maximal consecutive subword cPrime of the tail c that still appears consecutively in the normal form of a cyclically reduced conjugate rPrime of r has length strictly greater than (1 - j \lambda) |rPrime|, where j is the number of piece factors.

universe u v variable {X : Type u} {F : Type v} [Group F] namespace FreeGroupBasis -- Proof sketch: use the `C'(\lambda)` estimate on each piece in the initial semi-reduced -- factorization, compare the total deleted length with `j * \lambda * |r'|`, and use maximality -- of `cPrime` together with the cyclically reduced conjugacy of `rPrime` to identify the -- surviving consecutive letters of the tail `c` inside `rPrime.toWord`. theorem maximal_common_consecutive_subword_length_gt_of_condition_c_prime (basis : FreeGroupBasis X F) (R : Set F) (lambda : ℝ) {r : F} (hR : C'(lambda)[basis, R]) (hr : r ∈ R) (j : ℕ) {pieces : Vector (List (SignedLetter X)) j} {c : List (SignedLetter X)} (hpieces : ∀ piece ∈ pieces.toList, basis.is_piece R piece) (hfactor : (basis.repr r).toWord = pieces.toList.flatten ++ c) {rPrime : FreeGroup X} (hconj : IsConj rPrime (basis.repr r)) (hrPrime_cyclic : FreeGroup.IsCyclicallyReduced rPrime.toWord) {cPrime : List (SignedLetter X)} (hcPrime_left : cPrime <:+: c) (hcPrime_right : cPrime <:+: rPrime.toWord) (hcPrime_max : ∀ ⦃t : List (SignedLetter X)⦄, t <:+: c → t <:+: rPrime.toWord → t.length ≤ cPrime.length) : (cPrime.length : ℝ) > (1 - (j : ℝ) * lambda) * (rPrime.toWord.length : ℝ)

Lemma 5.9.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-9-2 (1): any subset of a Noetherian topological space, with the induced topology, is Noetherian.

universe u variable (X : Type u) [TopologicalSpace X] #check TopologicalSpace.NoetherianSpace.set

Theorem 5.9.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-9-3: if the relator set R satisfies C'(1 / 8) with respect to basis, and no relator in R is trivial, then R satisfies Condition J; equivalently, no relator in R is conjugate to its inverse.

universe u v variable {X : Type u} {F : Type v} [Group F] namespace FreeGroupBasis -- Proof sketch: if some `r ∈ R` were conjugate to `r⁻¹`, pass to a cyclically reduced conjugate -- of `basis.repr r`, compare it with the inverse cyclic word, and apply the Section `9` -- one-eighth overlap estimate to obtain a common subword longer than half the relator length. -- The nontriviality hypothesis rules out the degenerate empty-word case, and the overlap bound -- contradicts the `C'(1 / 8)` piece inequality. theorem satisfiesConditionJ_of_condition_c_prime_one_eighth (basis : FreeGroupBasis X F) (R : Set F) (hR : C'((1 / 8 : ℝ))[basis, R]) (hrel_ne_one : ∀ ⦃r : F⦄, r ∈ R → r ≠ 1) : J[R]

Definition 5.9.7

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-9-7: Condition J for a relator set R says that no relator in R is conjugate to its inverse.

universe u variable {F : Type u} [Group F] namespace Set def SatisfiesConditionJ (R : Set F) : Prop := ∀ ⦃r : F⦄, r ∈ R → ¬ IsConj r r⁻¹

5.10 — Small Cancellation Products

Theorem 5.10.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-10-1: for a free product F = *ᵢ Aᵢ and a relator set R satisfying the Section 10 free-product small-cancellation hypothesis C'(1 / 8) on the symmetrized family of weakly cyclically reduced relators generated by R, every finite-order element of the quotient G = F / N is either conjugate to the image of a finite-order element from one of the factors Aᵢ, or comes from a relator r ∈ R that is a genuine proper power r = root ^ n with 1 < n, with the quotient element conjugate to a power of the image of root in G. Via freeGroupEquivCoprodI, this also covers the free-group case in the textbook statement.

universe u v open Monoid variable {ι : Type u} {A : ι → Type v} variable [∀ i, Group (A i)] namespace Monoid.CoprodI open Word notation:55 "C'(" lambda ")[" R "]" => condition_c_prime R lambda -- Proof sketch: apply the defining predicate `condition_c_prime` to the displayed symmetrized -- relator `q`, its cyclic occurrence of `piece`, and the assumption that `piece` is a piece. variable (R : Set (CoprodI A)) local notation "F" => CoprodI A local notation "N" => Subgroup.normalClosure R local notation "G" => F ⧸ N local notation "v" => QuotientGroup.mk' N -- Proof sketch: choose a shortest conjugate representative of a preimage of the finite-order -- quotient element and analyze its powers using the Section `10` free-product small-cancellation -- theory under `C'(1 / 8)`. If the representative already has finite order in the ambient free -- product, the free-product torsion theorem conjugates it into one factor, giving alternative -- `(i)`. Otherwise the reduction analysis forces a defining relator to be a genuine proper power -- `r = root ^ n` with `1 < n`, and the same normal-form argument shows that the original quotient -- element is conjugate to an integral power of the image of `root`. theorem torsion_theorem_for_small_cancellation_quotient (hR : C'((1 / 8 : ℝ))[R]) (w : G) (hw : IsOfFinOrder w) : (∃ i : ι, ∃ a : A i, IsOfFinOrder a ∧ IsConj (v (of a)) w) ∨ ∃ r ∈ R, ∃ root : F, ∃ n : ℕ, 1 < n ∧ r = root ^ n ∧ ∃ m : ℤ, IsConj ((v root) ^ m) w

Lemma 5.10.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-10-2: if the relator set R in the free product F = *ᵢ Aᵢ satisfies the Section 10 small-cancellation hypothesis C'(1 / 6), then the natural map from each factor Aᵢ into the quotient F / ⟪R⟫ is injective.

universe u v open Monoid variable {ι : Type u} {A : ι → Type v} variable [∀ i, Group (A i)] namespace Monoid.CoprodI local notation "F" => CoprodI A theorem factor_injective_in_quotient_of_condition_c_prime {R : Set F} (hR : C'((1 / 6 : ℝ))[R]) (i : ι) : Function.Injective ((QuotientGroup.mk' (Subgroup.normalClosure R)).comp (of : A i →* F))

Theorem 5.10.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-10-3: if X and Y are nontrivial and are not both of order 2, then the free product X ∗ Y is SQ-universal.

universe u v open scoped Monoid.Coprod variable {X : Type u} [Group X] [Nontrivial X] variable {Y : Type v} [Group Y] [Nontrivial Y] -- Layer triage: -- `source-facing`: a nontrivial free product `X ∗ Y` together with the exclusion of the unique -- exceptional case where both factors have order `2`. -- `core/canonical`: `Monoid.Coprod` with notation `X ∗ Y` is mathlib's owner abstraction for the -- free product of two groups, `Nat.card` is the project's standard universe-stable notion of group -- order, and `IsSQUniversal` from Proposition `2-5-31` is the owner predicate for -- `SQ`-universality. -- `bridge/view`: no auxiliary bridge is needed here, since the textbook statement is already -- about the canonical free-product owner object itself. -- Domain sampling: -- 1. `Monoid.Coprod` with notation `X ∗ Y` is the existing project-wide owner for free products -- of two groups. -- 2. Proposition `2-5-31` introduces `IsSQUniversal` as the canonical subquotient-universality -- predicate on a group. -- 3. Proposition `3-12-3` records the same exceptional branch for free products by the condition -- `¬ (Nat.card G₁ = 2 ∧ Nat.card G₂ = 2)`, so that is the established project encoding of -- “not the free product of two copies of `C₂`”. -- Primitive vs. derived: -- the primitive public data are only the two nontrivial factor groups and the exclusion of the -- order-`2`/order-`2` case; `SQ`-universality is the resulting owner-level property of the free -- product. -- Proof sketch: use the classical small-cancellation construction from the textbook. For any -- countable group `H`, adjoin `H` as a free factor to `X ∗ Y`, impose the symmetrized relators -- from the theorem text to obtain a `C'(1/10)` small-cancellation quotient, and apply the -- embedding theorem for factors in a small-cancellation product. The quotient is generated by the -- images of `X` and `Y`, hence is a quotient of `X ∗ Y`, so `H` embeds in a quotient of a -- subgroup of `X ∗ Y`. theorem isSQUniversal_freeProduct_of_nontrivial_of_not_both_order_two (hnot_order_two : ¬ (Nat.card X = 2 ∧ Nat.card Y = 2)) : IsSQUniversal (X ∗ Y)

Theorem 5.10.4

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-10-4: every countable group embeds in a two-generator complete Hopfian group. If H is finitely presented, then the ambient group can be chosen finitely presented. If H has no elements of order 5, then the ambient group can be chosen cohopfian.

universe u variable (H : Type u) [Group H] [Countable H] theorem exists_embedding_into_two_generator_complete_hopfian_group : ∃ (G : Type u) (_ : Group G) (_ : Group.FG G) (f : H →* G), Function.Injective f ∧ Group.rank G ≤ 2 ∧ IsHopfian G ∧ (Subgroup.center G = ⊥ ∧ MulAut.innerAutomorphismSubgroup G = ⊤) ∧ (Group.IsFinitelyPresented H → Group.IsFinitelyPresented G) ∧ ((∀ h : H, orderOf h ≠ 5) → IsCohopfian G)

Theorem 5.10.5

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-10-5: for the explicit quotient built from the relators r₀, r₁, ... attached to the chosen enumeration of H, the relator set satisfies C'(1 / 10), the quotient is a quotient of C₅ * C₇, the factors H, C₅, and C₇ embed, the quotient is Hopfian and complete, and it is cohopfian whenever H has no element of order 5.

universe u open scoped Monoid.Coprod namespace SmallCancellationProduct variable {H : Type u} [Group H] open Monoid.CoprodI open MulEquiv local notation "F" => Monoid.CoprodI (Factors H) variable (enumerate : ℕ → H) -- Proof sketch: this is the standard overlap computation from the textbook. The block structure -- forces every common piece to be shorter than one tenth of a relator. theorem quotient_properties (henum : Function.Surjective enumerate) : C'((1 / 10 : ℝ))[relatorSet enumerate] ∧ Function.Surjective (cyclicFactorsHom enumerate) ∧ Function.Injective (embedH enumerate) ∧ Function.Injective (embedC5 enumerate) ∧ Function.Injective (embedC7 enumerate) ∧ IsHopfian (quotient enumerate) ∧ (Subgroup.center (quotient enumerate) = ⊥ ∧ JA(quotient enumerate) = ⊤) ∧ ((∀ h' : H, orderOf h' ≠ 5) → IsCohopfian (quotient enumerate))

5.11 — Small Cancellation Theory over Free Products with Amalgamation and HNN Extensions

Definition 5.11.1

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-11-1: if w ≠ 1 is an element of the free product of G and H with A and B amalgamated along e, then a normal form of w is a reduced alternating word in the two factors whose product is w. The length-one exception for letters lying in the amalgamated part is already built into Reduced e.

universe u v open Monoid.CoprodI open Monoid.PushoutI open Monoid.PushoutI.NormalWord namespace Subgroup.amalgamatedProductAlong variable {G : Type u} {H : Type v} [Group G] [Group H] variable {A : Subgroup G} {B : Subgroup H} (e : A ≃* B) local notation "Family" => Monoid.CoprodI.twoFactorFamily G H local notation "Maps" => Monoid.CoprodI.twoFactorAmalgamatingMaps A.subtype (B.subtype.comp e.toMonoidHom) local notation "W" => Word Family local notation "P" => Subgroup.amalgamatedProductAlong e -- Primary domain: normal forms in free products with amalgamation. -- Layer triage: -- `source-facing`: a normal form of one fixed nontrivial element in the amalgamated product, -- recorded as a reduced alternating word that evaluates to that element. -- `core/canonical`: `Subgroup.amalgamatedProductAlong e` for the ambient amalgamated product, -- `Subgroup.amalgamatedProductAlong.Reduced` for the source reduced-word predicate, -- `Subgroup.amalgamatedProductAlong.ofWord e` for evaluation of a word in the amalgamated -- product, and `Monoid.PushoutI.syllableLength` for the chosen-transversal owner length on the -- ambient pushout. -- `bridge/view`: the textbook phrase “normal form of `w`” is the conjunction that the source -- reduced word represents the chosen element `w`; no second word owner is introduced. -- -- Domain sampling: -- 1. `Subgroup.amalgamatedProductAlong e` is the chapter-facing owner for a two-factor -- amalgamated product. -- 2. `Subgroup.amalgamatedProductAlong.Reduced e` already encodes exactly the source condition -- that successive syllables come from different factors and that syllables from the -- amalgamated subgroup are allowed only in the length-one case. -- 3. `Subgroup.amalgamatedProductAlong.ofWord e` is the canonical evaluation map from a reduced -- factor word to the represented element of the amalgamated product. -- 4. `Monoid.PushoutI.syllableLength` is the chapter's canonical owner for the chosen-transversal -- normal-form length, and the source length here differs from it only by the length-one -- convention for nontrivial elements of the amalgamated subgroup. -- -- Primitive vs. derived: -- the primitive public inputs are the amalgamating identification `e`, the represented element -- `w : P`, and the reduced word `u : W`. The textbook notion “`u` is a normal form of `w`” is the -- derived conjunction that `u` is reduced and evaluates to `w`, together with the source -- restriction `w ≠ 1`. The uniqueness-of-length statement is derived through the canonical owner -- `syllableLength`, not by a second parallel length API. def IsNormalForm (w : P) (u : W) : Prop := w ≠ 1 ∧ Reduced e u ∧ ofWord e u = w

Definition 5.11.2

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-11-2 (1): for a normal-form word, “cyclically reduced” is the existing chapter owner predicate Monoid.CoprodI.Word.IsCyclicallyReduced, specialized to the amalgamating maps ιA and ιB.

universe u open Monoid.CoprodI Monoid.PushoutI open Subgroup.amalgamatedProductAlong namespace Subgroup.amalgamatedProductAlong variable {G H : Type u} [Group G] [Group H] variable {A : Subgroup G} {B : Subgroup H} variable (e : A ≃* B) local notation "Family" => Monoid.CoprodI.twoFactorFamily G H local notation "P" => amalgamatedProductAlong e local notation "ιA" => A.subtype local notation "ιB" => B.subtype.comp e.toMonoidHom #check (Word.IsCyclicallyReduced ιA ιB) -- Proof sketch: this is the length-`0`/length-`1` branch built directly into the definition.

Definition 5.11.3

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-11-3: the symmetrized relator family generated by R in the amalgamated product P = ⟨G * H, A = B⟩ consists of the weakly cyclically reduced words whose value in P is conjugate to some relator from R or to its inverse.

universe u open Monoid.CoprodI namespace Subgroup.amalgamatedProductAlong variable {G : Type u} {H : Type u} [Group G] [Group H] variable {A : Subgroup G} {B : Subgroup H} {e : A ≃* B} local notation "Family" => Monoid.CoprodI.twoFactorFamily G H local notation "W" => Word Family local notation "P" => Subgroup.amalgamatedProductAlong e def symmetrizedRelatorFamily (R : Set P) : Set W := { q | IsWeaklyCyclicallyReduced e q ∧ ∃ r ∈ R, IsConj (ofWord e q) r ∨ IsConj (ofWord e q) r⁻¹ }

Definition 5.11.8

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-11-8: {x₁, x₂} is a blocking pair for the subgroup A ≤ H if x₁ ≠ x₂, neither x₁ nor x₂ belongs to A, and for every nontrivial a ∈ A, every choice of y, z ∈ {x₁, x₂}, and every signs ε, δ : ℤˣ, the element y ^ (ε : ℤ) * a * z ^ (δ : ℤ) does not belong to A.

universe u variable {H : Type u} [Group H] namespace Subgroup structure IsBlockingPair (A : Subgroup H) (x₁ x₂ : H) : Prop where /-- The two displayed elements of a blocking pair are distinct. -/ distinct : x₁ ≠ x₂ /-- Every element of the displayed pair lies outside the subgroup. -/ not_mem {x : H} (hx : x ∈ ({x₁, x₂} : Set H)) : x ∉ A /-- No signed sandwich `y^ε a z^δ` with nontrivial `a ∈ A`, `y, z ∈ {x₁, x₂}`, and signs `ε, δ : ℤˣ` returns to `A`. -/ blocked {a y z : H} (ha : a ∈ A) (ha_ne_one : a ≠ 1) (hy : y ∈ ({x₁, x₂} : Set H)) (hz : z ∈ ({x₁, x₂} : Set H)) (ε δ : ℤˣ) : y ^ (ε : ℤ) * a * z ^ (δ : ℤ) ∉ A

Theorem 5.11.9

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-11-9: if P = ⟨H ∗ K, A = B⟩ is a free product with amalgamation, B is a proper subgroup of K, and there is a blocking pair for A in H, then P is SQ-universal.

universe u v variable {H : Type u} {K : Type v} [Group H] [Group K] open Subgroup theorem isSQUniversal_amalgamatedProductAlong_of_exists_blockingPair (A : Subgroup H) (B : Subgroup K) (e : A ≃* B) (hB_proper : B < ⊤) (hblocking : ∃ x₁ x₂, A.IsBlockingPair x₁ x₂) : IsSQUniversal (amalgamatedProductAlong e)

Corollary 5.11.10

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Corollary 5-11-10: if H is free, A ≤ H is finitely generated of infinite index, and K contains a proper subgroup A' isomorphic to A, then the amalgamated free product (H * K; A = A') is SQ-universal.

universe u v w variable {H : Type u} {K : Type v} [Group H] [Group K] [IsFreeGroup H] open Subgroup theorem isSQUniversal_amalgamatedProductAlong_of_freeGroup_subgroup_fg_of_infiniteIndex (A : Subgroup H) (hA_fg : A.FG) (hA_quotient : Infinite (H ⧸ A)) (A' : Subgroup K) (hA'_proper : A' < ⊤) (e : A ≃* A') : IsSQUniversal (amalgamatedProductAlong e)

Definition 5.11.11

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Definition 5-11-11: the relator set R satisfies the Section 11 small-cancellation condition C'(λ) when every piece prefix of a symmetrized relator from R has length strictly less than λ times the relator length, and every symmetrized relator has length strictly greater than 1 / λ.

universe u open Monoid.CoprodI namespace Subgroup.amalgamatedProductAlong variable {G : Type u} {H : Type u} [Group G] [Group H] variable {A : Subgroup G} {B : Subgroup H} {e : A ≃* B} local notation "Family" => Monoid.CoprodI.twoFactorFamily G H local notation "W" => Word Family local notation "P" => Subgroup.amalgamatedProductAlong e def condition_c_prime (R : Set P) (lambda : ℝ) : Prop := (∀ {q : W} (_ : q ∈ symmetrizedRelatorFamily R) {piece : List (Σ b, Family b)} (_ : piece <+: q.toList) (_ : is_piece R piece), (piece.length : ℝ) < lambda * q.toList.length) ∧ ∀ {q : W} (_ : q ∈ symmetrizedRelatorFamily R), (1 / lambda : ℝ) < q.toList.length

Lemma 5.11.15

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Lemma 5-11-15: a one-relator presentation with at least two generators is equivalent to one on the same generator type whose relator has exponent sum 0 in some distinguished generator.

universe u variable {X : Type u} lemma exists_equivalent_presentation_with_zero_generatorExponentSum (r : FreeGroup X) (hX : 2 ≤ Cardinal.mk X) : ∃ (r' : FreeGroup X) (_ : PresentedGroup ({r} : Set (FreeGroup X)) ≃* PresentedGroup ({r'} : Set (FreeGroup X))) (t : X), generatorExponentSum t r' = 0

Theorem 5.11.16

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

Theorem 5-11-16: a group with a one-relator presentation on at least three generators is SQ-universal. This item adds no new public declaration beyond the canonical theorem isSQUniversal_of_three_generator_one_relator_group; the Chapter V HNN-extension argument is a proof route for that already existing owner-level statement.

universe u variable {X : Type u} #check (isSQUniversal_of_three_generator_one_relator_group : (r : FreeGroup X) → (hX : 3 ≤ Cardinal.mk X) → IsSQUniversal (PresentedGroup (Set.singleton r)))